1. Introduction
With the increasing depletion of fossil energy and the deterioration of the environment becoming more and more serious, the development and utilization of RESs is the main direction of future energy development. Distributed energy based on RESs such as wind power, solar energy, and biomass widely exist as clean energy, and development technology is becoming more and more mature, which has received wide attention around the world. Many countries have set future renewable energy development goals to promote sustainable energy development [
1].The share of RESs in Finland will increase to 38% by 2020 [
1]. The Chinese government has formulated four energy revolutions, aimed at achieving the goal of nonfossil energy accounting for 15% and 20% of primary energy consumption in 2020 and 2030, respectively [
2]. Meanwhile, in traditional power systems, a large part (40–60%) of fossil energy or renewable energy sources may be wasted because traditional power systems cannot effectively use the input energy [
3]. Further, the fluctuation of renewable energy output, market electricity prices, and load demand make it hard to elevate comprehensive energy utilization efficiency and reduce greenhouse gas emissions. In this regard, one of the effective solutions to solve these problems is the combined heating and power (CHP) microgrid. As an independent and controllable energy supply system, CHP microgrids can satisfy local energy demands and guarantee the safe and stable operation of a power system, reduce extra costs, and improve the environment [
4,
5]. However, with the huge penetration of renewable energy, the operation mode of the CHP microgrid has changed dramatically, which needs a new mode of operation to match the current situation.
The economic dispatch of a CHP microgrid with renewable energy has been the focus of significant efforts in the literature. Reference [
6] presents a multiobjective economic emission dispatch to minimize total operation cost and CO
2 emission cost for the microgrid that contains RESs. Reference [
7] developed an optimization dispatching model for CHP plants to minimize system operation costs and satisfy the end-consumers’ demands. For the CHP-direct heating(CHP–DH) system, the authors established energy balance, system control, and operation constraints in [
8], where the optimized objective function is to minimize system operation cost and net acquisition for heat and power. A stochastic optimization model considefing the economic and environmental impacts and reliability is proposed in [
9], whose ultimate goal is to minimize fuel cost. In order to solve the multiobjective economic dispatch of the microgrid, a particle swarm optimization (PSO) method and enhanced firefly algorithm (FA) have been adopted, respectively, in [
10,
11]. Reference [
12] developed an optimization model from both the environmental and economic viewpoints and proposes optinal operation and management strategy for the mincrogrid. Reference [
13] introduces an economic model and optimization method for the CHP system, including solar, wind, and fuel cell. The optimization goal is to minimize system operation costs and maintenance costs and compare the effects of genetic algorithms and particle swarm optimization algorithms in the model. In these references, although the fuel cost and the operation cost of the generator units are taken into account, the impact of environmental on the economic operation of CHP microgrids is rarely considered.
Further, several studies focus on the impact of auxiliary functions, such as demand response and batteries on CHP microgrid operation. Reference [
14] satisfies the electrical and thermal load demands of end-users with minimal cost by implementing demand response (DR) programs. Reference [
15] considers the energy storage system as a non-ideal battery and analyzes the impact of battery on the economic operation of CHP microgrid. Reference [
16] considers DR as a virtual generator, replacing user-side resources with supply-side energy, which can effectively improve energy efficiency and reduce energy costs. References [
17,
18] believe that under fluctuation of electricity prices, real-time electricity prices can improve the adaptability of DR and thus propose a price-based DR management model for real-time electric and thermal loads. However, most studies about DR focus on the power load but neglect the heat load. Therefore, it is necessary to propose a novel DR model.
In addition, the impact of volatility of renewable energy on grid shocks has not been well resolved. Some of the literature has studied the uncertainty processing method. Reference [
19] describes the uncertainty of wind power output through normal distribution but cannot accurately characterize wind power changes. A weighting method is applied to systematically combine multiple predictive models in [
20], which can reduce the prediction bias. References [
21,
22,
23,
24] deal with wind power uncertainty through scenario generation and reduction methods. The solution to the uncertainty of renewable energy output are relatively simple and quite limited. It is urgent to find new ways to reduce the impact of renewable energy output uncertainty on the power system.
Through the analysis of the previous literature, we summarize the structure and influencing factors of the CHP microgrid system proposed in several typical studies and compare them with this paper, as shown in
Table 1.
To deal with the problem mentioned above, we propose an optimal scheduling method based on stochastic optimization against the uncertainty of renewable energy output and consider the economy, environment, as well as system flexibility aspects in this paper. In addition, the operation mode of the CHP microgrid system and the behavior of system operators in different situations are discussed, and operational strategies are designed, respectively based on the interests of system operators and the environment. In short, the main contributions of this paper are included as follows.
- (1)
Considering system operation cost, carbon dioxide emissions cost, as well as system flexibility in the objective function.
- (2)
Based on the fuzzy C-means (FCM) clustering method, the number of historical scenarios of wind power and solar energy output can be reasonably reduced to a fixed value. Then, a novel clustering evaluation method named CCQ is proposed to determine the number of best scenario categories, which can reduce the uncertainty of renewable energy output compared with previous literature methods.
- (3)
A new demand side management method called integrated energy demand response is employed. Compared with a single electrical or heat load demand response method, the flexibility of the CHP microgrid system is improved, and the system operation cost is reduced significantly.
- (4)
The operation mode of the CHP microgrid and the behavior of the system operator in different situations are discussed, and different operational strategies are put forward in the light of the interests of system operators and environmental benefits.
The rest of proposed paper is organized as follows:
Section 2 mainly describes the structure of the CHP microgrid and its mathematical model.
Section 3 presents the objective functions and constraints of the CHP microgrid, and the solution to the proposed model. Case studies and simulation results as well as the corresponding operational strategy are explained in
Section 4.
Section 5 summarizes the main conclusions in this paper.
3. CHP Microgrid Optimization Scheduling Strategy
Stochastic programming is a method for dealing with uncertain decision problems and has been widely used in the field of power systems. This paper takes the output power of CHP units and the renewable energy as the decision variables, via converting stochastic programming to mixed linear programming (MILP) and solving it with MATLAB R2016b (MathWorks company, Natick, MA, USA, 2016) and CPLEX Optimization Studio v12.8(IBM company, Amund, New York, NY, USA, 2017).
3.1. Objective Function
The operation cost of the CHP microgrid system, which contains the power exchange cost with the main network, the maintenance cost of each unit, the CO
2 emission cost, and natural gas procurement cost, should be minimized. The mathematical model of the objective function is shown below (14):
where
is the exchange power cost.
is the natural gas consumption cost.
is the gas turbine maintenance cost.
,
,
,
, and
represent the maintenance cost of each generating unit, respectively.
is the CO
2 emissions cost.
is the IEDR compensation cost.
3.1.1. The Cost of Exchange Power between the Main Network and the CHP Microgrid
where
and
are electricity procurement/sales costs from or to the main network.
and
are electricity procurement price and electricity sales price, respectively.
3.1.2. Natural Gas Consumption Cost
where
represents natural gas price.
and
represent the amount of natural gas consumed by gas turbines and gas boilers in interval t.
3.1.3. The Maintenance Cost of Each Unit
where
,
,
,
,
, and
indicate the unit power maintenance cost of each unit in the CHP microgrid in interval t, respectively.
3.1.4. Carbon Dioxide Emission Cost
where
represents carbon tax.
,
, and
indicate carbon dioxide emissions per unit of power from the GT, GB, and main network.
3.1.5. Integrated Energy Demand Response Compensation Cost
where
and
represent the compensation price of the unit electric load/thermal load in the demand response program.
3.2. CHP Microgrid System Constraints
This section mainly describes the CHP microgrid system constraints, including the balance of electrical energy and thermal energy, as well as the technical constraints of the units.
The technical constraints of the units are described in
Section 2.2 in detail. Please see Equations (1)–(11) for specific constraints.
3.3. RESs Uncertainty Sets
Since WT and PV power generation are subject to meteorological conditions, their power output has certain intermittency and volatility. In order to characterize the WT and PV, this paper adopted the improved fuzzy C-means (FCM) to cluster the historical power scenarios of wind power and solar energy out into several typical scenarios and then used the clustering comprehensive quality (CCQ) method to evaluate whether these scenarios are representative and determine the optimal number of scenarios [
26].
3.3.1. FCM Clustering Method
In this paper, the objective function was established with all the same scenarios as the minimum total distance from the cluster center:
where
represents the total number of scenarios;
is the number of clusters,
is the relative membership of the sample
to the category
; and
,
,
;
is the Euclidean distance between sample
and category
;
is the weight of the
-th indicator, and
;
is the clustering feature normalization number of the category
indicator feature value
,
.
Usually, there are differences in the dimension and magnitude of each indicator in the system, so each indicator needs to be normalized, then
is updated to:
where
and
are the maximum and minimum values of the
-th indicator eigenvalues in the scenario sets, respectively.
Finally, via constructing the Lagrange function, the cluster iteration can be obtained, and the loop iteration can realize the scenario clustering.
3.3.2. CCQ Evaluation Method
The appropriate number of clustering scenario categories can accurately reflect the characteristics of wind power and photovoltaic output and has a significant impact on the solution effect of the model. Thus, we adopted the CCQ method to determine the best scenario category.
The clustering effect can be expressed by the variance within the cluster. The larger the variance, the worse the clustering effect, and the smaller the variance, the better the clustering effect.
where
is the total number of scenes of set
;
is the sample mean value corresponding to the eigenvalue of an index; and
is the value of this indicator in the
-th scenario.
The distance between members within each cluster should theoretically be as close as possible, so the smaller the cluster density, the better.
where
is the sum scenarios after clustering;
represents the variance in category
.
The clustering proximity and the distance between clusters are inversely proportional. The smaller the clustering proximity, the more effectively each cluster can be separated, indicating that the classification effect is better.
where
is the clustering center in category
;
is a gaussian constant and
for simplified calculation.
In summary, this paper linearly combines clustering density and proximity, that is, the CCQ method, which can be expressed as:
where
is the weight of equilibrium clustering density and clustering proximity, which is 0.5 in this paper. The average comprehensive quality evaluation of each index
determines the optimal number of clustering, as shown in the following formula:
The solution procedure of the WT and PV scenario clustering and CHP microgrid optimization scheduling strategy is demonstrated in
Figure 4.
5. Conclusion
This paper mainly studied the operation mode of the CHP microgrid system with RESs and the behavior of system operators in different scenarios to improve the efficiency of RES use and proposes the operating strategy for system operators under the power market. Considering the battery and IEDR, a MILP model was utilized to optimize the CHP microgrid with RESs. Firstly, considering the uncertainty and volatility of WT and PV output, the FCM and CCQ clustering evaluation method were used to cluster the WT and PV historical power output data into the optimal scenario number, and then the clustering results were comprehensively evaluated to obtain the best typical scenario. Secondly, the MILP method was used to simulate the optimized scheduling of the CHP microgrid in different scenarios, and the operating status of each unit, total system cost, CO2 emission cost, and integrated demand response compensation cost in the system were obtained. Finally, the behavior of the system operator in different scenarios was analyzed. The results indicate that the battery and IEDR can increase system flexibility and reduce system operating cost. In the grid-connected mode, a shrewd system operator can profit handsomely by making an optimal operation strategy and trading electricity at the time-of-use price. In this regard, more social capital will be introduced to invest in the development of RESs in the future, which is conducive to environmental improvement and sustainable social development. Compared with previous works, the main research in this paper was as follows: (1) an economic operation strategy of the CHP microgrid with RESs considering CO2 emission, battery, and IEDR was conducted; (2) a novel scene clustering method (FCM–CCQ) was adopted, which can not only cluster but also evaluate whether the scene clustering number is the optimal result; and (3) considering the uncertainty of renewable energy output.
Additionally, we proposed five different operating modes for the CHP microgrid and obtained the results by simulation. After analyzing the results, the strategy choices of the system operators under different circumstances were presented.
Finally, there are some shortcomings in this paper, which are that we did not consider the start–stop cost of generator units, the climbing slope constraint, and the model being idealized. However, it is hoped that the work can provide a useful reference for CHP microgrid economic optimization scheduling containing RESs. In future research, the influence of various factors will be comprehensively considered, making the model construction closer to the actual situation.
This paper can provide useful reference for CHP microgrid economic optimization scheduling containing RESs. As a future work, the influence of various factors will be given more comprehensive consideration, making the model construction more dynamic and accurate. Further, the start–stop cost of generator units and the climbing slope constraint can be employed in this model.