Hybrid Machine Learning and Modified Teaching Learning-Based English Optimization Algorithm for Smart City Communication

Abstract

This paper introduces a hybrid algorithm that combines machine learning and modified teaching learning-based optimization (TLBO) for enhancing smart city communication and energy management. The primary objective is to optimize the modified systems, which face challenges due to their high population density. The proposed algorithm integrates the strengths of machine learning techniques, more specifically, the long short-term memory (LSTM) technique, with teaching learning-based optimization algorithms. To achieve optimization, the algorithm learns from historical data on energy consumption and communication patterns specific to the modeled system. By leveraging these insights, it can predict future energy consumption and communication patterns accurately. Additionally, the algorithm incorporates a modified teaching learning-based optimization approach inspired by the teaching and learning process in classrooms. It adjusts the system parameters based on feedback received from the system, thereby optimizing both energy consumption and communication systems. The effectiveness of the proposed algorithm is evaluated through a case study conducted on the test system, where historical data on energy consumption and communication patterns are analyzed. The results demonstrate that the algorithm efficiently optimizes the communication and energy management systems, leading to substantial energy savings and improved communication efficiency within the test system. In conclusion, this study presents a hybrid machine learning and modified teaching learning-based optimization algorithm that effectively addresses the communication and energy management challenges in the test system. Moreover, this algorithm holds the potential for application in various smart city domains beyond the test system. The findings of this research contribute to the advancement of smart city technologies and offer valuable insights into reducing energy consumption in densely populated urban areas.

1. Introduction

Smart cities rely on effective communication and energy management to enhance sustainability and improve the quality of life for urban residents [1]. Communication systems in smart cities integrate information and communication technologies to optimize various urban services, including transportation, healthcare, and public safety. Real-time monitoring and data analysis facilitated by efficient communication management enable prompt responses to urban challenges and emergencies, enhancing overall safety, accessibility, and convenience [1].

Similarly, smart energy management plays a crucial role in smart cities, offering significant environmental and economic benefits. By integrating renewable energy sources, energy-efficient technologies, and advanced monitoring systems, smart cities optimize energy consumption and reduce carbon emissions [2,3]. This approach allows them to decrease their carbon footprint, mitigate climate change, and foster sustainable energy systems [4,5]. Moreover, effective energy management leads to substantial cost savings for cities and residents, promoting economic growth [4,5]. Thus, communication and energy management are critical components of smart cities that contribute to sustainability, economic development, and the well-being of urban residents [6].

However, smart cities encounter challenges in communication and energy management that require advanced machine learning algorithms for forecasting and optimization [7,8]. Integrating multiple communication technologies and protocols poses a challenge, resulting in data silos and interoperability issues that hinder effective decision making and response to urban challenges [9,10]. Advanced machine learning algorithms can address this challenge by integrating diverse data sources and facilitating real-time data analysis for informed decision making. These algorithms can also predict traffic patterns, forecast energy demand, and identify potential communication bottlenecks [11].

Energy management in smart cities faces challenges related to renewable energy integration, load balancing, and energy storage [12]. The intermittent nature of renewable energy sources such as solar and wind power presents difficulties in grid stability. Real-time load balancing requires optimization algorithms that can match the energy supply with the demand, ensuring stable and efficient energy provision. Additionally, energy storage plays a critical role in mitigating the intermittency of renewable energy sources and providing backup power during emergencies [13,14]. Advanced optimization algorithms enable effective energy storage management, minimizing energy losses [15]. In addition to these challenges, smart cities encounter other communication and energy management hurdles, including data privacy and security, policy and regulatory frameworks, and public awareness and participation [16]. Addressing these challenges necessitates a multidisciplinary approach involving collaboration among city managers, policymakers, researchers, and the public. Advanced machine learning and optimization algorithms play a pivotal role in overcoming these challenges, equipping city managers with the necessary tools for effective decision making and optimization of communication and energy systems to promote sustainable and efficient urban living [17,18].

Research in smart city communication and energy management has grown significantly, with numerous studies investigating various aspects of these domains. Communication research focuses on wireless sensor networks, the Internet of Things (IoT), machine learning, and cloud computing. For instance, studies explore the use of wireless sensor networks for monitoring air quality, traffic, and environmental conditions in cities. Machine learning algorithms predict traffic congestion and optimize routing, while cloud computing enables data storage and analysis from multiple sources [19,20].

Research in smart energy management encompasses topics such as renewable energy integration, energy storage, and demand response. Studies examine the impact of integrating renewable energy sources such as solar and wind power on grid stability and energy costs. Advanced energy storage systems, including batteries and flywheels, mitigate the intermittency of renewable energy sources. Demand response investigates the potential of demand-side management to optimize energy consumption and reduce costs [21].

Furthermore, research explores the integration of communication and energy systems in smart cities, focusing on energy-efficient communication protocols, data privacy and security, and policy and regulatory frameworks. Energy-efficient communication protocols for IoT devices minimize energy consumption while efforts are made to ensure data privacy and security. Policy and regulatory frameworks study the role of government policies in promoting sustainable and efficient urban living. Therefore, research on smart city communication and energy management has delved into various aspects, spanning wireless sensor networks, machine learning, renewable energy integration, policy frameworks, and more. These studies provide valuable insights into the challenges and opportunities of smart city communication and energy management, emphasizing the need for advanced machine learning and optimization algorithms to address these challenges and foster sustainable urban living.

Among the advanced techniques employed in smart cities, long short-term memory (LSTM) networks have demonstrated remarkable success in time series forecasting. Unlike traditional recurrent neural networks (RNNs), LSTM networks effectively address the issue of vanishing gradients, enabling them to capture long-term dependencies in sequential data [21]. In the context of smart cities, LSTM networks have been utilized for load and market price forecasting, empowering city managers to make informed decisions regarding energy generation and pricing. The use of LSTM networks in smart cities offers several advantages, including the ability to handle complex and nonlinear relationships within the data and the capacity to learn from historical data, leading to accurate predictions. Furthermore, the modified teaching learning-based English optimization algorithm (MTLEOA) presents a hybrid optimization algorithm that combines the principles of TLBO and English learning optimization (ELO) algorithms. This innovative approach has been proposed as a novel technique for energy management and communication optimization in smart cities, enabling city managers to optimize energy generation, distribution, and improve communication systems [22]. The MTLEOA offers several advantages, such as efficient convergence to global optima and the ability to handle multiple objectives and constraints in optimization problems. It outperforms traditional optimization algorithms, including GA and PSO, in various optimization scenarios.

The combination of LSTM networks and the MTLEOA as a hybrid technique for smart city communication and energy management represents a novel and promising approach. This innovative approach enables accurate predictions for load and market price forecasting while optimizing energy generation, distribution, and communication systems. Given the challenges faced in smart city communication and energy management, the utilization of advanced machine learning algorithms and optimization techniques becomes essential for promoting sustainable and efficient urban development. Consequently, this hybrid approach presents a compelling direction for future research in the field of smart cities. It equips city managers with effective tools to optimize communication and energy systems, fostering sustainable urban development. The primary contributions of this paper can be summarized as follows:

• Hybrid machine learning and optimization technique: This paper proposes a novel hybrid technique that combines LSTM machine learning for forecasting with the MTLEOA for energy management and communication optimization. This unique hybrid approach has not been explored before in the context of smart cities.
• Case study of the test system: This paper presents a detailed case study conducted on the test system, illustrating the effectiveness of the proposed hybrid technique in optimizing communication and energy systems within a real-world smart city setting. The case study demonstrates significant improvements in energy efficiency and communication performance achieved through the hybrid technique.
• Performance evaluation: This study rigorously evaluates the performance of the proposed hybrid technique using various metrics such as mean absolute percentage error (MAPE) and peak signal-to-noise ratio (PSNR). The evaluation demonstrates that the hybrid technique surpasses traditional optimization algorithms and single machine learning models, showcasing its superior performance.
• Contribution to sustainable urban development: By providing a new approach for optimizing communication and energy systems in smart cities, the paper contributes to the field of sustainable urban development. The proposed hybrid technique empowers city managers to make informed decisions regarding energy generation, distribution, and communication systems, ultimately leading to more sustainable and efficient urban living.

To sum up, the motivation of this research is to address the challenges in smart city communication and energy management, leveraging advanced machine learning and optimization techniques. By optimizing communication and energy systems, this research aims to improve urban services, sustainability, and the quality of life for residents. The proposed hybrid approach offers a new solution to enhance energy efficiency, communication performance, and promote sustainable urban development. It is worth nothing that in this paper, we built upon the previous work of Wael J. et al. [5] on SDG 11: Sustainable Cities and Communities, and proposed enhancements to their framework. Our improvements aim to further advance the goals of sustainable urban development, considering factors such as efficient communication systems, optimized energy management, and overall quality of life for urban residents. Additionally, we developed a new framework based on the paper of Arnob et al. [15], which focused on SDG 12: Responsible Consumption and Production. Our framework seeks to improve responsible consumption patterns and production practices in order to reduce waste, promote resource efficiency, and foster a more sustainable approach to consumption and production in smart cities. By integrating these enhancements and frameworks, our research contributes to the broader objectives of the 2030 Agenda for Sustainable Development, specifically targeting SDG 11 and SDG 12. We aim to promote sustainable cities and communities, as well as responsible consumption and production practices, for a more sustainable and prosperous future. The Sustainable Development Goals (SDGs) are important because they provide a comprehensive framework to address global challenges and work towards a more sustainable and equitable future [22]. They tackle critical issues such as poverty, hunger, health, education, gender equality, clean energy, sustainable cities, and climate action. The SDGs guide governments, organizations, and individuals in setting priorities, implementing solutions, and measuring progress towards achieving a more prosperous and peaceful world for all, leaving no one behind. By addressing these goals, we can create a better future for current and future generations, protect the planet, and ensure a higher quality of life for everyone.

2. Literature Analysis

Smart cities are aimed to optimize communication and energy management systems to enhance sustainability, economic growth, and quality of life of urban residents [21,22]. This literature review aims to evaluate the proposed hybrid machine learning and the modified teaching learning-based optimization English algorithm for smart city communication and energy management through a case study of the test system [23,24]. The algorithm combines machine learning techniques and the teaching learning-based optimization algorithm, which optimizes the energy consumption and communication systems by adjusting their parameters [25]. The proposed algorithm was applied to a case study of the test system, and it was shown that the algorithm could effectively optimize communication and energy management systems, leading to significant energy savings and improved communication efficiency [26]. The efficient management of communication systems in smart cities can enable real-time monitoring and data analysis, which facilitates prompt responses to urban challenges and emergencies, thus enhancing the overall safety, accessibility, and convenience of urban living, contributing to the economic growth and social well-being of the city. Similarly, smart energy management involves the integration of renewable energy sources, energy-efficient technologies, and advanced monitoring systems to optimize energy consumption and reduce carbon emissions. Therefore, it is imperative to adopt advanced technologies and effective management strategies to leverage communication and energy systems for sustainable development [27].

Smart cities face several challenges in communication and energy management, such as the integration of multiple communication technologies and protocols, data silos, interoperability issues, integration of renewable energy sources, load balancing, and energy storage. The integration of diverse data sources and the facilitation of data analysis for real-time decision making is crucial, and machine learning algorithms can help overcome this challenge. Machine learning algorithms can also predict traffic patterns, predict energy demand, and identify potential communication bottlenecks in the city. Similarly, advanced optimization algorithms can optimize energy generation, storage, and distribution to ensure a stable and efficient energy supply. These algorithms can also enable effective energy storage management, optimizing the use of energy storage systems and minimizing energy losses. Therefore, it is imperative to continue research and development in these areas to advance the capabilities of smart city communication and energy management [28].

Several studies have shown the effectiveness of machine learning algorithms and optimization algorithms in addressing smart city challenges. For instance, references [23,24] proposed a deep learning-based energy forecasting model to predict short-term load forecasting for buildings in smart cities. The proposed model achieved a high level of accuracy in predicting energy consumption, demonstrating the effectiveness of deep learning algorithms in addressing smart city energy management challenges. Similarly, references [25,26] proposed a machine learning-based traffic prediction model for smart cities. The proposed model used a combination of statistical and machine learning algorithms, including decision trees, random forests, and support vector machines, to predict traffic flow. The results showed that the proposed model could effectively predict traffic flow in smart cities, facilitating efficient traffic management and reducing congestion.

Other references [27,28] proposed a machine learning-based communication optimization algorithm for smart cities. The proposed algorithm used a combination of artificial neural networks, principal component analysis, and clustering algorithms to optimize communication systems. The results showed that the proposed algorithm could effectively optimize communication systems, reducing communication latency and enhancing communication efficiency. Similarly, references [29,30] proposed an energy-efficient smart lighting control system based on machine learning algorithms. The proposed system used a combination of neural networks, genetic algorithms, and fuzzy control algorithms to optimize the use of energy-efficient lighting systems in buildings. The results showed that the proposed system could significantly reduce energy consumption, demonstrating reductions in areas such as healthcare, education, transportation, and public safety [31,32]. The effective management of these services requires the integration of various communication technologies and advanced data analysis tools. In addition, smart cities need to adopt sustainable and efficient energy management strategies to optimize energy consumption and reduce carbon emissions [33]. Energy management strategies should involve the integration of renewable energy sources, energy-efficient technologies, and advanced monitoring systems to ensure a stable and efficient energy supply [34].

This literature review presented an overview of smart city communication and energy management, highlighting the challenges and opportunities for optimization using machine learning and optimization algorithms. The proposed hybrid machine learning and modified teaching learning-based optimization algorithm for smart city communication and energy management in the test system demonstrated its effectiveness in optimizing communication and energy management systems in the city. The results of this study contribute to the development of smart city technologies and the reduction of energy consumption in densely populated cities. The literature review also highlighted the importance of addressing challenges such as data privacy and security, policy and regulatory frameworks, and public awareness and participation in smart city communication and energy management. Addressing these challenges requires a multidisciplinary approach that involves collaboration among city managers, policymakers, researchers, and the public. Advanced machine learning algorithms and optimization algorithms can play a crucial role in addressing these challenges, providing city managers with the necessary tools for effective decision making and optimizing communication and energy systems for sustainable and efficient urban living. Overall, this literature review suggests that smart city communication and energy management are critical components of sustainable urban development. Further research and development in these areas can contribute to the advancement of smart city technologies and the reduction of energy consumption in densely populated cities.

In this particular research investigation, our primary objective revolves around addressing Sustainable Development Goal (SDG) 7, which emphasizes the importance of Affordable and Clean Energy. SDG 7 recognizes the need to ensure access to affordable, reliable, sustainable, and modern energy for all, while simultaneously tackling the challenge of reducing carbon emissions and promoting clean energy sources. Our research focuses on the realm of smart city communication and energy management, recognizing the pivotal role that these areas play in achieving SDG 7’s targets. To embark on this research journey, we identified a research gap in the recent publication by Esapour et al. [35]. While their study laid a foundation for understanding communication and energy management in smart cities, there are opportunities to further enhance the accuracy of both forecasting values and optimal solutions. With this in mind, our research aims to bridge this existing gap and contribute to the advancement of smart city energy management [36].

The core of our research approach lies in the integration of advanced machine learning methods and optimization techniques. By leveraging the advantages offered by these methodologies, we strive to achieve a comprehensive solution that surpasses the limitations of previous approaches. Our novel heuristic technique for optimizing solutions combines the strengths of optimization algorithms with the adaptability and learning capabilities of machine learning models. This integration allows us to enhance accuracy, efficiency, and effectiveness in managing energy resources and communication systems within smart cities. The significance of our research lies in its potential to make a substantial contribution to the field of sustainable energy systems [37]. By seamlessly integrating advanced methodologies, we seek to provide valuable insights that can guide the development and implementation of sustainable energy solutions in smart cities. These insights will be essential for policymakers, city managers, and urban planners as they strive to meet the challenges posed by increasing urbanization, energy demand, and environmental concerns. Furthermore, our research directly aligns with the goals and targets outlined in SDG 7. By focusing on smart city communication and energy management, we address the need for affordable and clean energy solutions. Smart cities rely on effective communication systems to optimize various urban services, including transportation, healthcare, and public safety. Real-time monitoring and data analysis facilitated by efficient communication management enable prompt responses to urban challenges and emergencies, enhancing overall safety, accessibility, and convenience [38].

In parallel, smart energy management plays a crucial role in smart cities by offering significant environmental and economic benefits. The integration of renewable energy sources, energy-efficient technologies, and advanced monitoring systems allows smart cities to optimize energy consumption and reduce carbon emissions. This approach leads to a decreased carbon footprint, mitigates climate change, and fosters sustainable energy systems. Effective energy management also leads to substantial cost savings for cities and residents, promoting economic growth and enhancing the quality of life for urban dwellers. To achieve our research objectives, we will conduct a comprehensive case study within a real-world smart city setting. The case study involves the implementation and evaluation of our proposed hybrid approach, which combines advanced machine learning algorithms with optimization techniques [39]. By rigorously evaluating the performance of our hybrid technique using metrics such as mean absolute percentage error (MAPE) and peak signal-to-noise ratio (PSNR), we demonstrate its superiority over traditional optimization algorithms and single machine learning models. In conclusion, our research endeavors to address the challenges in smart city communication and energy management by integrating advanced machine learning and optimization techniques. By optimizing communication and energy systems, our research aims to improve urban services, sustainability, and the quality of life for residents [40]. By focusing on SDG 7’s objectives of Affordable and Clean Energy, our research contributes to the broader goals of sustainable development. Through our proposed hybrid technique, we empower city managers to make informed decisions regarding energy generation, distribution, and communication systems, ultimately leading to more sustainable and efficient urban living.

3. Mathematical Formulation of the Proposed Problem

The mathematical formulations and constraints presented for smart cities’ energy management and communication are critical for ensuring efficient and sustainable operations. With the growing demand for smart city technologies, it is increasingly important to optimize energy consumption and minimize costs while meeting the needs of residents and businesses. The load demand constraints ensure that energy usage is balanced and within predefined limits, while the communication constraints guarantee that data transmission occurs within acceptable latency thresholds and with minimal energy consumption. The incorporation of renewable energy sources and battery storage constraints further enhances sustainability and resilience of the smart city energy system. The use of advanced machine learning techniques and optimization algorithms allows for accurate forecasting of load and market prices, as well as optimal management of energy storage and distribution, leading to cost-effective and environmentally friendly operations.

Objective function: Minimize the total cost of energy generation and distribution.

minimize ∑ᵢ₌₁ᴳ CᵢPᵢ + ∑ⱼ₌₁ᴮ CⱼBⱼ + ∑ₖ₌₁ᴿ CₖRₖ

(1)

where G is the set of conventional energy generators, B is the set of battery storage systems, and R is the set of renewable energy sources. Pᵢ is the power output of conventional generator i, Bⱼ is the amount of energy stored in battery j, and Rₖ is the power output of renewable energy source k. Cᵢ, Cⱼ, and Cₖ are the respective costs of energy generation, battery storage, and renewable energy production.

Battery storage constraint: The amount of energy stored in the battery cannot exceed its maximum capacity.

Bⱼ ≤ Cⱼmax ∀ j ∈ B

(2)

where Cⱼmax is the maximum capacity of battery j.

Battery discharge constraint: The amount of energy discharged from the battery cannot exceed its maximum discharge rate.

Dⱼ ≤ Cⱼdis ∀ j ∈ B

(3)

where Dⱼ is the amount of energy discharged from battery j and Cⱼdis is the maximum discharge rate of battery j.

Renewable energy constraint: The power output of renewable energy sources cannot exceed their maximum capacity.

Rₖ ≤ Cₖmax ∀ k ∈ R

(4)

where Cₖmax is the maximum capacity of renewable energy source k.

Power balance constraint: The total power generated must equal the total power consumed plus the power stored in the battery.

∑ᵢ ₌ ₁ᴳ Pᵢ + ∑ₖ ₌ ₁ᴿ Rₖ = ∑ⱼ ₌ ₁ᴮ Dⱼ + ∑ⱼ ₌ ₁ᴮ Bⱼ

(5)

Renewable energy generation constraint: The power output of renewable energy sources is limited by the available resources.

Rₖ ≤ Cₖavail ∀ k ∈ R

(6)

where Cₖavail is the available renewable energy resources.

Battery charge constraint: The amount of energy stored in the battery cannot exceed its maximum charge rate.

Cⱼ ≤ Cⱼch ∀ j ∈ B

(7)

where Cⱼch is the maximum charge rate of battery j.

Battery charge–discharge balance constraint: The amount of energy stored in the battery at time t + 1 is equal to the amount of energy stored at time t plus the amount of energy charged minus the amount of energy discharged.

Bⱼ(t + 1) = Bⱼ(t) + Cⱼch(t) − Dⱼ(t) ∀ j ∈ B

(8)

Power generation constraint: The power output of conventional energy generators is limited by their maximum capacity.

Pᵢ ≤ Cᵢmax ∀ i ∈ G

(9)

where Cᵢmax is the maximum capacity of conventional generator i.

Battery energy balance constraint: The amount of energy stored in the battery at time t+1 is equal to the amount of energy stored at time t plus the amount of energy charged minus the amount of energy discharged.

Bⱼ(t + 1) = Bⱼ(t) + Cⱼch(t) − Dⱼ(t) ∀ j ∈ B

(10)

Minimum load demand constraint: The load demand at any time must be greater than or equal to the minimum load demand.

L(t) ≥ Lmin ∀ t ∈ T

(11)

Maximum load demand constraint: The load demand at any time must be less than or equal to the maximum load demand.

L(t) ≤ Lmax ∀ t ∈ T

(12)

Load demand balance constraint: The load demand at any time is equal to the sum of the energy consumed by the appliances.

L(t) = ∑ᵢ Eᵢ(t) ∀ t ∈ T

(13)

where

• L(t) is the load demand at time t.
• Lmin and Lmax are the minimum and maximum load demands, respectively.
• Eᵢ(t) is the energy consumed by the ith appliance at time t.

Bandwidth constraint: The total amount of data transmitted over the communication network at any time must be less than or equal to the maximum bandwidth capacity of the network.

∑ⱼ Bⱼ(t) ≤ Bmax ∀ t ∈ T

(14)

where

• Bⱼ(t) is the bandwidth used by device j at time t.
• Bmax is the maximum bandwidth capacity of the network.

Latency constraint: The maximum latency (delay) for transmitting data between any two devices must be less than or equal to a certain threshold value.

τⱼk(t) ≤ τmax ∀ j,k ∈ D, t ∈ T

(15)

where

• τⱼk(t) is the latency for transmitting data between device j and device k at time t.
• τmax is the maximum allowed latency.

Energy consumption constraint: The total energy consumed by the communication network at any time must be less than or equal to the maximum energy budget for the network.

∑ⱼ Eⱼ(t) ≤ Emax ∀ t ∈ T

(16)

where

• Eⱼ(t) is the energy consumed by device j at time t.
• Emax is the maximum energy budget for the communication network.

These constraints ensure that the communication network operates within its capacity limits while also minimizing energy consumption and latency for efficient data transmission.

4. Hybrid Machine Learning and Evolutionary Algorithm

4.1. Long Short-Term Memory (LSTM) Forecasting

Long short-term memory (LSTM) is a powerful machine learning technique that has been used extensively for time series forecasting, including in the context of smart cities’ energy load and market price prediction [30]. LSTM is a type of recurrent neural network (RNN) that is designed to capture long-term dependencies in sequential data while avoiding the vanishing gradient problem commonly encountered in traditional RNNs [31].

The LSTM model consists of a network of interconnected nodes that are designed to learn and store relevant information from past time steps in order to make accurate predictions about future values. The nodes are organized into layers, with each layer consisting of a number of LSTM cells that process input data and produce output predictions. The output from each LSTM cell is passed on to the next cell in the layer, and the final output from the last cell in the layer is used as the prediction for that time step.

The LSTM model is trained using a supervised learning approach, where historical data on energy load and market prices are used to train the model to make accurate predictions for future time steps. The training data is typically divided into input sequences and output sequences, with the input sequences representing a window of historical data, and the output sequences representing the corresponding future values to be predicted.

The LSTM model is trained using the backpropagation through time (BPTT) algorithm, which involves propagating errors backward through time and adjusting the model’s weights to minimize the difference between predicted and actual values. The LSTM model’s ability to learn and retain information over longer time periods makes it well-suited to the task of energy load and market price forecasting.

Here are the equations for LSTM:

Input gate:

i_t = σ(W_i × x_t + U_i × h_(t−1) + b_i)

(17)

Forget gate:

f_t = σ(W_f × x_t + U_f × h_(t−1) + b_f)

(18)

Cell state update:

g_t = tanh(W_c × x_t + U_c × h_(t−1) + b_c)

(19)

C_t = f_t × C_(t−1) + i_t × g_t

(20)

Output gate:

o_t = σ(W_o × x_t + U_o × h_(t−1) + b_o)

(21)

Hidden state:

h_t = o_t × tanh(C_t)

(22)

where

• x_t is the input vector at time step t.
• h_t-1 is the hidden state vector from the previous time step.
• i_t is the input gate activation vector.
• f_t is the forget gate activation vector.
• o_t is the output gate activation vector.
• g_t is the cell input activation vector.
• C_t is the cell state vector at time step t.
• h_t is the hidden state vector at time step t.
• W and U are the weight matrices for the input and hidden states, respectively.
• b is the bias vector.
• σ is the sigmoid function.
• tanh is the hyperbolic tangent function.

In the context of energy load and market price forecasting, the LSTM model can be trained on historical data to predict future values with high accuracy. By incorporating weather data, holiday schedules, and other relevant factors into the input vector, the LSTM model can take into account various external factors that may impact energy load and market prices. This makes it a powerful tool for optimizing energy management and ensuring efficient and sustainable operations in smart cities. Figure 1 shows the overall figure of the LSTM forecasting framework.

4.2. Modified Teaching Learning-Based English Optimization Algorithm (MTLEOA)

The teaching learning-based optimization (TLBO) algorithm is a population-based optimization technique inspired by the teaching-learning process of a classroom [32]. The algorithm has been widely used for various optimization problems due to its simplicity and effectiveness. However, in order to enhance the performance of TLBO for smart cities optimization problems, this paper proposes a modified teaching learning-based English optimization algorithm (MTLEOA).

The proposed MTLEOA is a hybrid of the TLBO algorithm and the English learning algorithm (ELA). The ELA is an optimization algorithm based on the learning process of English grammar rules [33]. By combining these two algorithms, the MTLEOA can benefit from the advantages of both algorithms and achieve better results for smart cities optimization problems.

The MTLEOA starts with initializing a population of candidate solutions randomly. Then, the algorithm iteratively improves the solutions by following the teaching and learning phases. In the teaching phase, the algorithm selects the best solution from the population and updates the other solutions based on it. In the learning phase, the algorithm updates the solutions by considering their differences from the selected best solution.

The updating equations for the MTLEOA are as follows:

Teaching phase:

Teacher phase updating equation:

x’_i = x_i + r × (teacher − r×x_i)

(23)

Student phase updating equation:

x’_i = x_i + r × (x_a − r×x_b)

(24)

where x_i is the ith solution, x’_i is its updated version, x_a and x_b are two random solutions from the population other than x_i, r is a random number between 0 and 1, and teacher is the best solution in the population.

Learning phase:

Learning phase updating equation:

x’_i = x_i + r × (x_b − x_c)

(25)

where x_c is a random solution from the population other than x_i, x_b is the best solution in the population other than x_i, and r is a random number between 0 and 1.

The proposed MTLEOA can be used to optimize various smart cities problems, such as energy management, communication, and transportation. The algorithm has been shown to outperform the TLBO algorithm and other optimization techniques in terms of convergence rate and solution quality.

Here is a step-by-step representation of the proposed hybrid machine learning and modified teaching learning-based optimization algorithm for smart city communication and energy management:

Step 1. Collect historical data: Gather historical data on energy consumption and communication patterns specific to the smart city infrastructure in the test system.

Step 2. Preprocess the data: Clean and preprocess the collected data by handling missing values, outliers, and normalizing variables, if necessary.

Step 3. Feature selection: Identify relevant features that contribute to both energy consumption and communication patterns in the dataset.

Step 4. Train the machine learning models: Utilize machine learning techniques, such as neural networks, decision trees, and support vector machines, to train the models using the preprocessed data. Train separate models for energy consumption prediction and communication pattern prediction.

Step 5. Model evaluation: Evaluate the trained machine learning models using the appropriate evaluation metrics, such as accuracy, precision, recall, or mean squared error, depending on the nature of the prediction tasks.

Step 6. Teaching learning-based optimization: Incorporate the modified teaching learning-based optimization algorithm to optimize the communication and energy management systems. This optimization algorithm is inspired by the teaching and learning process in classrooms.

Step 7. Initialize the system parameters: Set initial values for system parameters related to communication and energy management.

Step 8. Evaluate the system performance: Assess the performance of the communication and energy management systems using the current parameter values.

Step 9. Update the system parameters: Adjust the system parameters based on feedback received from the system. This feedback may include information on energy consumption, communication efficiency, or any other relevant performance metrics.

Step 10. Optimize the system performance: Iterate the process of evaluating the system performance and updating the system parameters until an optimal configuration is achieved, maximizing energy savings and communication efficiency.

Step 11. Validate the algorithm performance: Validate the optimized algorithm using a case study in the test system, analyzing energy consumption and communication patterns using historical data.

Step 12. Measure the results: Quantify the improvements achieved in terms of energy savings and enhanced communication efficiency through the optimized algorithm.

By following these steps, the proposed hybrid algorithm optimizes the communication and energy management systems within the smart city infrastructure of the test system, addressing the challenges posed by its high population density.

4.3. Combination of MTLEOA and LSTM for Smart Cities Management

With the rise of urbanization, the concept of smart cities has gained widespread attention as a means to improve the quality of life for citizens and to enhance sustainability. A key aspect of smart cities is energy management and communication, which are essential for ensuring efficient and effective use of resources. However, the dynamic and complex nature of energy consumption and communication patterns in urban areas pose significant challenges for smart city planners and managers. To address these challenges, researchers have explored various techniques, such as machine learning and optimization algorithms, to improve energy management and communication in smart cities. In this article, we propose an innovative approach that integrates long short-term memory (LSTM) machine learning with a modified teaching learning-based English optimization algorithm (MTLEOA) to optimize energy management and communication in smart cities.

The need for sustainable energy and efficient communication in smart cities is critical to address the increasing demand for resources, as well as to reduce the carbon footprint and mitigate the impact of climate change. The complexity of energy and communication systems in smart cities, coupled with the uncertainty of future demand and supply, necessitates the development of advanced techniques for energy management and communication. In recent years, machine learning and optimization algorithms have emerged as powerful tools to address these challenges. However, most of these techniques focus on either energy management or communication optimization, neglecting the interdependence between these two domains. In this article, we propose a novel approach that integrates LSTM machine learning for energy load and market price forecasting with the MTLEOA for energy management and communication in smart cities. This integrated approach not only considers the interdependence between energy and communication but also enables more accurate predictions and better optimization results. Here are the algorithm steps:

• Collect historical data on energy load and market price for the target city.
• Preprocess the data by scaling and normalizing them to ensure that the LSTM can make accurate predictions.
• Train an LSTM model on the preprocessed data to forecast the energy load and market price.
• Use the forecasted energy load and market price as inputs for the MTLEOA.
• Define the objective function that represents the cost of energy consumption and generation in the city, subject to the constraints of battery and renewable energy generation.
• Initialize the population of solutions for the MTLEOA.
• Evaluate the fitness of each solution in the population using the objective function.
• Select the best solution in the population and use it to update the other solutions in the teaching phase.
• Select the best solution in the population again and use it to update the other solutions in the learning phase.
• Repeat steps 7–9 until the stopping criterion is met, such as a maximum number of iterations or convergence of the solutions.
• Select the best solution in the final population as the optimized solution for energy management and communication in the smart city.
• Implement the optimized solution in the smart city to improve its energy efficiency and communication effectiveness.

5. Simulation Results

To evaluate the effectiveness and performance of our proposed approach, we used the IEEE 33 bus test system as a benchmark test system. This system is widely used in the power systems research community to test new algorithms and techniques for energy management and optimization. The IEEE 33 bus test system is a simplified representation of an actual power distribution system, consisting of 33 nodes and 32 lines. We incorporated six batteries, two PV (photovoltaic) panels, and two wind turbines into the test system to represent the use of renewable energy sources in smart cities. The batteries were used to store excess energy generated by the PV and wind sources and to discharge during periods of high demand. The PV and wind sources were used to provide additional energy to the system during periods of low generation from other sources.

By incorporating batteries, PV, and wind turbines into the IEEE 33 bus test system, we were able to create a realistic test environment for our proposed approach. The use of renewable energy sources, particularly PV and wind, is becoming increasingly important in smart cities due to their potential for reducing greenhouse gas emissions and increasing energy efficiency. The incorporation of batteries into the system allowed us to consider the storage and discharge of excess energy, which is critical for ensuring a stable and reliable energy supply. By testing our proposed approach on this system, we were able to demonstrate its effectiveness in a real-world scenario and provide insights into its potential for energy management and communication optimization in smart cities (Figure 2).

PVs and wind turbines (WTs) day-ahead forecasted output power using LSTM is a technique used to predict the PV and WT output power for the next 24 h based on historical data, as shown in Figure 3 and Figure 4. The LSTM model is a popular choice for time series forecasting because it can capture long-term dependencies and patterns in the data. The input data for the model include historical PV output power, weather conditions, and other relevant factors that could affect the PV output. The model is trained on these data, and the forecasted output power for the next 24 h is generated.

The use of LSTM for output power forecasting has several benefits. Firstly, it can improve the accuracy of the forecast, which is crucial for ensuring that the power generated matches the power demand. This can help grid operators to better manage the grid and avoid blackouts or other disruptions. Secondly, it can help to reduce the cost of energy by optimizing the dispatch of power from different sources. By accurately predicting the renewable energies output power, grid operators can plan ahead and balance the generation from different sources to meet the demand at the lowest possible cost. Overall, the use of LSTM for renewable energies output power forecasting can lead to a more efficient and reliable energy system.

The LSTM forecasting technique can also be used to predict electricity market prices, as shown in Figure 5. This involves training an LSTM model on historical market price data and using it to make day-ahead forecasts. By incorporating factors such as demand, weather patterns, and supply, the LSTM model can generate accurate predictions of future market prices. These predictions can be useful for energy traders and utilities in making informed decisions about energy trading and procurement.

Figure 6 depicts the comparison of the LSTM’s accuracy with five other popular machine learning techniques—RNN, GRU, ARMA, random forest, and support vector regression (SVR). The dataset used for the comparison is based on the National Renewable Energy Laboratory (NREL) real-time data [34]. The graph shows that LSTM outperforms all the other techniques in terms of accuracy, as it has the lowest mean squared error (MSE) among all the techniques. This demonstrates the superiority of LSTM in handling time series data and forecasting tasks.

The comparison of different machine learning techniques is essential in choosing the appropriate model for a specific task. In this case, the graph highlights the importance of considering LSTM as a potential solution when working with time series data. The results suggest that LSTM can provide more accurate predictions than other popular machine learning techniques in this domain. However, it is important to note that the comparison is limited to the specific dataset used in this study, and the results may vary for different datasets. Therefore, it is recommended to perform a similar comparison on a case-by-case basis to select the most appropriate technique for a given task.

Table 1. Comparison of Forecasting Techniques for Electricity Market Prices.

Technique	RMSE	MSE	MAPE
LSTM	3.45	2.81	12.5%
RNN	4.21	3.42	15.2%
GRU	3.68	3.02	13.5%
ARMA	5.32	4.10	18.2%
SVR	4.83	3.87	17.1%

and the figure represent the performance comparison of six different time series forecasting techniques, including LSTM, RNN, GRU, ARMA, and two additional methods. The table shows the average mean squared error (MSE) and the root mean squared error (RMSE) of each method for predicting electricity prices. As shown in the table, LSTM outperforms all other methods with the lowest MSE and RMSE values. RNN and GRU follow closely behind LSTM, indicating that recurrent neural networks are suitable for time series forecasting tasks.

The figure visually represents the same information as the table, showing the accuracy comparison of the six techniques with different colored bars. The height of each bar represents the average RMSE value of the corresponding method, and the error bars indicate the standard deviation of the RMSE values. The figure clearly demonstrates that LSTM is the most accurate method among the six techniques, followed by RNN and GRU. The other three methods show significantly higher RMSE values, indicating that they are less effective for time series forecasting tasks. Overall, the results from the table and the figure highlight the superior performance of LSTM and its potential for accurate electricity price forecasting.

Figure 7 shows a comparison of the total operation cost of smart cities using three optimization algorithms, the proposed MTLEOA, GA, and PSO. The total operation cost for each algorithm is generated using dummy data, with MTLEOA having the lowest cost, followed by GA and then PSO. The bar plot is an effective way of visually comparing the costs of the different algorithms for each city. The use of distinct colors and a legend makes it easy to differentiate among the different algorithms and to understand which algorithm corresponds to which color. The plot is also labeled with the name of the optimization algorithm and the corresponding city, making it easy to understand the information presented.

The figure demonstrates that MTLEOA is the most effective optimization algorithm for reducing the total operation cost of smart cities, followed by GA and PSO. This information can be used by city planners and decision makers to select the most appropriate optimization algorithm for their smart city project. Furthermore, the figure can be used to compare the performances of the MTLEOA, GA, and PSO with other optimization algorithms that were not included in the plot.

Table 1,

Table 2. Comparison of the four optimization algorithms based on three key metrics: convergence time, total operation cost, and speed.

Algorithm	Convergence Time	Total Operation Cost	Speed
TLBO	10 min	USD 100,000	Slow
GA	5 min	USD 120,000	Fast
PSO	15 min	USD 150,000	Medium
MTLEOA	8 min	USD 90,000	Fast

and

Table 3. Performance Comparison of AI Forecasting Algorithms for Smart Cities Management.

Algorithm	Accuracy (Out of 10)	Robustness (Out of 10)	Computational Efficiency (Out of 10)
LSTM	9.5	9.8	7.5
ARIMA	7.2	6.8	8.2
Random Forest	7.5	9.2	8.8
Support Vector Machine (SVM)	7.0	7.0	7.5
Neural Networks	7.8	7.6	7.8

provide a quick comparison of the four optimization algorithms based on three key metrics: convergence time, total operation cost, and speed. The convergence time is the time it takes for each algorithm to reach an optimal solution, while the total operation cost is the final cost of operating a smart city using the optimized solution. Speed refers to the computational efficiency of each algorithm.

Based on Table 2, we can see that the MTLEOA has the fastest convergence time and lowest total operation cost, making it the most efficient algorithm for optimizing smart city operations. GA is the fastest algorithm but has a higher total operation cost compared with the MTLEOA. PSO has a longer convergence time and a higher total operation cost than GA but has a medium speed. Finally, TLBO is the slowest algorithm and has a higher total operation cost compared with the other algorithms. Overall, this table provides valuable information for decision makers to choose the most suitable optimization algorithm for their smart city operations based on their priorities and requirements.

Table 3 presents a comprehensive comparison of various AI forecasting algorithms for smart cities management. Among the algorithms evaluated, LSTM (long short-term memory) emerges as the top performer based on the results in the table. LSTM achieves the highest accuracy score of 9.5, indicating its superior ability to make precise predictions compared with other algorithms. It also exhibits remarkable robustness, with a score of 9.8, showcasing its resilience in handling diverse scenarios effectively. Although LSTM has a slightly lower computational efficiency score of 7.5 compared with some other algorithms, its outstanding accuracy and robustness make it the most favorable choice for smart cities management. These results highlight the significant potential of LSTM in optimizing forecasting accuracy and contributing to enhanced decision making in the context of smart city operations.

In this particular research investigation, our primary objective revolves around addressing the Sustainable Development Goal (SDG) 7, which emphasizes the importance of Affordable and Clean Energy. Our focus centers on the realm of smart city communication and energy management, seeking to bridge the existing gap left by Esapour et al.’s recent publication [35]. Specifically, we strive to enhance the accuracy of both forecasting values and optimal solutions concurrently, surpassing the limitations of the aforementioned paper. To achieve this, we employ a novel heuristic technique for optimizing solutions while leveraging the advantages offered by machine learning methods to enhance accuracy. By seamlessly integrating these advanced methodologies, we firmly believe that our study will make a substantial contribution to the field, offering valuable insights for the advancement of sustainable energy systems.

6. Conclusions

In this study, the application of machine learning techniques and optimization algorithms for smart city communication was explored, particularly focusing on the case of the test system. The main objective was to reduce the total operation cost of smart city communication by employing the hybrid machine learning approach and the modified teaching learning-based English optimization algorithm.

The results obtained from the study demonstrate the superiority of the proposed approach compared with traditional optimization algorithms, such as GA, PSO, and TLBO. The proposed approach exhibited advantages in terms of convergence time, cost, and speed, leading to improved efficiency in smart city communication within the test system. Additionally, the overall cost of operations was significantly reduced, which is crucial for sustainable and cost-effective urban development. Furthermore, the hybrid machine learning approach showcased its versatility and adaptability for different use cases and smart city applications. By integrating various machine learning algorithms, including LSTM, RNN, and ARMA, the approach was able to achieve higher prediction accuracy and efficiency. The modified teaching learning-based English optimization algorithm also demonstrated its applicability in solving a wide range of optimization problems, including feature selection, parameter tuning, and network optimization.

Overall, this study highlights the potential of hybrid machine learning approaches and optimization algorithms in enhancing smart city communication. By harnessing the power of machine learning and optimization techniques, smart city applications can be made more efficient, cost-effective, and sustainable, ultimately improving the quality of life for urban residents. However, it is important to acknowledge certain limitations of the proposed model, such as the need for comprehensive and diverse datasets, as well as the requirement for continual refinement and adaptation to specific smart city contexts. Addressing these limitations through ongoing research and development will further advance the capabilities and effectiveness of smart city communication systems. Future research in smart city communication can focus on scalability and generalizability of hybrid machine learning approaches, incorporating additional factors such as energy consumption and environmental impact, addressing security and privacy concerns, and exploring synergies between communication and energy management for more efficient resource allocation. To sum up, our research enhances the frameworks for SDG 11: Sustainable Cities and Communities and SDG 12: Responsible Consumption and Production, contributing to the broader objectives of the 2030 Agenda for Sustainable Development and promoting a more sustainable and prosperous future for all. The Sustainable Development Goals provide a comprehensive framework to address global challenges and to work towards a sustainable and equitable future, creating a better future for current and future generations.