Review of Power Market Optimization Strategies Based on Industrial Load Flexibility

Abstract

New power systems, predominantly based on renewable energy, necessitate active load-side management to effectively alleviate the pressures associated with balancing supply-side fluctuations and demand-side energy requirements. Concurrently, as power markets continue to evolve, both the energy market and ancillary services market offer valuable guidance for the optimal economic dispatch of industrial loads. Although substantial energy-saving potential exists within industrial production processes, their inherent complexity, dynamic nature, and mixed continuous–discrete modal characteristics present significant challenges in achieving accurate and efficient demand-side response. Conversely, the ongoing advancement of industrial internet techniques lays a robust technical foundation for the reliable, stable, and economically efficient operation of new power systems with large-scale industrial load response. This paper starts from the industrial load, discusses the resources and advantages and disadvantages of the industry itself, and carefully distinguishes the advantages and disadvantages of participating in the power market to make decisions. This paper provides a comprehensive review of intelligent optimization and regulation of industrial load flexibility in response to new power systems. Firstly, it synthesizes the three prevalent demand response strategies (load shedding, load shifting, and load substitution), along with their associated regulatory techniques, considering the operational characteristics of various industrial sectors. It then examines the trading strategies and modeling challenges of flexible industrial loads within two power market environments: the energy market and the ancillary services market. Subsequently, using the non-ferrous industry electrolytic process as a case study, it explores the optimization of production process parameters under energy usage planning. Finally, from the perspectives of market, technical innovation, and stakeholder engagement, it highlights the unresolved issues and provides insights into future research directions concerning the intelligent, digital, and market-driven integration of flexible industrial load flexibility.

1. Introduction

1.1. Background

New power systems with renewable energy as the primary supply source feature interactive dynamics between “generation–grid–load–storage” and multi-energy complementarity, making them an integral part of the new energy framework and a vital enabler of achieving the “dual carbon” goals. In China, the generation rate of renewable energy in total electricity has exceeded 13%, with the installed capacity of renewable energy-based power generation reaching 1.516 billion kilowatts, accounting for 51.9% of the national total power generation capacity, surpassing that of fossil fuels, which lays a solid foundation for the industrial green transformation [1,2]. However, the integration of high-proportion renewable energy into the power grid presents formidable challenges in maintaining supply–demand balance within the power system. In the power market, China has continued to deepen power market reforms, gradually expanding the market scope, while the scale of market transactions has rapidly grown, further advancing the degree of market liberalization. In 2023, the total electricity traded in the national market reached 5.67 trillion kWh, marking a 7.9% year-on-year increase and accounting for 61.4% of total national electricity consumption [3]. The number of registered market participants across electricity trading platforms totaled 743,000, a 23.9% increase from the previous year, signaling the emergence of a diversified competitive market structure [3]. As the national unified power market system continues to be developed, it provides the necessary conditions for industrial loads to effectively engage in demand-side response through market participation.

In January 2024, the National Development and Reform Commission, the Nation-al Bureau of Statistics, and the National Energy Administration jointly issued the “Notice on strengthening the connection between green power certificates and energy conservation and carbon reduction policies to vigorously promote non-fossil energy consumption”, which emphasizes the encouragement of renewable energy consumption commitments for new projects [4]. The notice also accelerates the establishment of mandatory renewable energy consumption mechanisms for energy-intensive industries, with specific regions, such as those in the electrolytic aluminum sector, requiring renewable energy consumption ratios of up to 70%. In October 2024, the Ministry of Industry and Information Technique of China published the “Industrial Internet and Power Sector Integration Application Reference Guide (2024)”, ref. [5], highlighting the necessity of integrating the industrial internet with the power sector, and emphasizing the transition of users from passive electricity consumers to active participants in the energy services ecosystem.

Due to high electricity consumption, reliable control systems, advanced automation, and stable communication techniques, industrial loads have emerged as a crucial provider of demand-side flexibility. Energy-intensive, high-carbon industries, such as steel plants, wastewater treatment plants, electrolytic aluminum plants, cement factories, and pulp and paper mills, possess significant potential to provide flexible load [6]. The adjustable potential of energy-intensive industries is estimated at approximately 108 million kilowatts, accounting for 19.3% of the electricity supply–demand gap [7]. Effective utilization and regulation of industrial load flexibility can help to mitigate energy supply fluctuations in new power systems, enhance the dynamic matching of electricity supply and demand, and improve energy demand flexibility, thus generating economic benefits through market participation.

To illustrate the sources of adjustable potential, this section examines production processes of two typical energy-intensive industries—steel and cement. The distribution of various load types in steel production is presented in

Table 1. Production load of steel industry.

Load Type	Devices	Electricity Load Proportion
Main production loads	Electric arc furnace, blast furnace, converter, continuous casting machine, sintering machine, etc.	Exceeding 75%
Auxiliary production load	Water pump, hydraulic pump, axial fan, etc.	About 8%
Protective load	Exhaust gas and dust recovery suction fan, circulating cooling water pump, public protection electric equipment, etc.	Exceeding 10%

, where the controllable load accounts for 20% of total consumption in steel plants [8]. The load distribution of key production equipment in the cement industry is shown in

Table 2. Production Load of Cement Industry.

Load Type	Devices	Electricity Load Proportion
Main production loads	Raw material mill, cement mill, ball mill, etc.	55–60%
Auxiliary production load	Drive fan, drive belt motor, etc.	15–20%
Protective load	Cooling water pump, lubricating oil pump, etc.	8–15%
Non-production load	Office electrical equipment, central air conditioning, etc.	2–5%

. The cement production cycle is not constrained by technical processes, but rather aims to enhance production efficiency through continuous operation. Consequently, the production cycle in the cement industry is flexible, and the industry possesses interruptible load potential. With a flexible production schedule, controllable loads can account for approximately 24% of total electricity consumption in cement plants [8]. These data demonstrate the energy-saving and consumption-reduction potential of adjustable loads in two energy-intensive industrial processes.

Power markets provide an environment that enables industrial enterprises to harness flexibility, using market mechanisms to reduce energy costs. There are two primary forms of industrial participation in power markets:

• Direct participation: Industrial loads, as independent load entities, directly engage in power markets [9]. This includes integrating other loads or energy storage systems within industrial parks, in which industrial production systems are integrated with other energy systems—such as industrial parks [10] or microgrids [11]—to participate in power markets.
• Indirect participation via intermediaries: Industrial loads participate in the market through the aggregation of flexibility resources from various enterprises and energy systems, facilitated by intermediaries such as aggregators [12] and demand response providers [13]. This can take the form of virtual power plants (VPPs) [14].

With the continuous development of renewable energy generation and energy storage techniques, the coordination between renewable power plants, storage units, and conventional thermal power plants has gained increasing attention from industry stakeholders.

In recent years, substantial research has been conducted on the flexibility of power systems both domestically and internationally. However, there remains no unified definition. There are several widely cited definitions of power system flexibility, as shown in

Table 3. Definition of power system flexibility.

Reference	Year	Source	Definition
[15]	2008	International Energy Agency	The ability of the power system to quickly respond to large power and energy fluctuations on both sides of supply and demand under certain economic cost constraints.
[16]	2008	North American Electric Reliability Corporation	The ability to use system resources to meet load changes.
[17]	2011	Paul Denholm	General characteristics of the system’s ability to respond to changes and uncertainties in load.
[18]	2012	Eamonn Lannoye	The ability of the system to deploy its resources in response to changes in load.
[19]	2013	Holttinen H	Adaptive loads produce variability and uncertainty in the balance, while maintaining a satisfactory level of performance on any timescale.
[20]	2016	Zhao J	Under the constraints of time and cost, the maximum adaptability of the power system to uncertainty fluctuations.
[21]	2022	China Power Roundtable project research group	The ability of various resources of the power system to rapidly change their own power generation characteristics to maintain the active power balance of the system.

. Power system flexibility encompasses generation-side flexibility, grid-side flexibility, and load-side flexibility. No clear definition exists for industrial load flexibility on the load side. In this paper, we define industrial load flexibility as the industrial load’s ability to adjust resource utilization within a given timeframe to meet the power system supply–demand balance, subject to constraints such as economic considerations, production plans, and operational conditions. In this context, terms such as “adjustment capability” and “demand response”, which refer to load adjustments based on inherent characteristics to shift energy consumption periods, are collectively referred to as flexibility.

In summary, the rapid evolution and critical demand for new power systems, coupled with the steady advancement of power market reforms and the large-scale deployment of industrial internet techniques, have provided the technical foundation and infrastructural platform essential for the intelligent optimization and regulation of industrial production processes in response to power system requirements. These developments create the preconditions for the large-scale implementation of industrial load demand-side response. The current literature tends to discuss load flexibility to ensure stable operation of power grids from the perspective of power grids, and less from the perspective of industry. This paper starts from the industrial load itself, discusses the resources and advantages and disadvantages of the industry itself, and carefully distinguishes the advantages and disadvantages of participating in the power market to make decisions. This paper systematically reviews industrial load participation in demand-side response within the context of emerging power systems from three dimensions:

• Mechanisms for flexibility provision in industrial processes and their sector-specific applications: Industrial load flexibility extends beyond electricity demand management to encompass the modulation of key process parameters in industrial operations to accommodate demand fluctuations in power market. Different industrial processes, depending on their operational characteristics, can deliver distinct types of flexibility. Through an in-depth analysis of the sources and application contexts of flexibility within various industrial processes, this dimension establishes the theoretical foundation for leveraging flexibility resources in the formulation of market participation strategies for industrial loads.
• Market participation strategies for industrial enterprises in the electricity and ancillary services markets: Building on a comprehensive understanding of the sources and capacities of flexibility within industrial processes, enterprises formulate and optimize trading strategies to ensure effective market participation. This dimension focuses on the commodification of industrial load flexibility, designing trading strategies based on the type of market and dynamic demand, which involves considering both internal and external characteristics of industrial loads. The objective is to balance supply and demand in the power system, which simultaneously maximizes the economic returns for enterprises in the power market.
• Optimization and control of industrial parameters in the electrolytic industry. Taking the electrolytic industry as a case study, this section explores the optimization of energy consumption planning through the intelligent setting and regulation of process parameters within the production cycle. This approach not only enhances production efficiency, but also improves the responsiveness of the industrial process to fluctuations in power market demand, thereby demonstrating the practical value of industrial load flexibility. In this dimension, industries must consider the dynamic interplay between external characteristics and internal characteristics, with a primary focus on optimizing internal process parameters—such as current density, temperature, and concentration—ensuring the safe and stable operation of industrial devices.

1.2. Policies for Industry Participation in Demand-Side Response

The commitment to energy structure transformation and carbon neutrality in China has positioned demand-side management as a pivotal mechanism for regulating energy consumption. By encouraging consumers to modify their electricity consumption patterns, demand-side management aids in achieving load balancing, ensuring the stable and secure operation of the power system, and facilitating the integration of renewable energy sources. A well-developed policy infrastructure and comprehensive demand-side management protocols are essential foundational conditions for enabling industrial participation in power markets.

In the 1990s, China introduced the concept of “demand-side management” in the electricity sector. In 1998, China officially implemented the Energy Conservation Law of the People’s Republic of China [22], which explicitly recognized energy conservation as a long-term strategic directive for the national economic development, promoting an energy-efficient trajectory for the national economy. Simultaneously, the law encouraged the development and utilization of renewable energy. In 2000, the Electricity Saving Management Measures [23] were introduced, affirming the effectiveness of demand-side management in energy conservation and environmental protection. Subsequently, a series of regulations were issued to further promote demand-side management in the electricity sector. In the industrial domain, in 2011, the Ministry of Industry and Information Technique (MIIT) issued the Guidelines on Promoting Demand-Side Management in the Industrial Sector [24], which significantly advanced the management of industrial electricity consumption, addressing the tension between limited electricity resources and rapid industrial development. In 2016, the MIIT released the Special Plan for Demand-Side Management in the Industrial Sector (2016–2020) [25], establishing an action plan for demand-side management in industrial sectors. In 2017, the ministry issued the Notice on Advancing Supply-Side Structural Reforms and Improving Demand-Side Management in the New Context and the Implementation Plan for Addressing the Curtailed Hydropower, Wind, and Solar Energy [26,27], which outlined the need for demand-side management to not only continue its role in energy savings and emission reduction, but also to focus on advancing power system reforms, promoting the absorption of renewable energy, and enhancing demand-side responsiveness to renewable power generation. In 2022, the National Development and Reform Commission (NDRC) released the 14th Five-Year Plan for Modern Energy Systems, which emphasized the importance of increasing electricity load flexibility and advancing demonstrations of virtual power plants that aggregate various resources, including industrial adjustable loads and user-side energy storage.

In 2023, the National Development and Reform Commission (NDRC) issued the Measures for Electricity Demand-Side Management (2023 Edition) and the Measures for Electricity Load Management (2023 Edition), which established that the scale of orderly electricity usage should reach 30% of the historical peak load. These measures aim to optimize the allocation of electricity resources through feasible technical, economic, and managerial strategies, thereby enhancing the security, carbon reduction, and efficiency of the power system. The main content covers areas such as electricity conservation, demand response, green electricity usage, electricity substitution, smart electricity usage, and orderly electricity consumption. In the same year, the NDRC also issued the Notice on Further Accelerating the Development of the Electricity Spot Market (NDRC Reform [2023] No. 813), which explicitly emphasized the promotion of new entities, such as energy storage systems, virtual power plants, and load aggregators, in roles related to peak shaving, valley filling, and optimizing power quality. In October 2024, the NDRC published the Guiding Opinions on Vigorously Implementing Renewable Energy Substitution Actions (NDRC [2024] No. 1537), which highlighted the need to strengthen industrial electricity demand-side management and jointly promote the green and low-carbon transformation of industrial energy use [28]. A summary of the demand-side management regulations from 2022 to 2024 for various provinces is provided in Appendix A.

2. Ways in Which Industrial Load Provides Flexibility

Industries exhibit complex continuous–discrete dynamic characteristics, with intricate energy supply–production mechanisms. These systems are marked by multi-system, multi-energy, and multi-timescale coupling. Based on the understanding of load flexibility provision in reference [29], we categorize the methods by which industries offer flexibility into three general types, according to the distinct operational characteristics of industrial processes:

• Load Reduction: Industries proactively reduce electricity consumption by optimizing device operation processes or adjusting production schedules, achieving this reduction while maintaining production objectives or with minimal impact on them.
• Load Shifting: Industries adjust the operational timing of devices, shifting energy-intensive operations or non-essential devices to periods of low electricity demand or lower electricity prices, thereby alleviating pressure on the grid.
• Load Substitution: Industries achieve self-sufficiency in power supply by leveraging renewable energy generation systems or their own device to partially replace grid-supplied electricity.

To investigate strategies that yield optimal economic performance and enhance grid stability in the electricity market across various industrial sectors and enterprises of differing scales, conducting a quantitative comparative analysis is indispensable. For instance, Yun et al. [30] demonstrated that process optimization targeting load shifting can reduce electricity costs by 15% without adversely affecting production output. Moreover, load substitution in an industrial park integrating photovoltaic systems and combined heat and power (CHP) facilities resulted in a 20% reduction in reliance on grid electricity [31], achieving a payback period of less than eight years.

The flexibility provided by industrial loads not only contributes to the optimization of energy efficiency within industrial operations, but also plays a crucial role in supporting grid scheduling requirements. Through load shedding, load shifting, and load substitution, industrial loads can significantly contribute to both electricity energy markets and ancillary services. In the energy market, industries can reduce peak loads through load shedding and shifting, thereby assisting in balancing grid supply and demand and enhancing the economic efficiency of the power market. In the frequency regulation market, industrial loads can swiftly respond to demand fluctuations, participating in frequency control services to maintain grid frequency stability. Furthermore, industrial load flexibility enables the provision of reserve services, particularly in the event of unforeseen grid disturbances or forecasting errors. By mobilizing internal or renewable energy resources, industries can ensure grid reliability. Therefore, industrial load flexibility not only enhances operational efficiency within the industry, but also provides vital peak-shaving, frequency regulation, and reserve capabilities, thereby contributing to the stability and economic performance of the power system.

2.1. Load Shedding

In the context of demand response and industrial load flexibility, load shedding has emerged as a critical regulatory mechanism that is widely applied across various industrial sectors, particularly during situations involving electricity supply shortages or elevated electricity prices, demonstrating significant regulatory benefits. Applicable to energy-intensive industries with interruptible production processes, such as steel manufacturing and electrolytic aluminum production, this approach offers potential reductions in electricity costs. However, it may impact production continuity, necessitating careful consideration of the efficiency losses associated with adjustments to production schedules. The implementation of load shedding can typically be analyzed from multiple perspectives, including adjustments to device operational modes, the optimization of production processes, the enhancement of device energy efficiency through technical upgrades, and process control optimization. Each approach, tailored to the specific characteristics of different industrial sectors, combines advanced technical tools and optimization strategies to achieve the objectives of reducing electricity demand, improving power utilization efficiency, and enhancing the overall flexibility value. Below, a detailed discussion of several common load shedding strategies is provided.

(1)

Adjusting device operating modes

Adjusting device operating modes refers to altering or even interrupting the operation of a specific device (such as modifying operational modes, reducing load, or shutting down non-essential devices) to mitigate peak system demand. In the steel industry, for example, to address fluctuations in grid frequency, under-frequency load shedding (UFLS) serves as a load reduction method, achieved by adjusting the operating modes of the device to meet frequency regulation requirements [32]. In steel plants, the rolling process exhibits interruptible characteristics, and load reduction is achieved by interrupting operations during specific time periods [33].

(2)

Optimizing production processes

Optimizing production processes involves analyzing the industrial resource characteristics and applying advanced optimization methods to adjust and refine production workflows. By designing scientifically based production processes, energy efficiency is enhanced, leading to more efficient electricity consumption patterns. For instance, a simulation-based analysis was conducted on a typical steelmaking process [34]. By utilizing a network diagram approach, an input–operation–output model was established, revealing a 47.6% improvement in energy utilization efficiency throughout the steelmaking process. Moreover, in electric arc furnace steelmaking, the production workflow’s optimization is attained through a comprehensive analysis of energy transformation dynamics within electric arc furnaces. Optimization strategies encompass adjusting the molten steel volume [35], implementing a continuous charging methodology with preheated scrap [36], enhancing the predictive accuracy of melt durations [37], and augmenting electrode immersion [38]. Empirical findings indicate that modifying the molten steel volume yields a 16% reduction in electrical power consumption within the electric arc furnace [36].

(3)

Device and technique upgrades

Device and technique upgrades refer to enhancing the efficiency of electrical power consumption throughout the entire device lifecycle through version upgrades and technique transformations. For instance, in electric arc furnace steelmaking, the integration of ultra-high-power electric arc furnaces has improved device efficiency, thereby reducing the energy consumption per ton of steel [39]. Similarly, in wastewater treatment plants, adjusting the aeration on–off ratio has optimized the intermittent aeration process [40]. In pulp and paper mills, advancements such as the upgrade of lignin extraction [41], the adoption of double-disk refiners [42], and improvements in black liquor concentration [43] have not only enhanced energy efficiency, but also increased the production of byproducts, thereby reducing the load profile and realizing the flexibility value of the industrial load.

(4)

Optimizing process control

The optimization of process control discussed in this paper refers to utilizing advanced control techniques and algorithms to monitor and regulate the operational state, based on existing devices and techniques. This ensures that device parameters are optimized to achieve load reduction, with production deadlines met and reliable operation maintained. For instance, in the non-ferrous metal production line, an advanced control system monitors parameters such as current, voltage, and temperature in electrolytic cells, optimizing the electrolysis process to reduce energy consumption and achieve load reduction [44]. Similarly, in the electric arc furnace steelmaking, precise control of electrode positions and arc length optimizes energy consumption [45]. Chapter five will provide a detailed discussion on how process control optimization can facilitate load reduction, with a specific focus on the non-ferrous metal industry.

2.2. Load Shifting

Load shifting refers to the practice in which industrial entities, in response to external factors such as time-of-use electricity pricing (with electricity prices varying across different periods—typically higher during peak hours and lower during off-peak hours), proactively modify devices’ operation schedules. This approach shifts non-essential or flexible electricity demands from high-price periods to lower-price periods, thereby reducing energy costs and alleviating peak load demand on the grid. Without compromising production schedules or product quality, industrial loads can be redistributed across various time periods, facilitating temporal and spatial reallocation of electricity consumption. Under time-of-use pricing policies, this approach is suitable for industries with certain flexibility in production processes, such as cement manufacturing and wastewater treatment. It offers potential reductions in electricity costs, though it may require energy storage systems or adjustments to production schedules, potentially increasing management complexity. Currently, industrial load shifting methods include adjusting industrial production schedules and converting energy carriers.

(1)

Adjusting industrial production schedules

Adjusting industrial production schedules involves modifying the operational modes of industrial devices to optimize production processes, facilitating load shifting during specific time intervals and enabling flexible load dispatch. Factories can optimize their production schedules by adjusting the timing of tasks. For instance, high-energy consumption processes can be scheduled during off-peak electricity price periods, whereas low-energy or non-critical tasks are shifted to peak price periods, without compromising order delivery. Leveraging industrial internet and smart manufacturing techniques enables real-time monitoring of and dynamic adjustments to production devices. Through big data analysis and predictive models, plants can anticipate changes in electricity prices and automatically generate adjusted production schedules. Different industrial sectors, depending on their specific operational characteristics, identify distinct targets for load shifting; for example, electric arc furnaces in steel plants [46], electrolytic cells in electrochemical industries [47], aeration in wastewater treatment plants [48], crushers and cement mills in cement plants [49], and ball mills in ceramic manufacturing plants [50].

(2)

Converting energy carriers

Converting energy carriers refers to a process in industrial production that involves acquiring and converting energy during off-peak electricity price periods. The energy is then stored in appropriate carriers (such as electricity, heat, or gas) and released during peak price periods to support industrial production, effectively shifting energy consumption across time intervals. Electricity and heat are the primary energy carriers used to meet industrial load demands, relying primarily on storage techniques such as thermal storage [51], battery storage [52], and compressed air energy storage [53]. Additionally, the coupling relationship between hydrogen storage systems, water storage systems, and industrial combined heat and power systems is examined [54]. By operating hydrogen and water storage systems, industrial heat and power load shifting can be achieved. Furthermore, some industrial processes, due to their specific production characteristics, can develop customized energy storage methods to enable load shifting. For example, a load shifting scheme for ammonia synthesis plants through the storage of ammonia products was proposed [55]. The potential for load shifting in ethylene oxide production was demonstrated in [56] by storing intermediate products like oxygen and hydrogen.

2.3. Load Substitution

Industrial parks meeting internal energy demands through self-generation present an attractive alternative to reliance on large power grids. This is primarily achieved through combined heat and power generation and renewable energy generation. Load substitution focuses on optimizing internal resources to enable self-sufficiency in power supply, thereby reducing dependence on grid electricity from the demand side. This approach is not merely about decreasing the total load or adjusting the timing of electricity consumption, but rather about altering the source of electricity supply. This approach is suitable for industrial parks with access to renewable energy resources or waste heat and pressure, such as wastewater treatment plants and chemical and paper industries. It facilitates a reduction in dependency on the grid and lowers carbon emissions, though it necessitates substantial initial investments. In industrial parks, the payback period for photovoltaic energy storage systems can be shortened to within 8 years [57], though this is constrained by regional solar irradiance conditions. Furthermore, under China’s dual carbon policy, the development of certain industries is linked to the capacity of renewable energy installations, promoting the integration of heavy industries with renewable energy generation systems.

(1)

Combined heat and power

The technical potential for new combined heat and power (CHP) installations across the U.S. is an additional 149 million kilowatts of generation capacity [58]. CHP systems can be categorized into two types: those integrated within industrial processes and those newly established in industrial parks. CHP devices within industrial processes refers to systems designed to supply both heating and power by adjusting the CHP system based on constraints such as production tasks, device operation, natural gas prices, electricity prices, and raw material costs. Furthermore, various industrial systems or devices, such as industrial boilers, steam systems, wastewater treatment systems, cement kiln tail systems, and kiln cooling machines, often have substantial waste heat resources. Consequently, CHP systems are widely implemented in industries including cement, pulp and paper, wastewater treatment, and steel production. For instance, in wastewater treatment plants, strategies for maximizing resource utilization within the facility can be explored to enable on-site power generation for load substitution [59]. These strategies include electricity recovery from wastewater head height differences [60] and heat recovery from waste hot water [61,62]. Additionally, anaerobic digestion systems in wastewater treatment plants can be integrated with CHP systems to utilize biogas produced through anaerobic digestion to power the CHP system, facilitating on-site self-sufficiency while effectively utilizing waste heat and biogas [63]. The integration of biogas production, consumption, and storage with CHP systems was explored, and a predictive electricity price-based value order model was introduced to develop energy dispatch strategies for wastewater treatment plants in Germany under dynamic electricity pricing conditions [64].

The installation of CHP in industrial parks refers to the addition of such devices within the park, integrated with existing facilities and potential distributed or centralized renewable energy generation systems. This integration supplies both thermal and electrical energy for production and non-production devices within the park. To address cooling demands, certain industrial parks incorporate CHP systems [65]. Under the trends of high electrification and carbon reduction, the deployment of combined heat and power systems faces limitations imposed by energy conversion techniques and principles, which may restrict their role in advancing green and low-carbon development within the park [66].

(2)

Renewable energy generation

The installation of distributed or centralized renewable energy generation systems within industrial parks offers a competitive and sustainable low-carbon pathway for directly supplying power to on-site devices. Common configurations include distributed photovoltaic systems, distributed wind power systems, and hybrid centralized–distributed systems. With the increasing adoption of electric vehicles, solar-storage-charging systems and wind-solar-storage-charging systems have garnered significant attention from various institutions and organizations, driving the development of demonstration projects. Renewable energy systems, along with integrated energy storage solutions and electric vehicles or vehicle clusters, enable broader energy scheduling and load substitution across the industrial park.

The potential for integrating solar systems with wastewater treatment facilities to enhance renewable energy utilization is highlighted [67]. The coupling of electric arc furnaces with hydrogen storage systems and thermal energy storage systems is explored [68], demonstrating their economic feasibility and capability to accommodate renewable energy generation. The integration of solar systems with electrolytic hydrogen production offers a cost-effective and low-carbon option for steel enterprises employing electric arc furnaces with direct iron reduction processes [69]. Furthermore, reducing the costs associated with electrolytic cells is expected to facilitate the adoption of pure hydrogen direct reduction ironmaking techniques [70].

The decision factors influencing industries’ choice of whether to adopt flexible loads to participate in the electricity market route include production continuity requirements, device adjustability, and resource endowment. Production continuity requirements refer to the fact that continuous process industries give priority to non-interruption strategies (such as load substitution); equipment adjustability refers to the fact that industries with intermediate product storage capabilities (such as chemicals) are more likely to implement load transfer; and resource endowment refers to the fact that regions with rich wind and solar resources tend to replace new energy, and industries with sufficient waste heat resources are suitable for CHP. It is also possible to coordinate and optimize paths through multiple paths to achieve cross-strategy combination efficiency.

3. Techniques for Achieving Industrial Demand-Side Response

The integration of industrial participation in demand-side response within power systems relies on three key technical dimensions: distribution network techniques, system integration techniques, and industrial internet techniques. These three dimensions are interdependent, with distribution network and system integration techniques providing essential support for industrial internet. Together, they enable the effective implementation of industrial demand-side response, as illustrated in Figure 1.

Distribution network techniques encompass flexible AC/DC distribution networks, smart distribution network situation awareness, new forms and planning of distribution networks, and distribution network protection.

The flexible AC/DC distribution network technique allows for the adaptable transmission of electricity in AC or DC power transmission modes, meeting diverse load demands and enhancing power transmission efficiency [71]. This technique optimizes grid operation by modifying voltage levels and power flow directions, ensuring efficient, reliable, and economical electricity delivery.

Smart distribution network situation awareness employs advanced sensing devices and data analysis techniques to monitor grid operations in real time. It identifies potential faults or anomalies promptly and implements preventive measures, improving grid safety and operational stability [72]. New forms and planning of distribution networks address the challenges posed by renewable energy integration and the widespread adoption of electric vehicles. This technique focuses on developing innovative grid architectures and optimization strategies to ensure that future networks become more intelligent and environmentally sustainable [73].

Distribution network protection techniques for distribution systems leverage modern information techniques and automation devices to effectively manage and control distribution networks. These systems rapidly and accurately identify and isolate faulted areas, minimizing the extent and duration of outages and safeguarding the quality of power supply for users [74].

System integration techniques include integrated energy conversion techniques, energy storage techniques, and virtual power plant techniques.

Integrated energy conversion techniques enable the supply of diverse energy demands by converting between multiple energy carriers, such as electricity, heat, natural gas, and hydrogen. These techniques expand the possibilities for flexible energy scheduling in industrial energy systems. Among the most representative systems is industrial CHP generation, which demonstrates significant potential for flexible energy dispatch by integrating electricity, heat, and natural gas. Typically located within or near industrial facilities, these systems provide heating and electricity for industrial production. Industrial CHP systems are characterized by mature technical development, rapid response times, and dispatchability. Their flexibility can be further enhanced through the integration of thermal storage units, electric boilers, and heat pumps [75]. Advances in waste heat recovery techniques, such as organic Rankine cycle and heat pump systems, have facilitated the utilization of abundant waste heat resources in industries like steelmaking and cement plants [76,77]. Coupling techniques between energy systems, including steel steam systems and byproduct fuel systems, and waste heat recovery systems enable the interconnected and hierarchical utilization of multiple energy carriers, such as byproduct gas, steam, heat, and electricity [78].

In power markets, combined heat and power systems can participate in energy markets [79] and provide ancillary services, including frequency regulation [80] and reserve [81]. Micro-scale CHP systems can also meet the thermal and electrical demands of industrial parks, facilitating on-site resource flexibility and scheduling [82]. Furthermore, studies have revealed the potential of integrating combined heat and power systems with solar [83] and wind power [84] systems to enhance flexibility. Research has also begun to explore the potential of larger interconnected multi-energy systems and the integration of process industries with wind and solar power generation [85].

Hydrogen production from industrial waste heat and wastewater offers an attractive solution for cascading energy use, improving resource efficiency, and generating clean energy. Utilizing industrial combined heat and power systems for hydrogen production through waste heat recovery represents a primary method [86,87]. Scholars have also explored the feasibility of hydrogen production from wastewater in plants such as food processing [88]. Additionally, alkaline electrolysis-based hydrogen production systems can also be integrated with other systems, such as photovoltaic power generation systems [89], demonstrating potential for providing flexibility [90].

Advanced energy storage techniques are a critical component in enabling flexibility in load profiles. Advanced energy storage technologies serve as a pivotal component for enabling flexibility in load profiles on the demand side, and are also crucial for achieving system balance within power grids [91]. For industry, energy storage systems can store and release energy based on energy demand and price variations, optimizing energy scheduling strategies and enhancing the capacity for flexible energy supply [92]. Specific forms of energy storage include electrochemical storage [93], thermal storage [94], and cold storage [95]. Additionally, industrial loads can integrate energy storage system (ESS) service models by paying a service fee to store surplus renewable energy in ESS facilities, which can be discharged during peak periods to meet demand. This strategy optimizes energy usage, enhances renewable energy integration, and reduces annual operational costs [96]. The integration of solar thermal systems with industrial processes also provides flexible scheduling solutions, with applications in water purification [97], water heating [98], food processing [99], papermaking [43], and steel plants [100]. As industrial electrification increases and electric vehicles grows, vehicles within industrial parks can act as storage units, contributing to industrial flexibility scheduling [101]. Furthermore, advancements in hydrogen techniques have positioned hydrogen storage as a viable option for enhancing flexibility [102].

Supported by techniques such as smart meters, advanced metering infrastructure, and coordinated control systems, the virtual power plant has emerged as a business model capable of aggregating and managing distributed renewable energy, energy storage systems, and flexible loads. This model demonstrates substantial potential for flexibility in energy scheduling [103]. By integrating various flexible resources, virtual power plants reduce output errors caused by forecasting inaccuracies and unexpected events, facilitating improved industrial energy management, enhanced power supply reliability, and greater economic feasibility.

In research, industrial loads are often aggregated as interruptible or responsive loads within virtual power plants. These loads can be combined with commercial loads [104] and residential loads [105], electric vehicle clusters [106,107], and generation units such as hydropower [108], solar [109], wind power [110], and distributed generation [111]. The advancement of multi-agent system techniques further improves the coordination and flexibility of resource scheduling within virtual power plants [112]. For industrial loads, research on the aggregation characteristics of internal systems has gained increasing attention, with industrial virtual power plant techniques emerging as a focal area [113]. Specific industrial processes, such as manufacturing [114] and breweries, along with distributed generation, electric vehicles, and energy storage systems within industrial parks, can collectively form industrial virtual power plants. These systems enable participation in power markets on a more refined scale, driving higher levels of efficiency and market integration.

From the industrial perspective, techniques within the industrial internet encompass industrial big data, industrial Internet of Things, and industrial control and automation techniques.

The industrial big data technique spans the management of the entire data lifecycle, from data acquisition to analysis and mining, with the aim of enhancing decision-making efficiency, optimizing production processes, and reducing costs. It includes three key components: data acquisition and processing, data storage and management, and data analysis and mining.

Data acquisition and processing form the foundation of industrial big data. This typically involves real-time collection of large volumes of data on production, device status, and energy consumption through various sensors, devices, and monitoring systems [115]. The data include information from industrial devices, production planning and scheduling, and energy usage and demand, as well as market data. As these data may originate from a variety of sources, such as industrial devices, sensors, control systems, and external market environments, the acquisition technique must ensure high precision, real-time capability, and diversity to maintain data integrity and accuracy.

Following data acquisition, data storage and management become critical. With the rapid growth in data volume, traditional storage methods are no longer adequate. Modern industrial big data storage solutions leverage distributed storage architectures [116], such as cloud computing platforms, distributed databases, and big data frameworks. These techniques enable efficient storage and management of massive datasets, while ensuring high availability and reliability. Furthermore, data management includes processes such as data cleansing, preprocessing, and archiving, which help to ensure data quality, eliminate redundancy and noise, and improve the accuracy of subsequent analyses, such as industrial trading strategies.

Data analysis and mining represent the core aspect of industrial big data. By utilizing advanced machine learning and artificial intelligence algorithms, stored data can be deeply analyzed to uncover underlying patterns and trends. This process involves predictive analysis (e.g., forecasting market prices, load demands) [117], pattern recognition (e.g., identifying device failure modes, production optimization patterns) [118], and optimization algorithms (e.g., optimizing energy trading strategies) [119], all of which aid industrial loads in making informed decisions. Through the seamless integration of data acquisition, storage, and analysis, the industrial big data technique offers real-time monitoring and forecasting capabilities, enabling market participants to devise more precise trading strategies, improve energy efficiency, and optimize production scheduling.

Industrial Internet of Things techniques facilitate real-time monitoring and data transmission of industrial devices through intelligent sensing and interconnected techniques. It relies on highly reliable communication techniques [120], such as 5G and industrial Ethernet, to ensure the rapid and efficient flow of information between industrial devices, cloud platforms, and market participants. Edge computing [121] further enhances local processing capabilities, enabling industrial loads to adapt more swiftly to market fluctuations and engage in real-time transactions.

Industrial control and automation techniques [122] provide foundational support for efficient industrial operations. These techniques leverage advanced control algorithms and intelligent adjustment techniques to automatically optimize production processes and enable flexible load management. Automation systems adjust energy consumption based on power market price signals and production demands, ensuring production targets are met, while minimizing energy usage and transaction costs.

The integration of these techniques not only enhances industrial flexibility, but also enables industrial loads to participate in power markets with greater precision and intelligence. This leads to a dual benefit of improved economic performance and enhanced energy efficiency.

4. Flexible Industrial Load Trading Strategy and Modeling in Power Market

4.1. Trading Strategy

In power markets, trading strategies for flexible industrial loads are crucial for optimizing power resource allocation, enhancing system efficiency, and ensuring grid stability. With the continuous evolution of market mechanisms, industrial loads contribute flexibility through load shedding, load shifting, and load substitution. These contributions are significant not only in energy markets, but also in ancillary services markets, by providing essential services such as frequency regulation and reserve capacity. Depending on the market clearing mechanisms, the trading strategies for flexible industrial loads vary accordingly.

Industrial participation in power markets typically spans two types of markets: the energy market and the ancillary services market [123]. Based on these markets, clearing mechanisms are generally categorized into two types: independent clearing and joint clearing [124]. Independent clearing refers to market outcomes determined separately within either the energy or ancillary services market, with no notable interdependencies between them. Joint clearing, on the other hand, involves outcomes determined simultaneously or sequentially across both markets, reflecting temporal interconnections. Each market participation model has unique mechanisms and requirements, necessitating the precise optimization of trading strategies for flexible industrial loads. These strategies must align with market demands, constraints, and clearing sequences. The following sections provide a detailed discussion of trading strategies for flexible industrial loads under independent and joint clearing mechanisms, along with their applications and associated challenges in energy scheduling.

4.1.1. Independent Clearing Market

Independent clearing refers to the process in which different markets operate separately, with bids and clearings conducted independently and settlements managed by operators for each market. The independent clearing model for energy markets and ancillary services markets has been implemented in various countries and regions, including the Nordic countries, Germany, and the United Kingdom. This section provides a comprehensive review of research on flexible industrial loads participating in independent clearing markets, focusing on their applications and optimization strategies in energy markets and ancillary services markets.

• Energy market

The energy market represents a critical domain for industrial load participation in power markets, encompassing market trading mechanisms across various temporal scales. Based on differences in market clearing times and scheduling cycles, the participation forms and research hotspots of industrial loads will be introduced with regard to three aspects: day-ahead energy markets, real-time energy markets, and multi-timescale energy markets.

(1)

Day-ahead energy market

In the power markets of various countries and regions, day-ahead energy markets typically clear once per day, with clearing conducted the previous day to determine the prices and volumes for energy transactions on the following day. In these markets, electricity is generally sold at higher prices during peak periods, characterized by elevated demand, and at lower prices during off-peak periods, marked by reduced demand. Industrial loads participate in day-ahead energy markets either directly or through intermediaries, submitting their load profiles the day before to engage in the market. By voluntarily adjusting consumption patterns, industrial loads aim to reduce electricity costs or enhance profitability. Practical cases of participation in day-ahead energy markets can be observed in energy-intensive industries, such as steel plants, wastewater treatment plants, electrolytic aluminum plants, cement plants, and pulp and paper mills.

Research on trading strategies for industrial loads in day-ahead energy markets focuses on achieving optimal scheduling based on factors such as the power grid, the external industrial environment, and internal industrial resources. These factors include generation capacity, electricity prices, natural gas prices, industrial order demand, energy consumption requirements, and device conditions. By leveraging optimized scheduling algorithms, industrial load flexibility can be effectively utilized within the specified day-ahead time to respond optimally to market price signals, with research primarily aimed at cost minimization. For instance, the principles of group technique were introduced to linearize the scheduling model of discrete unit manufacturing systems, achieving an approximate 40% reduction in energy costs [125]. The integration of scheduling and optimization problems for steel plants with self-owned power plants in day-ahead energy markets was investigated [126], considering factors such as time-of-use electricity pricing, self-generation costs, and feed-in tariffs, aiming to optimize production planning and electricity costs. An evolutionary algorithm was combined with branch-and-bound methods to linearize mixed-integer nonlinear programming models [127], addressing optimal scheduling for industrial microgrids with energy storage systems to minimize total costs. Production scheduling strategies under time-based and incentive-based electricity pricing plans were analyzed [128], utilizing Markov chains to simulate peak and off-peak electricity periods and adjust production systems, thereby reducing operational costs for manufacturers. In some regions of the United States, electricity pricing plans include demand charges based on the highest average energy consumption within a specified time [129]. Energy scheduling for parallel machines under time-of-use electricity pricing and demand charges was further explored [129], demonstrating that this approach can lower peak demand and reduce overall energy costs for industrial operations.

(2)

Real-time energy market

The real-time energy market operates on decision-making intervals ranging from a few minutes to several hours, with a significantly shorter time span compared to the day-ahead market. Clearing intervals in real-time energy markets vary across countries and regions. For instance, in the United States, the PJM, CAISO, and MISO markets utilize 5 min clearing intervals, whereas Nord Pool in Europe adopts 15 min intervals. In these markets, system operators determine power generation and consumption curves based on day-ahead market clearing results and ultra-short-term supply–demand forecasts. They organize market transactions and allocate resources, resulting in awarded real-time market power and clearing prices. Unlike day-ahead prices, real-time prices reflect actual market conditions more accurately [130].

Under the influence of day-ahead market clearing, price differentials can lead to new demand peaks and troughs in the real-time market [131]. Furthermore, inaccuracies in supply–demand forecasts may contribute to discrepancies between day-ahead and real-time prices [132]. By participating in real-time energy markets, industrial loads can dynamically adjust their electricity consumption in response to market price signals, optimizing energy costs. Participation methods include direct involvement or engaging through intermediaries. Load profiles are submitted for market participation shortly before each clearing interval, typically 5 or 15 min.

Industrial loads participating in real-time markets can access more precise and immediate flexibility scheduling solutions. Research on trading strategies for industrial loads in real-time energy markets focuses on achieving optimal scheduling by considering factors such as the power grid, external industrial environment, and internal industrial resources. This ensures that industrial loads respond optimally to market price signals within specified real-time intervals, addressing the high-frequency volatility and short-term decision-making requirements of power markets. Key research areas include high-frequency decision-making models [30], uncertainty analysis [133], and assessments of reliability and stability [134]. High-frequency decision-making models involve the development of frameworks for frequent decision-making in real-time environments, such as 5 min or 15 min dispatch intervals.

(3)

Multi-timescale energy market

Focusing solely on energy scheduling strategies for industrial loads in single-timescale markets, such as day-ahead or real-time markets, increases the complexity of maintaining grid supply–demand balance and may result in network congestion. Recognizing the adoption of “intraday” markets in some countries and regions, researchers often consider optimization across three timescales: day-ahead, intraday, and real-time markets [134]. Industrial loads participate in multi-timescale energy markets either directly or through intermediaries, by submitting load profiles before the clearing intervals of specific timescale markets.

Multi-timescale markets integrate day-ahead, intraday, and real-time markets, aiming to minimize forecasting deviations and scheduling errors through coordinated and hierarchical optimization mechanisms. This approach enhances the economic efficiency and security of the power system, while reducing energy costs for industrial enterprises. Within these markets, industrial loads can establish load plans in day-ahead markets and dynamically adjust load response strategies in intraday and real-time markets to accommodate price fluctuations and grid conditions.

Research in multi-timescale markets primarily focuses on optimizing industrial load management across different timescales. Key areas of interest include multi-timescale coordination and optimization methods [135], multi-scale forecasting and uncertainty analysis [136], and hierarchical scheduling models [137]. Significant focus is placed on industrial production systems and integrated energy systems, such as industrial parks or aggregators, which combine industrial loads with other energy resources.

2.

Ancillary services market

The ancillary services market imposes specific requirements on participating demand-side resources, including adjustable capacity, continuous response time, and reliability. For industrial loads, effective participation in frequency regulation and reserve ancillary services markets hinges on their ability to achieve efficient load adjustment and provide reliable response times. This paper examines trading strategies for industrial participation in single-category ancillary services markets, with the discussion divided into two sections: frequency regulation and reserves.

(1)

Frequency regulation market

Frequency regulation, an essential active power balancing service, plays a critical role in maintaining power quality and ensuring the security of the power system. This service is typically divided into three types based on timescales. Primary frequency regulation adjusts active power output to reduce frequency deviations caused by system frequency straying from its target, with adjustments completed within seconds to tens of seconds. Secondary frequency regulation addresses imbalances not resolved by primary regulation, adjusting power output in real time to maintain frequency and interconnection power control, with response times ranging from seconds to approximately 15 min. Tertiary frequency regulation, implemented in some countries and regions, provides longer-timescale adjustments, typically between 15 min and several hours. Predominantly conducted on real-time timescales, frequency regulation addresses significant volatility in grid demand, requiring prompt responses to stabilize system frequency. Although certain frequency regulation services are pre-scheduled in day-ahead markets, most operational decisions and dispatch activities are executed within real-time markets.

Industrial loads, such as electrolytic aluminum and hydropower silicon, function as thermal energy storage loads with high thermal inertia. Short-term adjustments have minimal impact on normal production, making these loads suitable for dynamic and rapid adjustments in power systems [138,139]. Industrial users participating in frequency regulation markets can provide flexible load responses and receive payments for ancillary services, increasing their economic returns [140]. Additionally, demand-side frequency response services contribute to enhancing the stability and reliability of the power system [141]. Research in this area primarily focuses on optimizing frequency control methods for the dynamic characteristics of industrial loads.

In electrolytic aluminum loads, it can participate in solving the frequency control issue of isolated microgrids. For example, Ding et al. [141] proposed a coordinated frequency control scheme for isolated industrial microgrids. By combining the characteristics of electrolytic aluminum loads, feedback control signals are used to achieve power regulation, thereby maintaining frequency stability under power imbalance. Du et al. [142] proposed an adaptive frequency stability control method for the flexibility of electrolytic aluminum loads. By optimizing parameters, flexible load regulation is achieved and frequency response performance is enhanced. With the popularization of renewable energy, Liu et al. [143] proposed a frequency control scheme based on distributed model predictive control (MPC), which can cope with frequency deviations under different system topologies, while considering voltage constraints and effectively alleviating frequency fluctuations caused by renewable energy fluctuations. Du et al. [144] further improved the frequency response speed and stability of electrolytic aluminum loads through inertial response control. Experiments have shown that this method is effective in improving frequency regulation performance. In addition, Xing et al. [145] proposed a source–load coordinated control strategy for electrolytic aluminum loads participating in DC grid frequency regulation. The hierarchical control method was used to optimize frequency regulation and reduce regulation costs, and the economic benefits of the strategy in ensuring system stability were verified. In order to solve the time delay problem of industrial load frequency regulation, Liao et al. [146] proposed a coordinated control method of a saturated reactor and generator exciter to adjust the electrolytic aluminum load in order to balance the power imbalance. Bao et al. [147] verified the effectiveness of an electrolytic aluminum load participating in frequency control through field experiments, designed a rapid power reduction and wind abandonment experiment, and achieved a good frequency response effect. Liao et al. [148] proposed a control method based on electrolytic aluminum load regulation, which achieved smooth regulation of wind power fluctuations, while Bao et al. [149] proposed a solution to provide secondary frequency regulation auxiliary services through industrial load demand response through a day-ahead-real-time hierarchical structure.

(2)

Reserve market

Reserve capacity is designed to mitigate the risk of active power imbalances caused by forecasting errors and unforeseen events. Typically, after the energy market clearing process, the remaining capacity of cleared units can be combined with other units to participate in reserve markets through independent clearing mechanisms. In reserve markets, operational reserves are categorized based on unit status as either online reserves (also known as spinning reserves) or offline reserves (non-spinning reserves). The timeframes for reserve activation are generally divided into 10 min and 30 min intervals. In recent years, researchers have explored various flexible scheduling strategies for reserve capacity, aiming to optimize its application in independently cleared reserve markets. For instance, Zhou et al. [150] developed an optimization model for demand-side participation in grid reserves, addressing the issue of configuration optimization in the reserve market for demand-side response. A bi-level uncertainty optimization model was proposed [151] by integrating particle swarm optimization and differential evolution algorithms to develop optimal reserve scheduling plans. These plans were analyzed for maximum user profits on the demand side and minimum operating costs in microgrids. Additionally, Guo et al. [152] considered a fully decentralized peer-to-peer market mechanism, using a chance-constrained optimization method to determine the reserve capacity required from demand-side resources.

Industrial loads and other electricity-consuming resources within industrial parks can actively adjust their consumption over the short term to participate in reserve markets. For instance, a generalized modeling framework was developed for industrial steam system flexibility, combining physics-based and data-driven approaches [153]. The feasibility and economic benefits of industrial steam systems in reserve markets were validated through case studies involving steam devices in steel and cement plants. A parallel particle swarm optimization algorithm was applied to optimize energy scheduling for industrial participation in reserve markets, aiming to maximize profits [154]. A hierarchical control structure was proposed for electrolytic aluminum plant loads [155]. The upper-level control facilitates participation in reserve ancillary markets, while the lower-level control employs model predictive control principles to regulate direct current and respond to frequency deviations.

4.1.2. Joint Clearing Market

A joint clearing market integrates energy markets with ancillary services markets, such as frequency regulation and reserves. Based on system security constraints, industrial loads participate in co-optimized clearing markets through unified mathematical models, resulting in more cost-effective and strategic outcomes. Shi et al. [156] provided a comprehensive review of approaches to co-optimization in energy and ancillary services markets. Joint clearing typically adopts one of two methods: sequential clearing or joint clearing. In the sequential clearing model, markets are cleared in succession, with each subsequent market proceeding only after the successful clearing of the preceding one. Countries and regions implementing sequential optimization include the ERCOT market in the United States. The ERCOT market applies this model, and has introduced a Real-Time Co-Optimization Initiative aimed at achieving real-time coordination of energy and ancillary services markets by 2025. In contrast, joint optimization clears energy and ancillary services markets simultaneously. Regions adopting joint optimization include Italy, Australia, and the PJM and CAISO markets. The joint clearing market in Italy spans day-ahead, intraday, and real-time balancing markets. Within these markets, energy and ancillary services are co-optimized and settled across different timescales. Australia employs 5 min interval joint optimization for energy and ancillary services during real-time operations. The PJM market integrates day-ahead and real-time energy markets with ancillary services markets, including reserves and frequency regulation, with co-optimization conducted every 5 min. In China, provinces such as Guangzhou [157], Shandong [158], Zhejiang [159], and Inner Mongolia [160] have piloted market structures for co-optimized energy and ancillary services markets. For instance, the Guangzhou wholesale power market integrates day-ahead and real-time energy markets with ancillary services, optimizing residual energy [157].

Numerous studies have examined the potential and specific applications of flexible loads, such as industrial resources, in joint clearing markets. These studies explore the feasibility, economic benefits, scheduling flexibility, and impact of flexible loads on power systems within such markets. Yang et al. [161] analyzed the characteristics of flexible resources and identified their feasibility for participation in joint energy and ancillary services markets. Sun et al. [162] categorized loads into three types—interruptible, shiftable, and movable—and evaluated the economic benefits of interruptible and shiftable loads in joint energy and ancillary services markets. Tong et al. [163] introduced a three-stage model for load participation in both energy and ancillary services markets. The results indicate that the flexibility provided by industrial loads can alleviate grid balancing pressures and reduce electricity market costs. Zhang et al. [164] investigated the potential of the industrial natural gas in joint energy and reserve markets, highlighting electricity costs as the primary driver for achieving economic viability through flexible scheduling of air separation devices.

Research on trading strategies for specific industrial participation in joint clearing markets has primarily focused on plants such as metal smelting [165], chemical plants [166], and CHP systems [167]. Most studies adopt cost minimization [168,169] as the primary optimization objective, while a smaller portion focuses on profit maximization [167] and load flexibility potential [33]. Some studies employ multi-objective optimization approaches, including models that jointly consider minimizing energy operating costs and maximizing ancillary service benefits [165]. For specific market types, most research emphasizes the joint clearing of day-ahead energy and reserve markets [166]. A smaller portion addresses real-time markets or other ancillary services, including frequency regulation. Examples include studies on the joint clearing of real-time energy and reserve markets [170] and the integration of day-ahead energy and frequency regulation markets [167].

4.2. Trading Strategy Modeling Types and Solutions

The development of trading strategies for industrial participation depends not only on changes in market conditions, but also on the formulation of appropriate optimization models to represent these dynamics. Within this framework, the selection of models and corresponding solution methods becomes a critical factor in ensuring the effective implementation of trading strategies. This section begins by introducing common modeling approaches, typically categorized into deterministic and stochastic models. It then delves into the use of advanced solution algorithms to address multi-objective, multi-constraint, or uncertainty-related challenges within these models, ultimately enhancing the efficiency of flexible industrial load participation in power markets.

4.2.1. Modeling Types

In studying trading strategies for industrial load participation in power markets, these strategies are typically represented through optimization models. Depending on the characteristics of the market environment and the specific problem, optimization models are generally categorized into two main types: deterministic models and uncertainty models, as shown in

Table 4. Modeling types of trading strategies for industrial loads.

Modeling Type	Research Direction	Focal Points of Research
Deterministic models	Mixed-integer linear programming models	Satisfying load shifting constraints and equipment operational restrictions
	Nonlinear programming models
Uncertainty models	Dynamic programming models	The dynamic adaptability of industrial loads in response to changes in market conditions
	Scenario analysis
	Stochastic optimization
	Robust optimization
	Information gap decision theory

.

Deterministic models assume that all input parameters are fully known and remain constant during the modeling process. In the context of industrial load participation in power markets, these models are typically developed based on stable power grids, external industrial environments, and internal industrial resources, such as generation capacity, electricity prices, natural gas prices, industrial order demands, energy consumption requirements, and device conditions. The objective is to achieve optimal scheduling for industrial loads within a well-defined market environment. Deterministic models are relatively straightforward to construct, and commonly include mixed-integer linear programming models [135], nonlinear programming models [171], and dynamic programming models [172]. Industrial loads are scheduled according to predicted electricity prices, with the primary objective often being cost minimization [125] or profit maximization [173]. These models generally do not account for dynamic adjustments, and focus instead on satisfying load shifting constraints and device operational restrictions. They are also frequently used as the initial stage in multi-timescale optimization [137], enabling the development of long-term load plans that ensure overall industrial stability and feasibility. However, deterministic models overlook uncertainties, such as renewable energy variability, price fluctuations, changes in industrial demand, and device operational states. This limitation can lead to significant discrepancies between optimized results and actual performance, making it challenging for industrial loads to adapt to dynamic market environments or unexpected events.

Uncertainty models incorporate randomness and variability in input parameters, aligning more closely with real-world market environments and industrial operating conditions. In trading strategies for industrial load participation in power markets, the stochastic nature of renewable energy generation, frequent market price fluctuations, dynamic changes in industrial demand, and variability in device operational states are key factors in uncertainty modeling. By integrating these factors, uncertainty models account for their potential impacts on industrial load scheduling, providing more flexible and robust decision-making support for market participation. Common approaches to uncertainty modeling include scenario analysis [174], stochastic optimization [175], robust optimization [176,177], and information gap decision theory [136]. The primary objectives of these models are typically cost minimization or profit maximization. In addition to load planning, these models introduce flexibility constraints and risk metrics to ensure stability and feasibility under various scenarios. Compared to deterministic models, uncertainty models emphasize the dynamic adaptability of industrial loads in response to changes in market conditions. For instance, industrial loads can adjust energy usage plans based on real-time market price fluctuations or renewable energy output variability, mitigating economic losses caused by scheduling deviations during random events such as sudden drops in renewable generation [133] or extreme price volatility [178]. Uncertainty models can also be integrated into multi-timescale optimization [137], enabling short-term plans to guide real-time industrial response strategies and simultaneously maintaining long-term stability. This integration provides a more flexible optimization framework, balancing long-term objectives with short-term adaptability, and enhancing industrial participation in power markets. Although uncertainty models involve higher modeling complexity and require extensive data and computational resources, they are more effective in addressing dynamic market changes. This approach improves the practical implementation of optimization results, offering industrial loads a competitive edge in market participation.

It is worth noting that research on trading strategies leveraging machine learning concepts has emerged as a prominent area of focus. Although machine learning algorithms are not inherently part of optimization models, they can support the resolution of optimization problems or be integrated into optimization frameworks. For instance, machine learning can be utilized to predict or generate input data for optimization models, approximate complex constraints or objective functions in high-dimensional optimization models, or directly produce trading strategies. In the next section, “Model Solution Methods”, we will explore the application of machine learning approaches to solving trading strategies.

On this basis, the construction of the model not only requires considering the optimization objectives (such as cost minimization, benefit maximization, etc.), but also needs to combine multiple constraints, such as grid constraints, equipment performance limitations, and market rules. When constructing a trading strategy model, it is first necessary to clarify the optimization objectives and constraints of the model. For example, the optimization objectives of industrial loads may include minimizing electricity costs, maximizing revenue, or increasing the flexibility of market participation. Constraints include the physical constraints of the power system, power market rules, grid stability, and load scheduling constraints. In an environment with strong uncertainty, we usually use methods such as stochastic optimization models, robust optimization, or dynamic programming to deal with uncertain factors such as market price fluctuations and load forecasting errors.

The parameter settings in the model are based on market data, forecasting models, and experimental data. The selection of key parameters includes the fluctuation range of electricity market prices, load forecasting accuracy, demand response capabilities, equipment operating characteristics, etc. For uncertainty models, it is usually necessary to set the probability distribution of uncertain parameters based on historical data or market simulation data. For example, the uncertainty of new energy power generation may depend on meteorological data and historical power generation data. When determining parameter values, it is necessary to consider the dynamic changes in the actual power market to ensure the adaptability of parameter settings to the actual market environment.

It is important to note that while the optimization models introduced—both deterministic and uncertainty-based—offer significant advantages in capturing the operational complexity of industrial load participation in power markets, they also face practical limitations. Uncertainty models, for instance, typically require large amounts of high-quality historical and real-time data to accurately characterize the stochastic nature of renewable energy generation, market price fluctuations, and industrial load variations. In many cases, the availability and reliability of such data can be a major challenge, leading to potential inaccuracies in the model inputs. Additionally, the increased computational complexity associated with uncertainty modeling, particularly in scenarios involving robust or stochastic optimization, can impede real-time decision-making, which is critical in dynamic market environments.

To address these limitations, several potential strategies can be considered. First, the integration of advanced data acquisition and big data analytics platforms can improve the quality and availability of real-time data, thus enhancing the accuracy of uncertainty models. Second, the application of machine learning techniques to preprocess data and predict uncertain parameters can help to reduce the computational burden by providing more accurate input forecasts. Third, adopting model reduction and decomposition techniques may facilitate the partitioning of complex optimization problems into smaller subproblems that can be solved more efficiently in real time. Furthermore, leveraging high-performance computing resources or cloud-based platforms could mitigate computational challenges, ensuring that the models remain practical for real-time market operations. Future research should focus on hybrid approaches that combine these techniques to develop scalable, robust, and computationally efficient optimization frameworks for industrial load participation.

4.2.2. Model Solution Methods

In the power market, industrial load flexibility plays a critical role in regulating supply–demand balance and optimizing resource allocation. Its efficiency not only depends on the scientific nature of the model, but also on the sophistication and applicability of the solution methods. The optimization process of trading strategies typically involves complex multi-objective, multi-constraint, and multi-uncertainty problems, particularly when considering uncertainties in renewable energy output, market price fluctuations, and industrial energy demand. This increases both the scale and complexity of the model. In the face of these challenges, efficient solution methods are crucial for translating theoretical models into practical optimization solutions. By employing advanced solution techniques, industrial loads can better navigate market volatility, efficiently manage resource scheduling, and optimize trading strategies, thereby fully unlocking the potential of industrial load flexibility.

The model-solving methods for industrial participation in power market trading strategies can be divided into three categories: game theory, optimization algorithms, and machine learning or artificial intelligence, as shown in

Table 5. Model solution methods of trading strategies for industrial loads.

Solving Method	Research Direction	Focal Points of Research
Game theory	Cooperative games	Shapley values, Nash bargaining solutions, and coalitional games
	Non-cooperative games	Stackelberg game and evolutionary game theory
Optimization algorithms	Traditional mathematical optimization methods	Dual methods and feasible direction methods
	Heuristic and meta-heuristic techniques	Particle swarm optimization
Machine learning and artificial intelligence methods	Supervised learning	Forecast generation, energy demand prediction
	Unsupervised learning	Clustering based on generation system similarity, trading participant preferences, trading behaviors, and industrial loads
	Reinforcement learning	Dynamic pricing, production scheduling, energy dispatch, and trading optimization
	Deep learning	Power market forecasting
	Hybrid intelligence	Participant selection, dynamic pricing, energy dispatch, and trading strategy optimization

.

Game theory helps industrial enterprises to predict the decisions of other market participants, understand and manage market risks, and make more rational decisions, thereby formulating more flexible optimal trading strategies. Game theory can be classified based on the nature of the cooperation among participants, the order of decision-making, and the level of information available. It is divided into cooperative and non-cooperative games based on cooperation, static and dynamic games based on the sequence of decisions, and complete and incomplete information games based on the completeness of information. Ji et al. [179] reviewed the application of game theory in demand-side management, covering various scenarios and methodologies. In industrial participation in power market trading strategies, game theory is typically applied in industrial integration systems, aggregators, and peer-to-peer trading. For instance, Chen et al. [180] outlined the development of game theory in industrial parks, while Tushar er al. [181] provided a comprehensive review of its application in industrial peer-to-peer trading. Research on industrial load flexibility in power market trading strategies primarily focuses on cooperative games [182] and dynamic games within non-cooperative games [183]. Cooperative games emphasize collective rationality and cooperation, with participants forming alliances to maximize joint benefits. The core challenge lies in how to fairly distribute the benefits generated from cooperation, which may involve allocation methods such as Shapley values, Nash bargaining solutions, and coalition games. In industrial participation in the market, cooperative games can be employed to fairly distribute the joint benefits obtained by alliances formed between industrial consumers and other participants, utilizing tools like Shapley values [184], Nash bargaining solutions [185], and coalitional games [186]. Dynamic games in non-cooperative games focus on the order of participants’ decisions and the visibility of information. The key issue is how to make optimal decisions at each stage and predict opponents’ responses, maximizing the participant benefits. In essence, dynamic games emphasize the temporal evolution of strategies and their mutual interdependencies, with each decision influencing not only the immediate outcome, but also future decisions and results. In the context of industrial participation in market trading, common dynamic game methods include the Stackelberg game [137,187] and evolutionary game theory [188].

Optimization algorithms include traditional mathematical optimization methods and heuristic and meta-heuristic techniques. Traditional mathematical optimization methods refer to mathematical tools and approaches used to solve optimization problems, typically relying on mathematical models of the problem. In the context of optimizing industrial load flexibility in power market trading strategies, commonly used mathematical optimization methods include dual methods and feasible direction methods. Dual methods transform the original optimization problem into a dual problem, solving the dual problem to obtain the solution or bounds of the original problem. These methods include Lagrangian duality, linear duality, and convex duality, which are effective in addressing complex decision-making problems involving cost optimization, resource allocation, risk management, and uncertainty, all of which entail multiple constraints. Among these, Lagrangian duality is widely used for solving industrial problems related to revenue optimization [185,189], price limit optimization [190], and trading optimization under generation volatility [191]. Feasible direction methods involve finding optimization directions within the feasible domain and iteratively improving the objective function value until the optimal solution is reached. Common examples include gradient projection methods [192], sequential feasible direction methods, and the Frank–Wolfe algorithm [193]. For instance, the gradient projection method has been applied to determine optimal power decisions for industrial participation in point-to-point trading [192]. Heuristic and meta-heuristic methods simulate strategies inspired by nature or human behavior to quickly identify solutions close to the global optimum. These methods are well suited for complex, nonlinear, and high-dimensional optimization problems, providing approximate optimization solutions. Common algorithms in this category include particle swarm optimization (PSO), genetic algorithms (GAs), simulated annealing (SA), and ant colony optimization (ACO). Among these, PSO is frequently used in industrial production scheduling [194], trading optimization [195], and energy storage capacity planning [196].

Machine learning and artificial intelligence methods encompass supervised learning, unsupervised learning, reinforcement learning, deep learning, and hybrid intelligence. A comprehensive review of the current applications of machine learning and artificial intelligence on the demand side is provided in [197], which focuses on pricing mechanisms and demand-side incentives. Generally, machine learning and artificial intelligence methods employ data-driven approaches for making trading decisions in power markets, enhancing the accuracy of predictions and decisions. However, their operational mechanisms differ. Supervised learning utilizes labeled data to train models, enabling them to make accurate predictions on new, unseen data. The core challenge lies in selecting the appropriate features and models, relying on labeled data to ensure the quality and diversity of training data. In the context of industrial participation in trading strategies, supervised learning is employed to forecast generation [198], predict energy demand [199], and assist decision-makers in anticipating future market fluctuations. Unsupervised learning identifies hidden patterns and structures within unlabeled data. The focus is on clustering or dimensionality reduction algorithms that reveal underlying relationships in data. In industrial participation in trading strategies, research in unsupervised learning primarily revolves around cluster analysis of the supply and demand sides. This includes clustering based on generation system similarity [200], trading participant preferences [201], trading behaviors [202], and industrial loads [203,204], thereby optimizing trading decisions. Reinforcement learning learns optimal strategies through trial-and-error and feedback. The key challenge is to balance exploration of new strategies and exploitation of known optimal ones, with an emphasis on continuously refining strategies through reward mechanisms. Reinforcement learning is applied in dynamic pricing [205], production scheduling [206], energy dispatch [207,208], and trading optimization [209], facilitating industrial participation in power markets. Deep learning leverages multi-layer neural networks to automatically extract and learn complex data features. Its core is the use of large amounts of data and computational resources for training, enabling high-level pattern recognition and predictive capabilities. Deep learning includes techniques such as deterministic gradient methods, convolutional neural networks, and recurrent neural networks. A review of deep learning applications on the demand side is provided in [210]. In industrial trading strategies, deep learning is often applied to power market forecasting. For instance, Jamil et al. [211] used recurrent neural networks for predicting energy demand in power markets, while Hua et al. [212] employed convolutional neural networks for the comprehensive forecasting of generation, energy demand, and electricity prices. Hybrid intelligence refers to combining multiple intelligent methods to enhance decision-making capabilities, primarily referring to deep reinforcement learning in this context. Deep reinforcement learning combines deep learning with reinforcement learning, using neural networks to handle complex, high-dimensional data and learn optimal strategies. The key challenge is the use of deep neural networks for feature extraction and strategy optimization to address decision-making problems in complex environments. In industrial participation in trading strategies, deep reinforcement learning can be employed for tasks such as participant selection [213], dynamic pricing [214], energy dispatch [215], and trading strategy optimization [216], facilitating adaptive responses to fluctuations in power markets and promoting industrial engagement in power market participation.

5. Optimization of Non-Ferrous Metal Industries Under Production Planning

Section 4 discusses the economic analysis of industrial loads in the context of power markets, focusing on how industrial loads can participate in demand response. This section examines the optimization of internal process parameters, such as current density, temperature, and concentration, within the framework of production planning. Through methods such as simulation, process modeling, and experimental conditioning, the optimal values for these parameters can be determined, which, in turn, facilitates industrial load reduction—a process referred to as parameter optimization. Parameter optimization has long been a critical aspect of ensuring the closed-loop stability and decision-making efficacy of industrial manufacturing operations. However, industrial processes are inherently complex, with reaction dynamics that exhibit intricate characteristics. Reaction rates and product distributions can vary significantly depending on conditions, with these fluctuations sensitive to changes in input variables. These conditions involve several inter-related factors, such as raw material ratios, temperature, and pressure. The interaction between these variables forms complex coupling relationships, and adjustments to one parameter may affect others. Additionally, device constraints in production, such as the pressure and temperature tolerances of reactors and mixing efficiency, can limit the scope of optimization. Therefore, reviewing the current research on optimizing internal process parameters is vital for improving industrial efficiency and flexibility. Electrolysis, a key energy-intensive process in non-ferrous metal production, accounts for over 75% of energy consumption. Given its substantial potential for flexible regulation, this section focuses specifically on the optimization of electrical parameters, such as voltage and current, within industrial settings.

Non-ferrous metal smelting can be divided into two primary categories: pyrometallurgy and hydrometallurgy. Pyrometallurgy involves high-temperature melting and reduction reactions to process ores or concentrates for metal extraction, encompassing processes such as smelting, converting, and refining. Electrolytic aluminum production predominantly employs pyrometallurgical methods, specifically, the electrolysis of molten aluminum oxide to produce aluminum. Approximately 80% of global copper and 15% of zinc are produced using pyrometallurgical processes. Hydrometallurgy, on the other hand, utilizes chemical solutions to selectively recover valuable elements and prepare products, including processes such as leaching, extraction, and electrowinning. Globally, nearly 20% of copper and over 85% of zinc are produced through electrowinning [217]. Liu et al. [218] reviewed the current research on electrowinning within hydrometallurgy from the perspectives of electrolytes, electrodes, and electrolytic cells.

This section will review the status of parameter optimization setting in the three industries of electrolytic aluminum, electrolytic zinc, and electrolytic copper, respectively.

(1)

Electrolytic aluminum

In the field of electrolytic aluminum, the current large, prebaked cell production process technique in China primarily focuses on reducing energy consumption and overall costs. There are two main production process routes adopted by the domestic electrolytic aluminum industry. One is the “low-voltage” production process proposed by the Zhengzhou Research Institute of Aluminum Corporation of China in 2008, aimed at reducing energy consumption in the production process. This technique reduces the working voltage of the electrolytic cell by lowering the reactive power voltage, typically keeping the working voltage of the electrolytic cell below 3.95 V, and even lower than 3.90 V, thereby reducing the energy consumption per ton of aluminum produced. Currently, due to rising electricity prices and the pressure of energy conservation and emission reduction policies, most enterprises have adopted the “low-voltage” process. The other route is the “three-degree optimization” process developed by the Guizhou Aluminum-Magnesium Design Institute of Chinalco International in 2006, which optimizes the electrolyte temperature, initial crystal formation temperature, and superheating degree. Although this process has certain disadvantages in terms of energy consumption, leading many companies to abandon it, it is still used by a few enterprises, such as Sichuan Qimingxing, Yunnan Aluminum, Chinalco Guangxi Branch, and Shandong Huayu Aluminum. The optimal economic benefit of electrolytic aluminum often comes from a significant reduction in electricity costs.

The DC power consumption calculation formula of the electrolytic cell [219] is as follows:

$$W(\mathrm{KWH}/\mathrm{ton} \mathrm{aluminum})=2980\times V_{\mathrm{average}}/CE$$

(1)

where $W$ is the DC power consumption per ton of aluminum (kWh/t-AI), $V_{average}$ is the average cell voltage of the electrolytic cell (V), and $CE$ is the current efficiency of the electrolytic cell (%). The main way to reduce electricity costs is to improve the current efficiency and reduce the average voltage.

In the electrolytic aluminum process, key operational parameters that affect normal production and energy efficiency include the electrode distance, cell resistance, current, cell voltage, anode coefficient, electrolyte composition, alumina content, and temperature. To improve current efficiency and increase the aluminum output per unit of current, optimization can be achieved by reducing the electrolysis temperature [220], adjusting the electrolyte composition [221], optimizing the electrode distance [222], and lowering the alumina concentration [223]. Additionally, enhancing electrolyte conductivity, reducing anode overvoltage, improving conductor contact points, and minimizing anode effect sharing also help to reduce cell voltage, thus enabling a “low-voltage” production process. The average cell voltage is composed of polarization voltage, anode voltage drops, cathode voltage drops, electrolyte voltage drops, effect-sharing voltage, and conductor bus voltage drops. By reasonably combining low-voltage management, low alumina levels, short electrode distance, low temperature, and high current density parameters, a reduction in cell voltage can be achieved. It is noteworthy that adjusting the electrode distance appropriately can lower the voltage while maintaining current efficiency, reducing energy consumption.

According to existing processes, reducing the electrode distance by 1 mm results in a 30–40 mV decrease in cell voltage. As shown in Equation (1), for each 1 mV decrease in cell voltage, the direct current energy consumption per ton of aluminum can be reduced by approximately 3.2 kWh. The electrode distance, defined as the distance from the anode bottom to the cathode aluminum liquid surface, not only impacts the electrochemical reaction zone during electrolysis, but also directly affects the electrolysis temperature and current efficiency. Increasing the electrode distance can reduce the aluminum loss and improve the current efficiency, while reducing the electrode distance helps to lower the voltage and save energy. However, an excessively short electrode distance may increase aluminum loss and decrease current efficiency. In actual production, the electrode distance is typically maintained between 4.0 and 5.0 cm. Yang et al. [224] explored the relationship between cell voltage, current density, and electrode distance based on energy and heat balance in aluminum electrolysis cells.

(2)

Zinc electrolysis

The zinc electrolysis process consists of three primary stages: electrolyte preparation, zinc electro-deposition, and rectified power supply. Among these, zinc electro-deposition refers to the process in which zinc is deposited on the cathode from zinc sulfate electrolyte under the influence of a direct current using insoluble anodes. This process is a critical component of the hydrometallurgical zinc refining procedure, and represents the stage with the highest electrical energy consumption, accounting for 79–88% of the total energy consumption.

Yang et al. [225] systematically investigated the process of collaborative optimization and intelligent control methods, considering various production levels, such as the entire process, individual stages, and reactors. The zinc hydrometallurgical process consists of roasting, leaching, and electrowinning, with the entire operation being an energy-intensive procedure. Among these, the electrowinning stage accounts for 80% of the total energy consumption in the entire zinc hydrometallurgical process. The current efficiency serves as a key indicator for assessing the zinc electrowinning performance, representing the ratio of actual zinc mass to theoretical zinc mass. It is influenced by factors such as electrolyte temperature, current density, sulfuric acid concentration in the electrolyte, zinc ion concentration, and the presence of impurity ions in the electrolyte.

In zinc electrowinning, factors influencing energy consumption during the electrolysis process include the current density, electrolyte temperature, sulfuric acid concentration, zinc ion concentration (referred to as the electrolyte acid-to-zinc ratio), and content of impurity ions. This section primarily addresses research on the relationship between current density and other influencing factors. For instance, it examines the impact of impurity ion content [226], acid-to-zinc ratio [227], and both temperature and acid-to-zinc ratio on energy consumption [228] under stable current density conditions. Furthermore, strategies for controlling the electrolyte acid-to-zinc ratio during current density switching have been studied. Under time-of-use electricity pricing, optimization of the electrolyte acid-to-zinc ratio during current density transitions has been addressed [229] using techniques such as reinforcement learning [230] and deep learning [231]. Additionally, research has examined the synergy between the electrolysis process and external conditions, such as photovoltaic–zinc electrolysis coupling systems [232], offering innovative approaches to enhancing energy efficiency and reducing costs.

(3)

Copper electrolysis

Energy demand in copper production accounts for approximately 7% of global industrial process energy consumption [233]. In copper production facilities utilizing either pyrometallurgical or hydrometallurgical processes, the electrolysis step is often unavoidable. Impure cast anodes are used in electrorefining to produce cathode copper, whereas electrolytic deposition is used to produce cathode copper. Most cathode copper is produced through pyrometallurgical and electrorefining processes, which involve stages such as flash smelting, top-blown refining, anode furnace refining, and subsequent refining. Flash smelting involves smelting copper ores and fluxes into copper matte, with top-blown refining devices refining the matte into blister copper. Anode furnace refining further refines blister copper into anode copper, which is then cast and sent for subsequent stages. The essence of pyrometallurgical copper production is to gradually oxidize copper ores while removing other impurities. The oxygen stage provides oxygen for the pyrometallurgical copper production process, and the refining stage provides air to accelerate the oxidation of copper. Both stages supply a certain proportion of oxygen-enriched air to assist the oxidation of copper ores in the pyrometallurgical process. The sulfur dioxide gas generated during pyrometallurgical copper production is absorbed during the acid production stage and directed to acid towers for sulfuric acid production. During electrorefining, anode copper is electrolyzed in sulfur copper solutions at controlled temperatures to obtain cathode copper, and some precious metals can also be recovered.

In addition to electrorefining, other processes in copper production also require significant amounts of electricity. For instance, flash furnaces and converters require substantial oxygen and operate air separation units. Flash smelting furnaces, converters, and anode furnaces also require large quantities of compressed air, which necessitates the operation of air compressors. Moreover, environmental protection measures can account for up to 30% of the facility’s total energy consumption [234]. For instance, significant volumes of waste gases must be treated and cleaned before being released into the atmosphere, and the large quantities of sulfur dioxide (SO₂) produced during flash smelting and converting processes must be removed from exhaust gases. Hoang et al. [235] investigated the optimal control of parameters such as voltage, electrolyte composition, and temperature in copper electrorefining, focusing on minimizing energy consumption through control strategies.

6. Challenges

As a major energy consumer, industry provision of flexible electricity demand contributes to alleviating the pressures associated with maintaining stability in emerging power systems with high shares of renewable energy generation, thereby supporting industrial economic development. However, several challenges remain regarding industrial flexibility’s participation in power markets. This paper discusses these challenges from three perspectives: market, technique, and participants.

6.1. Market Challenges

Establishing a platform that enables the free flow of power supply and demand information, while ensuring secure data sharing, presents a significant challenge. Simultaneously, the platform must guarantee that electricity traded by different industrial entities remains uniform. In the future, the market will require the development of supporting institutional mechanisms to encourage participation from industries, aggregators, virtual power plants, and other intermediary agents in the power market.

To further incentivize industrial load flexibility, the current market mechanism requires improvements in both incentive schemes and supervisory measures. Specifically, the establishment of standardized performance evaluation metrics based on traded quantity and timescales, along with the introduction of dynamic pricing and differentiated compensation mechanisms, can effectively reward those participants who demonstrate superior flexibility performance. Moreover, to address potential issues of unfair competition, top-level design must clearly delineate the operational and developmental rights of various market participants, and multi-tier supervisory approaches—such as asymmetric supervision—should be explored to ensure equitable treatment in data transparency, real-time monitoring, and risk management. In parallel, refining the complementary policy framework and establishing a clear institutional design will provide a robust foundation for promoting industrial load participation and ensuring fair, efficient market transactions. Furthermore, the market should develop unified standards for the quantity and timescales of the electricity traded, and implement corresponding incentive and compensation policies tailored to the varying challenges and performance outcomes of different industrial entities, thereby ensuring stable operation within the power market.

6.2. Technical Challenges

Under the principle of safety and reliability, the development of integration techniques for industrial and other disruptive loads poses a significant challenge and an urgent issue. In processes such as steel production, electric arc furnaces use arc heat to melt metals, resulting in highly random loads that are strongly time-varying and nonlinear. These loads cause severe power quality issues, including voltage fluctuations, flickering, voltage dips, and harmonics, which, in turn, affect other electrical loads within the steel plant and other users of the public grid.

Through an analysis of industrial demand and the integration of external market prices, an optimal scheduling strategy is selected for the entire industrial production process. This strategy considers the inherent production costs and energy conversion efficiency of the various systems and devices involved, aiming to minimize energy costs and meet the diverse energy demands of industrial operations. A key feature of this approach is its ability to adjust in response to external price signals, thus proactively managing operational risks associated with fluctuations in future power markets, focusing on the interaction between external and internal characteristics.

Data collection, processing, and sharing techniques for industrial loads present significant challenges in the context of industrial participation in power market transactions. In terms of data collection, industries can utilize smart meters, sensors, and other hardware to accurately capture the real-time operational status of electrical devices, which forms the basis for regulating flexible industrial loads. Data collection schemes must be tailored to the specific production characteristics and load profiles of industrial facilities. Regarding data processing, industries must extract the flexibility potential of loads, while maintaining high production quality and efficiency. This enables data support for industrial participation in power markets. The combination of artificial intelligence and expert knowledge enhances rapid and accurate data processing, facilitating the prediction and optimization of production processes within defined timeframes. For data sharing, it is essential to ensure the security and privacy of industrial load information as industries align with power market participation requirements.

6.3. Market Participants

To support industrial loads in adjusting to power systems, existing infrastructure is insufficient. Significant investments in both construction and transaction costs are required for industrial loads and the grid, along with other stakeholders, to negotiate flexibility trading agreements. Additionally, policy frameworks, market structures, and various uncertainties may dampen the willingness of participants. Currently, the involvement of industrial loads through intermediaries in power markets lacks sufficient regulatory oversight, potentially affecting participant engagement.

It is essential to tailor strategies based on local conditions for the integrated development and use of industrial production characteristics alongside energy sources such as wind, solar, geothermal, and biomass. Optimizing the layout of power, gas, thermal, cooling, and water supply networks is crucial. Approaches such as CHP systems, distributed renewable energy, and energy-smart microgrids can facilitate multi-energy complementarity and collaborative supply, enhancing energy provision for industrial parks and other heavy-load sectors. Moreover, leveraging national, regional, and municipal policies related to industrial participation in power markets can open avenues for exploring demand-side services within the power system.

7. Conclusions

Industrial load flexibility scheduling offers a powerful solution to mitigate the operational risks of new power systems from a demand-side perspective. It represents a win–win strategy for both the power grid and the industry. This paper reviews the research progress in industrial load flexibility scheduling. It investigates the potential and feasibility of flexible scheduling techniques for industrial loads, based on the characteristics of industrial energy dispatch. The review highlights the various methods through which industrial processes contribute flexibility, and their specific applications across different sectors. Subsequently, the paper explores the relationship between the external characteristics of industrial loads and incentives within power markets. It examines how power markets influence industrial loads, and surveys the trading strategies of industrial loads in two types of markets—energy markets and ancillary service markets. Finally, focusing on the electrolytic industry as a case study, this paper delves into the interaction between external and internal characteristics, with a particular emphasis on optimizing internal process parameters, such as current density, temperature, and concentration. It provides a comprehensive review of research on the optimization of process parameters in electrolytic industries. Despite ongoing challenges related to market, technical, and stakeholder aspects, industrial load flexibility remains a dynamic and promising area of research under the context of unified power markets and new power systems.