Load aggregation regulation and control method integrated with improved genetic algorithm and game combination weighting method

Abstract

The invention belongs to the field of power system optimization scheduling, and discloses a load aggregation regulation and control method integrated with an improved genetic algorithm and a game combination weighting method, which comprises the following steps: introducing a load step, and preprocessing power consumption data of industrial users; an industrial user load response model is constructed, and an optimal user load response excitation is solved by using an improved genetic algorithm; constructing an industrial user load response potential comprehensive evaluation system; using a triangular fuzzy number analytic hierarchy process and an improved entropy weight method to carry out combined weighting on indexes of the comprehensive evaluation system; comprehensively evaluating the load response potentials of the industrial users by using a TOPSIS method, and clustering according to the user load response potentials by using a K-means algorithm to realize division of feature groups; power grid load adjustment is realized by using industrial users with different types of load response potentials, the flexibility of power grid load adjustment is improved, and the stability of a power system is improved.

Description

Technical Field

[0001] The present invention belongs to the field of optimal dispatching of power systems, and particularly relates to a load aggregation regulation method integrating an improved genetic algorithm and a game combination weighting method. Background Art

[0002] Industrial load, as the load with the largest total electricity consumption, accounts for more than 60% of the total electricity consumption of the whole society every year. However, industrial loads are complex and diverse, and there are significant differences in the production status among enterprises. It is impossible to obtain the group adjustable potential based on representative users. Therefore, it is necessary to select some users with greater response potential for regulation. By guiding industrial users to participate in peak shaving, it can not only relieve the pressure of shortage of flexible resources in the system, but also help users obtain response benefits and improve their enthusiasm for participating in response.

[0003] Currently, existing research mainly focuses on the dispatching strategies and decision-making models for industrial users to participate in demand response, but there is little research on the clustering and classification of industrial user demand response potential and user optimization.

[0004] In the domestic research on the evaluation of the potential of industrial users to participate in demand response, some scholars evaluate the comprehensive performance of industrial users to participate in regulation from multiple aspects such as the technical level and potential level, and list in detail the benefits obtained from participating in the response. Some scholars use the method of secondary clustering to extract data features, evaluate the user response potential from three aspects of the interruptible potential, transferable potential, and production potential of users, and verify the superiority of the method through comparative experiments. Some scholars, in order to balance individual preferences, use the fuzzy rough entropy weight method and the stepwise weighted evaluation ratio analysis method to calculate the index weights, and integrate the fuzzy rough numbers into the improved multi-attribute boundary approximation region comparison method to calculate the load response potential, reducing the influence of judgment ambiguity and individual preferences on the results.

[0005] However, when weighting various indicators, due to the large differences in interest preferences and thinking modes among individuals, it is difficult to ensure the consistency and objectivity of thinking during weighting.

[0006] In summary, although there has been a lot of research on industrial loads currently, there are still the following deficiencies: (1) The existing evaluation models for the peak shaving potential of industrial loads are difficult to comprehensively consider the production characteristics of user loads from all aspects, affecting the accuracy of the final evaluation results.

[0007] (2) In the research on the evaluation of industrial adjustable potential, there is currently little combination of capacity allocation and user optimization to achieve optimal allocation of capacity for users.

[0008] Difficulty in Solving the Above Technical Problems: The safe operation of the power system is closely related to the reliable supply of electricity. However, with the establishment of a new power system, the problem of shortage of flexible resources in the system has become increasingly prominent. Therefore, flexible resources should be obtained from industrial loads on the load side to ensure the safe and stable operation of the system. In the process of industrial users participating in peak shaving, how to reasonably allocate response capacity according to the response potential of users is a difficult point. In addition, it is also a top priority to complete system response with the minimum cost while meeting the system regulation tasks. Summary of the Invention

[0009] The purpose of the present invention is to address the above problems by providing a load aggregation regulation method integrating an improved genetic algorithm and a game combination weighting method, constructing a scientific comprehensive evaluation system for the load response potential of industrial users from multiple aspects such as technical level, dispatching cost, and dispatchability; constructing an industrial user load response model, and using an AI model to solve for the optimal user load response incentive; combining the TOPSIS method with the K-means method to achieve accurate evaluation and classification of the load potential of industrial users, thereby differentially matching the grid regulation requirements, improving the accuracy and flexibility of demand response, and enhancing the stability of the power system.

[0010] To achieve the above object, the technical solution provided by the present invention is as follows: A load aggregation regulation method integrating an improved genetic algorithm and a game combination weighting method, comprising the following steps: Step 1: Introduce load steps to preprocess the electricity consumption data of industrial users; Step 2: Construct an industrial user load response model, and use the improved genetic algorithm to solve for the optimal user load response incentive according to the functional relationship between incentive and response; Step 3: Construct a comprehensive evaluation system for the load response potential of industrial users; Step 4: Use the triangular fuzzy number analytic hierarchy process and the improved entropy weight method to perform combined weighting on the indicators of the comprehensive evaluation system; Step 5: Combine the indicator weights obtained in Step 4, use the TOPSIS method to comprehensively evaluate the load response potential of industrial users, and use the K-means algorithm to cluster according to the user load response potential to achieve the division of characteristic groups; Step 6: According to the regulation requirements of the power system, use industrial users with different categories of load response potential to achieve grid load regulation, increase the flexibility of grid load adjustment, and improve the stability of the power system.

[0011] Further, in Step 1, the electricity consumption data of industrial users is preprocessed through load steps to remove redundant load fluctuations, extract effective information from the data, and obtain data characteristics; the load step refers to the user load curve with a relatively small local load change rate.

[0012] Local load change rate When it indicates that the load values of two adjacent points on the user load curve are at the same load step, that is, the user load is in a relatively stable state; when the local load change rate When it is, it indicates that the user's power consumption state is in an unstable state, representing different load levels caused by the switching of the overall equipment power consumption state during the actual production process of industrial users.

[0013] In step 2, the excitation-response function of the industrial user load response model is used to describe the functional relationship between user excitation and response. According to this functional relationship, the unit response excitation of the user can be obtained, which is convenient for allocating the response capacity.

[0014] Preferably, the excitation-response function of the industrial user load response model is:[[]] ; In the formula, is the response excitation of the user at time period , is the load amount participated in the response by the user at time period , is the random error brought by the user response volatility, and are the excitation-response function parameters in the current situation.

[0015] Preferably, in step 3, the indicators of the comprehensive evaluation system for the industrial user load response potential include load stability, annual production hours of the maximum load, peak-valley difference rate, power supply reliability, enterprise production characteristics, unit response excitation, response capacity, response time, response speed, order demand, and proportion of electricity cost expenditure.

[0016] Preferably, the step 4 specifically includes the following sub-steps: Step 4.1: Calculate the index weights of the comprehensive evaluation system for the industrial user load response potential by using the triangular fuzzy number analytic hierarchy process; Step 4.1.1: Establish a triangular fuzzy number judgment matrix, compare the indicators of the comprehensive evaluation system for the industrial user load response potential pairwise, and perform index scaling according to the difference in importance between the indicators to construct an n×n judgment matrix , where n is the number of indicators; Step 4.1.2: Calculate the geometric mean of the index according to the triangular fuzzy number in each row of the judgment matrix ; Step 4.1.3: Calculate the vector and its inverse vector of the geometric mean of each index p, according to and its inverse vector to obtain the fuzzy weight of index p ; Step 4.1.4: Defuzzification calculation to obtain the weight of each index p ; Step 4.1.5: Normalize the weight results of each index p after defuzzification to obtain the index weight of index p after normalization ; Step 4.2: Use the improved entropy weight method to calculate the index weights of the comprehensive evaluation system for industrial load response potential; Step 4.2.1: Establish an initial evaluation matrix based on the sample data of the comprehensive evaluation system , where m represents the number of rows of the initial evaluation matrix and n is the number of indicators; Step 4.2.2: Normalize the positive indicators and reverse indicators in the initial evaluation matrix H respectively; Step 4.2.3: Calculate the information entropy of each index p: Step 4.2.4: Obtain the weight of each index p determined by the improved entropy weight method according to the information entropy of the index; Step 4.3: Integrate the index weights calculated in Step 4.1 and Step 4.2 to obtain the comprehensive index weight.

[0017] Compared with the prior art, the beneficial effects of the present invention include: 1) The present invention constructs a comprehensive evaluation system covering multiple levels such as the technical level, scheduling cost, and schedulability, combines the TOPSIS method with the K-means method, realizes the accurate evaluation and classification of the load potential of industrial users, differentiates and matches the grid regulation requirements according to the load response potential of different types of users, can improve the accuracy and flexibility of demand response, enhance the stability of the power system; through refined load management, enhance the power system's ability to absorb the volatility of wind power and photovoltaic power; through accurately matching adjustable load resources, can improve the execution efficiency of demand response, reduce the demand for reserve capacity; the dynamic adjustment based on the response potential of industrial users can reduce the call of high-price peaking units and reduce the operating cost of the power system.

[0018] 2) The present invention constructs an industrial user load response model and uses the improved genetic algorithm to solve for the optimal user load response incentive, and the incentive strategy optimized by the improved genetic algorithm can improve the response accuracy of industrial users.

[0019] 3) The present invention introduces load steps to process the electricity consumption load data of users, removes load fluctuations, and uses the difference between adjacent load steps to describe the demand response ability of users. The load step is a load fluctuation removal algorithm for dynamic threshold filtering, which is different from the traditional fixed threshold method and can adaptively identify the "extra fluctuations" in the electricity consumption data of industrial users and retain the effective load characteristics.

[0020] 4) Based on the game theory method, the present invention fuses the weights of user potential indicators obtained by expert evaluation with the weights of user potential indicators obtained by the entropy weight method, achieving a balance between experience orientation and data-driven.

[0021] 5) The present invention ensures the fairness of user load incentives by constructing a scientific comprehensive evaluation system, which can improve user acceptance and promote the establishment of a long-term sustainable demand response mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention will be further described below in conjunction with the drawings and embodiments.

[0023] Figure 1 It is a schematic flow chart of the load aggregation regulation method provided by the embodiment of the present invention.

[0024] Figure 2 It is a diagram of the industrial load demand response potential evaluation system provided by the embodiment of the present invention.

[0025] Figure 3 It is a user incentive response diagram under ideal conditions provided by the embodiment of the present invention.

[0026] Figure 4 It is a comparison diagram of actual user response incentives provided by the embodiment of the present invention.

[0027] Figure 5 It is a user response potential diagram provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] As Figure 1 shown, the load aggregation regulation method integrating the improved genetic algorithm and the game combination weighting method includes: Step 1: Introduce load steps to preprocess the electricity consumption data of industrial users; In the embodiment, load steps are introduced to process the electricity consumption data, remove the excess load fluctuations, and extract the effective information of the data.

[0029] The load step is a data processing method that can describe the relatively stable state of the data, remove the excess load fluctuations, extract the effective information of the data, and the difference between two adjacent load steps can describe the demand response ability of the electricity consumption users.

[0030] A load step refers to a section of a user's load curve with a low load change rate lasting at least one hour.

[0031] Under normal circumstances, the local change rate indicates that the load values of two adjacent points can be regarded as being in the same load step, that is, the load is in a relatively stable state for a period of time. When the local change rate it indicates that the production state of the current user is in an unstable state, which represents different load levels caused by the switching of the overall equipment power consumption state during the actual production process of industrial users.

[0032] Under normal circumstances, the local change rate indicates that the load values of two adjacent points can be regarded as being in the same load step, that is, the load is in a relatively stable state for a period of time. When the local change rate it indicates that the production state of the current user is in an unstable state, which represents different load levels caused by the switching of the overall equipment power consumption state during the actual production process of industrial users. The local load change rate is: ; (1) ; (2) In the formula, α is the local change rate; p is the local load; , respectively represent the minimum and maximum power consumption loads of industrial users during the current time period; , are the minimum and maximum load changes at adjacent moments; represents the variance of the power consumption load of industrial users during the current time period.

[0033] The calculation steps of the load step are as follows.

[0034] 1) Define the initial load array and judgment conditions: Construct the initial load array according to the first 4 load data of the power consumption load time series data of industrial users, and use as the judgment condition for the load step; when the values of the first 4 load data are in the same load step, then take the average value of the first 4 load data as the initial value of the load step ; ; (3) 2) Iteratively update the load step value according to the power consumption load time series data of industrial users: ; (4) In the formula, , , , respectively represent the (t-3)th, (t-2)th, (t-1)th, and tth load data of the electricity consumption load time series data of industrial users, where t represents the time step; Update the load step value according to Equation (4). If is on the load step where is located, then ; if is not on the load step where is located, then construct a new load array using the tth, (t + 1)th, (t + 2)th, and (t + 3)th electricity consumption load data, and re-judge and calculate to obtain a new load step value a v , ; 3) Record the information of the th load step on the th day, ; (5) In the formula, respectively represent the average value, start time, and end time of the historical data of the th load step on the th day, and is the number of load steps within 15 days.

[0035] Therefore, the load step information matrix i on the th day is: ; (6) In summary, the load step information matrix obtained from the 15-day electricity consumption data of industrial users is: ; (7) Step 2: Construct an industrial user load response model. According to the functional relationship between incentives and responses, use an improved genetic algorithm to solve for the optimal user load response incentive; The incentive-response function of the industrial user load response model is used to describe the functional relationship between user incentives and responses. According to this functional relationship, the unit response incentive of the user can be obtained, which is convenient for allocating response capacity.

[0036] In the embodiment, the incentive-response function of the industrial user load response model is: ; (8) In the formula, is the response incentive of user at time period , and is the response incentive of user at time period The load capacity involved in the response at that time is the random error caused by the volatility of user response and are the incentive-response function parameters under the current situation

[0037] The improved genetic algorithm in the embodiment refers to the improved genetic algorithm disclosed in the paper "Microgrid Load Optimal Allocation Method Based on Improved Genetic Algorithm" by Tang Hongwei et al. published in the 2nd issue of "Electrical Technology and Economy" in 2025

[0038] Step 3: Construct a comprehensive evaluation system for the load response potential of industrial users, as Figure 2 shown

[0039] In the embodiment, the comprehensive evaluation system is divided into three levels: technical level, dispatching cost, and dispatchability. From the perspective of the enterprise's own load characteristics, the technical level quantifies the volatility and electricity demand of the load, helping the power grid to better understand the user load characteristics, so as to allocate control instructions more reasonably. The dispatchability level intuitively quantifies the load adjustable potential when the enterprise participates in demand response, helping the power grid to fully understand the enterprise's demand response ability, so as to allocate adjustment tasks according to the enterprise's response ability

[0040] The indicators of the comprehensive evaluation system are as follows (1) Load stability Load stability is a characteristic quantity that describes the change of load and reflects the adjustable potential of load according to the volatility of load over time. This indicator represents the fluctuation of the load. The larger the ratio, the more stable the load and the lower the adjustability

[0041] The load stability of industrial users The calculation formula is ; (9) In the formula is the load power value within 15 days is the maximum power during this period is the number of samples

[0042] (2) Annual production hours of the maximum load The annual production hours of the maximum load are used to describe the production status of the enterprise. Since different enterprises have different production methods and order situations, their annual production hours of the maximum load are also different. The larger the number of hours, the better the production status of the enterprise, and the lower the response reliability

[0043] The annual production hours of the maximum load of industrial users The calculation formula is ; (10) In the formula, is the annual electricity consumption, is the maximum power of the user's electricity consumption within a year.

[0044] (3) Peak-valley difference rate: The peak-valley difference rate is a characteristic quantity used to describe the load volatility, which is the ratio of the difference between the peak-valley average value to the peak value. The larger the value, the stronger the reliability of peak shifting and the greater the corresponding potential.

[0045] The peak-valley difference rate of industrial users The calculation formula is: ; (11) In the formula, is the average value of the maximum electricity consumption within 15 days, is the average value of the minimum electricity consumption within 15 days.

[0046] (4) Power supply reliability: Since different industries have different demands and dependencies on electricity for production, that is, the impact on production during power outages is different, the requirements for power supply continuity of different industries are also inconsistent. In addition, with the development of technology and the change of energy structure, the requirements for power supply reliability of each industry may also change.

[0047] (5) Enterprise production characteristics: Enterprise production characteristics are characteristic quantities used to judge the adjustable potential according to the production type of the enterprise. Since the working methods of many industrial industries are different, the production characteristics of the enterprise can be judged according to the production type of the enterprise, and then its adjustability can be obtained.

[0048] The production types of enterprises are divided into three types: all-day production type, daytime production type, and nighttime production type.

[0049] (6) Unit response incentive: The unit response incentive is used to describe the incentive cost of the enterprise when participating in demand response under the current response situation at the time of unit response (1000 kW), and the incentive cost is calculated according to the incentive-response function.

[0050] (7) Response capacity: The response capacity represents the maximum adjustable capacity that industrial users can respond to when participating in ancillary services. The larger the response capacity, the greater the response potential.

[0051] The response capacity of industrial users The calculation formula is: ; (12) In the formula, is the power value at the is the power value of the previous load step, is the power value of the next load step.

[0052] (8) Response time: The response time represents the longest time that an enterprise device can participate in the response without affecting the production process since it starts to participate in the response. The longer the response time, the greater the user potential.

[0053] (9) Response speed: The response speed describes how quickly a power user responds to grid dispatching. The response speed refers to the time from receiving the regulation signal to starting to respond. The shorter the time, that is, the faster the response rate, the higher its value.

[0054] (10) Order demand: The order demand is a characteristic quantity that describes the basic production behavior of users and is used to represent the importance of the current production order to the entire enterprise. The more important the order is to the user, the less likely the user is to participate in the response.

[0055] The order demand of industrial users The calculation formula is: ; (13) In the formula, is the total profit in the order, is the total production profit of the enterprise in the past year.

[0056] (11) Proportion of electricity bill expenditure: The proportion of industrial users' electricity bill expenditure in production is an important factor affecting users' willingness to participate in the response. When the proportion of electricity bill is relatively high, users are more sensitive to electricity prices, and at this time, users are more likely to participate in the response through the stimulation of electricity prices.

[0057] Step 4: Use the triangular fuzzy number analytic hierarchy process and the improved entropy weight method to combine the weights of the indicators of the comprehensive evaluation system; Step 4.1: Calculate the index weights of the comprehensive evaluation system of industrial user load response potential using the triangular fuzzy number analytic hierarchy process; Step 4.1.1: Establish a triangular fuzzy number judgment matrix, compare the indicators of the comprehensive evaluation system of industrial user load response potential pairwise, and perform index scaling according to the difference in importance between the indicators to construct an n×n judgment matrix , where n is the number of indicators; ; (14) Step 4.1.2: Calculate the geometric mean of the indicators according to the triangular fuzzy numbers in each row of the judgment matrix ; ; (15) Wherein: , , are the lower bound, median and upper bound of the index ; Step 4.1.3: Calculate the geometric mean of each index p vector and its inverse vector, according to and its inverse vector to obtain the fuzzy weight of index p ; ; ; (16) Wherein: represents the circle addition of the matrix; represents the circle multiplication of the matrix; l, m, u are respectively the lower bound, median and upper bound of the fuzzy weight of index p; Step 4.1.4: Perform defuzzification calculation to obtain the weight of each index p, ; (17) Wherein, is the weight result of each index p after defuzzification; Step 4.1.5: Normalize the weight result of each index p after defuzzification to obtain the index weight of index p after normalization ; ; (18) Step 4.2: Use the improved entropy weight method to calculate the index weights of the industrial load response potential comprehensive evaluation system; Step 4.2.1: According to the sample data of the comprehensive evaluation system, establish an initial evaluation matrix , where m represents the number of rows of the initial evaluation matrix and n is the number of indexes; Step 4.2.2: Normalize the positive indexes and negative indexes in the initial evaluation matrix H respectively; Normalize the positive indexes in the initial evaluation matrix H according to the following formula: ; (19) Normalize the negative indexes in the initial evaluation matrix H according to the following formula: Wherein, is the result after normalization.

[0058] ; (20) Step 4.2.3: Calculate the information entropy of each index p: ; (21) ; (22) In the formula, represents the information entropy of the p-th index, represents 's characteristic coefficient. If , then let .

[0059] Step 4.2.4: According to the information entropy of the index, obtain the weight of each index p determined by the improved entropy weight method; ; (23) ; (24) ; (25) In the formula, is the weight of the p-th index after improvement; is the average value of the information entropy of all indexes whose information entropy is not equal to 1; are all intermediate variables.

[0060] Step 4.3: Integrate the index weights calculated in Steps 4.1 and 4.2 to obtain the comprehensive index weight, specifically including: 1) Denote the index weight vectors obtained in Steps 4.1 and 4.2 as and , and use different combination coefficients to and for combination to obtain the combined index weight vector set ;; ; (26) In the formula, represents the s-th combined basic weight vector; represents the index weight vector , 's combination coefficient; 2) Use the game theory method to construct an optimization model of the combined index weight vector ; (27) When dealing with multi-attribute decision-making problems, the game theory method seeks consistency or compromise to minimize the deviation between the optimal solution and a single solution.

[0061] Based on the game theory method, minimize the combined index weight vector and the index weights , Minimize the deviation between, and optimize the combined coefficients to find the optimal solution in the set of combined index weight vectors .

[0062] 3) Solve the optimization model; According to the properties of the differential function, the optimal first-order derivative condition of the weight vector can be transformed into the following linear equations.

[0063] ; (28) Solve the linear equations to obtain , and perform normalization to obtain , and substitute it into Equation (26) to obtain the comprehensive weight vector .

[0064] ; (29) Step 5: Combine the index weights obtained in Step 4, comprehensively evaluate the load response potential of industrial users using the TOPSIS method, and use the K-means algorithm to cluster according to the user load response potential to achieve the division of characteristic groups.

[0065] In the embodiment, the specific process of comprehensively evaluating the adjustable potential of industrial users' participation in demand response using the TOPSIS method includes: 1) Data preprocessing: Construct the initial evaluation matrix , and perform positive processing on the index values in the initial evaluation matrix to obtain ; 2) Determine the positive and negative ideal solutions: The positive ideal solution and the negative ideal solution are respectively the maximum and minimum values of each index; ; (30) ; (31) where and represent the positive and negative ideal solutions respectively, respectively represent the maximum value of the th index, respectively represent the minimum value of the th index; 3) Calculate the distance metric: Calculate the Euclidean distances between the index sample data of each user and the positive ideal solution and the negative ideal solution respectively, ; (32) ; (33) wherein , respectively represent the Euclidean distances between the index sample data of user i and the positive and negative ideal solutions; 4) Calculate the response potential: ; (34) wherein represents the response potential of user i; According to the response potential of user i, perform a descending order arrangement to achieve the sorting of user response potentials.

[0066] Step 6: According to the regulation requirements of the power system, utilize industrial users with different types of load response potentials to achieve grid load regulation, increase the flexibility of grid load adjustment, and improve the stability of the power system.

[0067] Example: Select 20 power-consuming enterprises from seven industries including electroplating and permanent magnet material manufacturing as experimental samples from among numerous industrial sectors in a certain city. Select the power consumption data of the most representative 15 days in summer as the experimental samples. The user numbers range from 1 to 20, and sample data is collected at a frequency of once every 15 minutes.

[0068] Under ideal conditions, the user incentive response is as Figure 3 shown.

[0069] In actual situations, under the same incentive conditions, there are significant differences in the sensitivity of different users to incentives, as Figure 4 shown. When the maximum incentive cost is 331,600 yuan, between Industry 4 and Industry 2, the maximum and minimum response capacities can differ by 4,059.17 kW, and their sensitivities to incentives vary greatly. The results show that when the response incentives are the same, the capacity that User 4 can respond to will be much greater than that of User 2, providing a reference for the system to allocate response capacities. And according to the response functions of the industries to which each user belongs and the maximum response capacity of the user itself, it can be obtained that the response capacity of User 9 is the largest, and there are significant differences in the maximum response capacities of different users. Therefore, when allocating capacities, it should be allocated according to the relationship between the response capacity of the user and the incentive cost.

[0070] Considering the differences in benefit functions in different industries, when the total peak shaving amount of the power system changes, the capacity allocation also changes, resulting in changes in the optimal selection results. The present invention selects the maximum response - incentive results given by the system when each user is at the maximum response for user optimization.

[0071] ​​​When using TFAHP to assign weights to indicators, first, experts score each indicator according to its importance to the potential evaluation system. Then, the triangular fuzzy numbers are processed using the fuzzy rough number theory to obtain fuzzy rough variables.

[0072] Table 1 Fuzzy evaluation information between indicators

[0073] The 11 indicators are pairwise compared to obtain the importance scale of the indicators. The triangular fuzzy numbers are used to process each indicator. Finally, the fuzzy evaluation results shown in Table 1 are obtained according to the differences in indicator importance, and the weights of each indicator are calculated using TFAHP. In addition, after obtaining the first weight result of the indicators, its consistency needs to be tested. The final test result shows that the consistency parameter CR = 0.08 < 0.1, so the consistency test result is qualified. Then, the load steps are used to step the electricity load data. According to the obtained step data matrix and the load step diagram, the data characteristic information is extracted to obtain the user data of two important indicators: response capacity and response speed. Then, the improved entropy weight method is used to calculate the objective weights of the indicators according to the established initial evaluation matrix H. Due to the limitations of the two methods themselves, it is necessary to balance the combination coefficients between the two and select the most appropriate combined weight coefficient to obtain the optimal solution. Finally, the game theory combined weight method is used to calculate the optimal weight coefficient to obtain the optimal combined weight, as shown in Table 2.

[0074] Table 2 Comprehensive weight results

[0075] Finally, the TOPSIS method is used to calculate the comprehensive scores of each user according to the obtained comprehensive weights and rank them. The K-means method is used to cluster the user potential according to the obtained adjustable potential results, and they are divided into high-potential, medium-potential, and low-potential industries. In the quantification of load potential, from the clustering results, it can be seen that: the value of high-potential users is greater than 0.5695, the value of medium-potential users is in the range of [0.4243, 0.3054], and the value of low-potential users is less than 0.2618. The final scoring and clustering results are as Figure 5 shown.

[0076] From Figure 5It can be seen that User 9 is a typical high-potential user, while User 16 is a typical low-potential user. After analyzing the electricity consumption data and enterprise characteristics of User 9 and User 16, it is found that User 9 has a relatively large responsive capacity, a fast response speed, a relatively low requirement for power supply continuity, and a relatively large proportion of electricity costs in the cost expenditure. This makes this user more likely to participate in the response compared to other users, and can flexibly adjust the load to meet the regulation requirements of the power grid. Moreover, due to the relatively low load factor of this user, the comprehensive potential of this user to participate in demand response is relatively large. When the power operator allocates peak shaving tasks, users with relatively large response potential should be given priority to ensure the smooth completion of peak shaving tasks as much as possible.

[0077] According to the final results, it is found that there are large fluctuations in the adjustable potential of different users in the same industry. This phenomenon is because there are significant differences in the electricity consumption characteristics, their own production methods, and enterprise operation states among different users, resulting in large fluctuations in the final comprehensive potential scores. Therefore, when measuring the overall response ability of an enterprise, the characteristics of the enterprise itself, such as its sensitivity to external incentives, should be considered from multiple aspects to achieve optimal user selection.

[0078] The present invention proposes a comprehensive evaluation model based on TFAHP and improved entropy weight method to help the power grid effectively measure the adjustable characteristics of each power user during demand response, prevent misallocation and wrong allocation during the distribution of regulation tasks, which may lead to the failure to complete peak shaving tasks in a timely manner and affect the system stability, realize the evaluation and analysis of the potential of industrial users to participate in demand response, and help power suppliers effectively screen high-quality users who can participate in system response.

[0079] (1) Use the response model to accurately express the functional relationship between user incentives and responses. At the same time, an improved genetic algorithm is added. Aiming at the differences in response functions among users, with the goal of minimizing the overall response incentive of the system, the response capacity of multi-industry users is allocated to obtain the optimal user load response incentive.

[0080] (2) Use load steps to process user electricity consumption data to obtain user characteristic data. Introduce triangular fuzzy numbers to transform the personal preferences of experts, and use a combination of subjective and objective weighting methods for weighting. At the same time, use game theory to balance the influence of various factors on the results during the evaluation process.

[0081] (3) Finally, use the K-means clustering algorithm to cluster and divide user potential, realize the cluster classification of user potential, and provide a reference for power suppliers to allocate response capacity. And use numerical examples to analyze the results. The results show that this method can effectively comprehensively evaluate the response potential of multi-industry users and provide effective support for the power grid to provide peak shaving services.

Claims

1. A load aggregation regulation method integrating an improved genetic algorithm and a game combination weighting method, characterized in that It includes the following steps: Step 1: Introduce a load step to preprocess the electricity consumption data of industrial users; Step 2: Build an industrial user load response model, and according to the functional relationship between excitation and response, use the genetic algorithm to solve for the optimal user load response excitation; Step 3: Build a comprehensive evaluation system for the load response potential of industrial users; Step 4: Use the triangular fuzzy number analytic hierarchy process and the improved entropy weight method to combine the weights of the indicators of the comprehensive evaluation system; Step 5: Combine the indicator weights obtained in Step 4, use the TOPSIS method to comprehensively evaluate the load response potential of industrial users, and use the K-means algorithm to cluster according to the user load response potential to achieve the division of characteristic groups; Step 6: According to the regulation requirements of the power system, use industrial users with different categories of load response potential to achieve grid load regulation, increase the flexibility of grid load adjustment, and improve the stability of the power system.

2. The load aggregation regulation method integrating an improved genetic algorithm and a game combination weighting method according to claim 1, characterized in that In Step 1, the electricity consumption data of industrial users are preprocessed through a load step to remove redundant load fluctuations, extract effective information from the data, and obtain data characteristics; The load step refers to the user load curve with a relatively small local load change rate; Local load change rate When it indicates that the load values of two adjacent points on the user load curve are at the same load step, that is, the user load is in a relatively stable state; when the local load change rate When it is, it indicates that the user's power consumption state is in an unstable state, representing different load levels caused by the switching of the overall equipment power consumption state during the actual production process of industrial users.

3. The load aggregation regulation method integrating the improved genetic algorithm and the game combination weighting method according to claim 2, wherein, In Step 1, the calculation process of the load step is as follows: 1) Define the initial load array and the judgment condition: Construct the initial load array based on the first 4 load data in the time series data of the industrial user's electricity load , and use as the judgment condition for the load step; when the values of the first 4 load data are at the same load step, the average value of the first 4 load data is used as the initial value of the load step ; ；（1） 2) Iteratively update the load step value according to the electricity consumption time series data of industrial users: ；（2） In the formula, , , , respectively represent the (t-3)th, (t-2)th, (t-1)th, and tth load data of the industrial user's electricity load time-series data, where t represents the time step; Update the load step value according to Equation (2). If is at the load step where it is located, then ; if is not at the load step where it is located, then construct a new load array using the electricity consumption load data of the t-th, (t + 1)-th, (t + 2)-th, and (t + 3)-th, re-judge and calculate to obtain a new load step value a v , ; 3) Record the information of the th load step on the th day, ； （3） In the formula, respectively represent the average value, start time, and end time of the historical data of the th load step on the th day, is the number of load steps within 15 days.

4. The load aggregation regulation method integrating an improved genetic algorithm and a game combination weighting method according to claim 1 or 2 or 3, characterized in that, In Step 2, the excitation-response function of the industrial user load response model is: ；（4） Wherein, is the user responding to the incentive during the time period , is the load quantity participated by the user in the response during the time period , is the random error caused by the user response volatility and are the incentive-response function parameters in the current situation.

5. The load aggregation regulation method integrating the improved genetic algorithm and the game combination weighting method according to claim 4, characterized in that, In Step 3, the indicators of the comprehensive evaluation system for the load response potential of industrial users include load stability, annual production hours of the maximum load, peak-valley difference rate, power supply reliability, enterprise production characteristics, unit response excitation, response capacity, response time, response speed, order demand, and proportion of electricity bill expenditure.

6. The load aggregation regulation method integrating the improved genetic algorithm and the game combination weighting method according to claim 5, wherein, In step 2, the response capacity is calculated by the formula: ；（5） Among them, is the power value at the time step , is the power value of the previous load step, is the power value of the next load step.

7. The load aggregation regulation method integrating an improved genetic algorithm and a game combination weighting method according to claim 6, characterized in that Step 4 specifically includes the following sub-steps: Step 4.1: Use the triangular fuzzy number analytic hierarchy process to calculate the indicator weights of the comprehensive evaluation system for the load response potential of industrial users; Step 4.1.1: Establish a triangular fuzzy number judgment matrix, compare the indicators of the comprehensive evaluation system for the load response potential of industrial users pairwise, and perform index scaling according to the difference in importance between the indicators to construct an n×n judgment matrix , where n is the number of indicators; Step 4.1.2: According to the judgment matrix Calculate the geometric mean of the indexes for each row of triangular fuzzy numbers ; ; Step 4.1.3: Calculate the geometric mean of each index p of the vector and its inverse vector, according to and its inverse vector to obtain the fuzzy weight of index p ; Step 4.1.4: Perform defuzzification calculation to obtain the weight of each index p ; Step 4.1.5: Normalize the weight result of each index p after defuzzification to obtain the index weight of index p after normalization ; Step 4.2: Use the improved entropy weight method to calculate the indicator weights of the comprehensive evaluation system for the industrial load response potential; Step 4.2.1: Establish an initial evaluation matrix based on the sample data of the comprehensive evaluation system , where m represents the number of rows of the initial evaluation matrix and n is the number of indicators; Step 4.2.2: Normalize the positive and negative indicators in the initial evaluation matrix H respectively; Step 4.2.3: Calculate the information entropy of each indicator p; Step 4.2.4: According to the information entropy of the indicators, obtain the weight of each indicator p determined by the improved entropy weight method; Step 4.3: Combine the indicator weights calculated in Step 4.1 and Step 4.2 to obtain the comprehensive indicator weights.

8. The load aggregation regulation method integrating the improved genetic algorithm and the game combination weighting method according to claim 7, characterized in that In Step 4.2.4, the weight of each index p determined by the improved entropy weight method is calculated by the following formula: ；（6） ；（7） ；（8） ；（9） ；（10） Wherein, represents the information entropy of the p-th index, represents the characteristic coefficient of, if , then let ; is the average value of the information entropy of all indices whose information entropy is not equal to 1; are all intermediate variables, is the element in the initial evaluation matrix H after normalization.

9. The load aggregation regulation method integrating the improved genetic algorithm and the game combination weighting method according to claim 7 or 8, characterized in that Step 4.3 specifically includes: 1) Denote the index weight vectors obtained in steps 4.1 and 4.2 as and , respectively. Combine and using different combination coefficients to obtain the combined index weight vector set ; ；（11） In the formula, represents the s-th combined basic weight vector; represents the index weight vector , is the combination coefficient of 2) Build an optimization model for the combined indicator weight vector; 3) Solve the optimization model; According to the properties of the differential function, the optimal first-order derivative condition of the optimization model of the combined indicator weight vector is the following linear equation system, ；（12） Solving the linear equations gives and performing normalization gives and substituting it into Equation (11) gives the comprehensive weight vector , ；（13）。 10. The load aggregation regulation method integrating the improved genetic algorithm and the game combination weighting method according to claim 9, characterized in that, In step 2), the combined index weight vector is minimized according to the game theory method and the index weight , Based on the principle of minimizing the deviation between them, the optimization model of the combined index weight vector is as follows: ；（14）。

Images