Data processing method and device for optimizing peak-valley time-of-use electricity price strategy

Abstract

The invention discloses a data processing method and device for optimizing a peak-valley time-of-use electricity price strategy. The method comprises the following steps: S110, acquiring power data of a peak-valley period of a target area; s120, determining a net load scene of the peak-valley period of the target area according to the power data; s130, selecting a corresponding optimization algorithm based on the net load scene of the peak-valley period; and S140, optimizing the user response model according to a selected corresponding optimization algorithm by taking the minimum peak-valley load difference as a target under a constraint condition to obtain a peak-valley time-of-use electricity price optimization strategy of the target area. The peak-valley time-of-use electricity price optimization strategy of the target area can be accurately obtained on the basis of comprehensively meeting the requirements of users and the scheduling requirements of the system.

Description

Technical Field

[0001] The present invention mainly relates to the field of electric power technology, and in particular to a data processing method and device for optimizing a peak-valley time-of-use electricity price strategy. Background Art

[0002] The current rapid economic development has led to a continuous increase in electricity demand. The current power supply system still has a significant imbalance between supply and demand during peak and trough periods. In other words, there is a shortage of power during peak periods and an oversupply of power during trough periods. To alleviate this problem, time-of-use electricity prices, as an important means of power demand side management, have been widely used and have played a key role in alleviating peak pressure on the power grid and optimizing power load distribution.

[0003] However, the inventors found that most of the current electricity price optimization algorithms use load transfer rate, user satisfaction and peak-to-valley difference rate as optimization targets to establish a multi-objective pricing model for electricity prices, thereby determining the corresponding electricity prices. The determination of the load transfer rate in this method relies on surveys and statistics, which is not precise enough and will affect the accuracy of the optimization results. Moreover, this method ignores the user's wishes and is difficult to fully meet the user's needs and the system's scheduling requirements. Summary of the invention

[0004] In view of the technical problems existing in the prior art, the present invention provides a data processing method and device for optimizing peak-valley time-of-use electricity price strategy, which can obtain accurate electricity price optimization strategy.

[0005] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0006] A data processing method for optimizing peak-valley time-of-use electricity price strategy, comprising the steps of:

[0007] S110, obtaining power data of the target area during peak and valley periods;

[0008] S120, determining the net load scenario of the target area during the peak and valley periods according to the power data;

[0009] S130, selecting a corresponding optimization algorithm based on the net load scenario during peak and valley periods;

[0010] S140. Under the constraints, with the goal of minimizing the peak-valley load difference, the user response model is optimized according to the selected corresponding optimization algorithm to obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

[0011] Preferably, the specific process of step S140 is:

[0012] S1401. Under the constraints, with the goal of minimizing the peak-valley load difference, the user response model is optimized according to the selected corresponding optimization algorithm to obtain the first peak-valley time-of-use electricity price optimization strategy;

[0013] S1402, using a multi-agent deep reinforcement algorithm to perform reinforcement learning training on the first peak-valley time-of-use electricity price optimization strategy to obtain a second peak-valley time-of-use electricity price optimization strategy;

[0014] S1403. Determine whether the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution; if so, the optimization ends; if not, return to execute steps S1401-S1402 until the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution, and obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

[0015] Preferably, in step S1402, the multi-agent deep reinforcement algorithm includes a queuing model constructed with the minimum peak-valley load difference and the unchanged electricity cost as priority classes.

[0016] Preferably, in step S140, the user response model is:

[0017]

[0018] Among them, L i is the fitted load at time i after implementing the peak-valley time-of-use electricity price optimization strategy in the target area; L i0 The actual load at time i before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; The average value of the total load in the peak period, peak period, normal period and valley period in their corresponding periods before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; 12 , 13 , 14 , 15 , 23 , 24 , 25 , 34 , 35 , 45 It is the load transfer rate of peak-peak period, peak-normal period, peak-valley period, peak-valley period, peak-normal period, peak-valley period, peak-valley period, normal-valley period, normal-valley period, and valley-valley period; T1, T2, T3, T4, and T5 are the five periods of peak, peak, normal, valley, and valley.

[0019] Preferably, the specific process of step S120 is:

[0020] The power data is compared with a plurality of pre-determined net load scenarios to determine the net load scenarios of the target area during the peak and valley periods; wherein the plurality of pre-determined net load scenarios are determined based on clustering results of the power data of the peak and valley periods of the plurality of target areas.

[0021] Preferably, in step S120, the peak and valley periods of the target area are determined by:

[0022] Based on the power data, calculate the net load corresponding to each time point of each day in each season of the target area;

[0023] For any season in the target area, the net average load curve data of the season is calculated according to the net load corresponding to each time point of each day in the season, the weight coefficient and the total number of days in the season; wherein the weight coefficient is determined based on the point density of each time point;

[0024] The net average load curve data of the season is clustered to determine the peak and valley periods of the season; the peak and valley periods include peak period, peak period, normal period, valley period and low period.

[0025] Preferably, the specific process of step S130 is:

[0026] In the preset adaptive algorithm system, the convergence time, solution speed, solution accuracy and stability are used as indicators to match the corresponding algorithm for the net load scenarios in the peak and valley periods;

[0027] Among them, the preset adaptive algorithm system includes an algorithm library and a scenario type library; the algorithm library includes taboo search algorithm, simulated annealing algorithm, genetic algorithm, particle swarm optimization algorithm and differential optimization algorithm; the scenario type library includes a variety of net load scenarios during peak and valley periods.

[0028] The present invention also discloses a data processing device for optimizing the peak-valley time-of-use electricity price strategy, comprising:

[0029] The acquisition module is used to obtain the power data of the target area during the peak and valley periods;

[0030] A partitioning module is used to determine the net load scenarios during peak and valley periods in the target area based on power data;

[0031] A selection module is used to select the corresponding optimization algorithm based on the net load scenario during peak and valley periods;

[0032] The optimization module is used to optimize the user response model under the constraints and with the goal of minimizing the peak-valley load difference, according to the selected corresponding optimization algorithm, to obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

[0033] The present invention further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are executed.

[0034] The present invention also discloses a computer device, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, and when the computer program is run by the processor, the steps of the method described above are executed.

[0035] Compared with the prior art, the advantages of the present invention are:

[0036] The embodiment of the present invention determines the corresponding net load scenario through the power data of the target area; in order to ensure that an accurate electricity price optimization strategy is obtained, a corresponding optimization algorithm is selected for the net load scenario corresponding to the target area; finally, considering the user's wishes and needs, the peak-valley load difference is minimized as the objective function, and the user response model that can reflect the user's wishes is optimized according to the selected corresponding optimization algorithm, and the peak-valley time-of-use electricity price optimization strategy for the target area is obtained on the basis of fully meeting the user's needs and the system's scheduling requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart for implementing the peak-valley time-of-use electricity price optimization method provided by an embodiment of the present invention.

[0038] Figure 2 It is a schematic diagram of clustering results provided by an embodiment of the present invention.

[0039] Figure 3 It is a diagram of the preset adaptive algorithm system architecture provided by an embodiment of the present invention.

[0040] Figure 4 It is a functional architecture diagram of autonomous reinforcement learning provided by an embodiment of the present invention.

[0041] Figure 5 It is an optimization solution flow chart of the peak-valley time-of-use electricity price optimization method provided by an embodiment of the present invention.

[0042] Figure 6 It is a flow chart for implementing the peak-valley time period division provided by an embodiment of the present invention.

[0043] Figure 7 is a schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0045] like Figure 1 As shown, the peak-valley time-of-use electricity price optimization method provided by the embodiment of the present invention includes the steps of:

[0046] S110: Obtaining power data of the target area during peak and valley periods.

[0047] In this embodiment, the peak and valley periods of the target area may be predetermined. The power data may include the centralized load (ie, total power), new energy output, purchased power, and delivered power, in addition to which it may also include historical basic power prices, etc., which are not limited here.

[0048] S120: Determine the net load scenario of the target area during the peak and valley periods according to the power data.

[0049] In this embodiment, considering the differences in electricity demand patterns and power generation resource endowments in different regions in terms of electricity price cost control, resulting in different costs for electricity production and transmission, in order to scientifically formulate electricity prices, this embodiment determines the net load scenarios corresponding to each peak and valley period based on the electricity data corresponding to the target area, thereby providing a scientific basis for formulating differentiated electricity price policies. This method can help to reasonably control electricity prices while ensuring electricity supply and promote balanced development of regional economies.

[0050] Accordingly, in an optional embodiment, step S120 includes:

[0051] The power data is compared with a plurality of pre-determined net load scenarios to determine the net load scenarios of the target area during the peak and valley periods; wherein the plurality of pre-determined net load scenarios are determined based on clustering results of the power data of the peak and valley periods of the plurality of target areas.

[0052] In this embodiment, multiple net load scenarios can be pre-determined, and each peak and valley period in each net load scenario corresponds to different power data. For example, the first net load scenario includes unified load and new energy output; the second net load scenario includes unified load, new energy output and purchased electricity, etc. The above two embodiments are only for illustration and do not constitute any limitation to this solution.

[0053] The data type corresponding to the power data of the target area is compared with each net load scenario. When the data type corresponding to the power data of the target area is completely consistent with a certain net load scenario, the net load scenario is used as the net load scenario of the target area during the peak and valley periods.

[0054] The following describes the determination of the net load scenario:

[0055] In this embodiment, power data of multiple different target areas can be collected in advance, and the power data of each target area can be clustered. By clustering regions with similar power generation and consumption characteristics, common factors in the formation of electricity price costs can be identified, and multiple net load scenarios can be obtained based on the clustering results.

[0056] like Figure 2 As shown, in order to obtain Figure 2 The clustering results in the above data were collected and the power supply and demand balance and power generation efficiency data of each province in the country in the past five years were divided. Figure 2 As shown, according to the clustering results, the embodiment of the present invention divides the net load scenarios of various regions across the country into the following four types:

[0057] Category 1 regions have a generally balanced electricity supply and demand, are self-sufficient in electricity, and have a high degree of supply and demand matching; the power generation structure is relatively diversified, with a relatively balanced proportion of thermal power, hydropower and new energy power generation.

[0058] Category 2 regions have strong electricity demand, especially high-energy-consuming industries and rapid urbanization; there is a large gap in electricity supply, and demand needs to be met through external power purchases.

[0059] Category three regions have abundant electricity resources, their power generation is greater than local electricity demand, they have abundant wind and solar resources, and their power generation efficiency is high, and they usually need to "transmit electricity from west to east."

[0060] Category 4 regions have abundant electricity resources, power generation is greater than local electricity demand, and the proportion of new energy installed capacity is high. When the output fluctuates greatly, it is occasionally necessary to purchase electricity from outside to balance supply and demand.

[0061] That is, based on Figure 2 According to the clustering results in , the embodiment of the present invention divides the net load scenarios into four categories. Among them, in the first net load scenario, there is new energy output, no out-of-region power purchase, and no out-of-region power transmission. This scenario is suitable for provinces that have a certain proportion of new energy power generation but are energy self-sufficient, and some small and medium-sized cities or regions in inland areas that rely on local wind power, solar energy and other renewable energy power generation. The power system in this type of area is mainly powered by new energy such as photovoltaic power generation and wind power generation, supplemented by a small amount of conventional power generation methods (such as thermal power and hydropower). However, due to the small electricity demand or infrequent external power exchange in these areas, there is no electricity trading with power grids outside the region, and the power system operates highly independently.

[0062] The second typical scenario of net load is that there is new energy output, out-of-region power purchase, and no out-of-region power transmission. This scenario is suitable for areas where new energy power generation accounts for a certain proportion but cannot fully meet local electricity demand, such as provinces with dense populations and developed industries. Although these areas have large new energy output, they have greater electricity demand, so they need to purchase electricity from other provinces or regions to balance electricity supply and demand. These areas have a large amount of solar power generation and wind power, but due to the rapid growth of economic and industrial demand, local power generation capacity is insufficient and they need to rely on out-of-region power purchase, especially during peak hours.

[0063] The third typical scenario of net load is that there is new energy output, no external power purchase, and external power transmission. This scenario is suitable for areas with abundant new energy power generation resources and the ability to transmit electricity to the outside. These areas are usually self-sufficient and do not rely on external power purchases. They mainly meet their own needs through local new energy power generation, and when local power is surplus, they transmit excess power to external areas. Typical such areas are mostly distributed in the western or northern regions with abundant new energy resources. These regions have a large number of wind power, photovoltaic and other renewable energy power generation facilities. Especially when there is sufficient sunlight or strong wind, new energy power generation greatly exceeds local demand, and exports electricity through the UHV transmission system to support the eastern or southern load centers with large power demand.

[0064] The fourth typical scenario of net load includes new energy output, out-of-region power purchase, and out-of-region power transmission. This scenario is suitable for regions that have a large amount of new energy power generation and need to export power to the outside, such as provinces in western and northern China that are rich in new energy resources. These regions have large-scale wind power and photovoltaic power generation facilities, and the output of new energy is large. In some periods, it exceeds local demand and transmits electricity to external provinces. At the same time, due to peak power consumption periods or large volatility of new energy, it is occasionally necessary to purchase electricity from outside to balance supply and demand. Category 4 The volatility of new energy power generation in these regions requires the power system to have a higher regulation capability, and two-way regulation of out-of-region power transmission and purchase helps maintain the balance of power supply and demand.

[0065] Step S130: Select a corresponding optimization algorithm based on the net load scenario during the peak and valley periods.

[0066] In this embodiment, considering that the data involved in different net load scenarios are different, the corresponding decision variables in the optimization process will also be different. Therefore, this embodiment selects different optimization algorithms for different net load scenarios to perform optimization solutions.

[0067] Accordingly, in an optional embodiment, step S130 includes:

[0068] In the preset adaptive algorithm system, convergence time, solution speed, solution accuracy and stability are used as indicators to match the corresponding algorithm for the net load scenarios in peak and valley periods.

[0069] Among them, the preset adaptive algorithm system includes an algorithm library and a scenario type library; the algorithm library includes taboo search algorithm, simulated annealing algorithm, genetic algorithm, particle swarm optimization algorithm and differential optimization algorithm; the scenario type library includes a variety of net load scenarios during peak and valley periods.

[0070] In this embodiment, considering that the method provided by this embodiment has a large number of decision variables in the objective function and the constraints when optimizing, each decision variable needs to consider different time periods, so the number of variables is large and the planning model is relatively large, so the optimization model selected in this embodiment is a nonlinear programming model. Nonlinear programming problems belong to mathematical nondeterministic polynomial (NP) difficult problems. Existing studies mainly use two algorithms to solve them, namely, linear solution algorithms and heuristic algorithms. This embodiment selects a heuristic algorithm by comparing the characteristics of linear solution algorithms and heuristic algorithms. Accordingly, the algorithm in the preset adaptive algorithm system is a heuristic algorithm.

[0071] like Figure 3 As shown, when selecting an optimization algorithm, the algorithm library and the scenario type library in the preset adaptive algorithm system are mapped. The algorithms in the algorithm library use convergence time, solution speed, solution accuracy and stability as indicators. According to the convergence time, solution speed, solution accuracy and stability of each optimization algorithm, the weight corresponding to each indicator is calculated. Finally, based on each indicator and its corresponding weight, the comprehensive score of each algorithm for the net load scenario of the target area is obtained, and the optimization algorithm with the highest score is used as the optimization algorithm corresponding to the net load area of ​​the target scenario.

[0072] In this embodiment, the convergence time refers to the time required for the algorithm to reach a preset convergence standard. The indicator is quantified by the time value, and the smaller the convergence time, the higher the indicator score.

[0073] The solution speed refers to the total time consumed by the algorithm from the beginning of execution to the result. The indicator is quantified by the time value. The smaller the solution speed, the higher the indicator score.

[0074] The objective function value (i.e., the solution accuracy is determined based on the objective function value) refers to the performance level achieved by the objective function value output by the algorithm. The indicator is quantified by the minimum economic cost, and the smaller the economic cost, the higher the indicator score.

[0075] Stability refers to the ability of the algorithm to maintain a relatively stable output without significant fluctuations or deviations in response to changes in input data, parameter fine-tuning, etc. The indicator is quantified by the variance of the objective function value. The smaller the variance, the higher the indicator score.

[0076] Accordingly, in the process of weighting indicators, that is, determining the weights corresponding to each indicator, considering that the results of the analytic hierarchy process (AHP) based on the knowledge and experience of experts are relatively subjective, and the results of the entropy weight method based on the information entropy theory are relatively objective, the embodiment of the present invention takes into account both subjective and objective factors in the evaluation process, adopts AHP and entropy weight method for subjective and objective weighting, and determines the indicator combination weight by weighted average. That is:

[0077] B g =ρw g +(1-ρ)v g

[0078] Where: B g is the combined weight of the g-th indicator; w g is the initial weight obtained by AHP; v g is the weight corrected by the entropy weight method; ρ is the AHP weight correction coefficient.

[0079] According to the weight of each indicator, the total evaluation score of each optimization algorithm under the net load scenario of the target area is calculated as follows:

[0080]

[0081] Among them, S w is the comprehensive evaluation score of the wth algorithm; y g is the score of the algorithm under the g-th indicator.

[0082] Step S140: Under the constraints, with the goal of minimizing the peak-valley load difference, the user response model is optimized according to the selected corresponding optimization algorithm to obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

[0083] In this implementation, considering the needs and wishes of users, a user response model should be set to reflect the fitted load of users in each time period after the implementation of the peak-valley time-of-use electricity price, so as to represent the user response in the peak-valley period, the user response in the peak-normal period, and the user response in the normal-valley period. This embodiment defines the load transfer rate as the ratio of the load transfer amount from the high electricity price period to the low electricity price period after the implementation of the time-of-use electricity price to the electricity price difference. Accordingly, the user response model can be expressed as:

[0084]

[0085] Among them, L i is the fitted load at time i after implementing the peak-valley time-of-use electricity price optimization strategy in the target area; L i0 The actual load at time i before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; The average value of the total load in the peak period, peak period, normal period and valley period in their corresponding periods before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; 12 , 13 , 14 , 15 , 23 , 24 , 25 , 34 , 35 , 45 It is the load transfer rate of peak-peak period, peak-normal period, peak-valley period, peak-valley period, peak-normal period, peak-valley period, peak-valley period, normal-valley period, normal-valley period, and valley-valley period; T1, T2, T3, T4, and T5 are the five periods of peak, peak, normal, valley, and valley.

[0086] Considering that the main purpose of implementing peak-valley time-of-use electricity prices is to reduce the peak-valley difference of the power grid and balance the load in each period, so as to ensure the safe and stable operation of the power grid. Therefore, taking the minimum peak-valley load difference as the objective function, the corresponding objective function can be expressed as:

[0087] min(maxL i -minL i )

[0088] Considering that the implementation of peak-valley time-of-use electricity prices aims to guide users' electricity consumption behavior in different time periods by adjusting electricity prices, thereby optimizing the allocation of power resources. Its core purpose is to balance the supply and demand of electricity by adjusting the electricity consumption time without reducing the user's electricity consumption, reduce the peak load pressure of the power grid, and improve the power utilization rate during the valley period. Therefore, taking the user's electricity consumption as the constraint condition, it can be expressed as:

[0089]

[0090]

[0091] W≈W ′

[0092] Where W is the user’s daily electricity consumption before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; W ′ It is the user's daily electricity consumption after implementing the peak-valley time-of-use electricity price optimization strategy in the target area.

[0093] The purpose of implementing peak-valley time-of-use electricity prices is to reduce users' electricity expenses. Therefore, after the implementation of this policy, the overall expenditure of users should be less than before the implementation. After the implementation of peak-valley time-of-use electricity prices, the peak and valley periods cannot be reversed. Therefore, the user's electricity cost is used as a constraint condition, and its expression is:

[0094] M=Pi0 W

[0095]

[0096] M ′ <M

[0097] Among them, P i0 To implement the unified electricity price before time-of-use electricity price, W i P is the electricity consumption in period i after the implementation of time-of-use electricity price. i It is the electricity price in period i after the implementation of time-of-use electricity price.

[0098] The peak-valley time-of-use electricity price needs to be priced based on the production cost of electricity, the local economic situation and the user's acceptance level, so as to formulate a reasonable peak-valley time-of-use electricity price. Therefore, the electricity price is used as a constraint condition, and its expression is:

[0099]

[0100] Among them, P 1,min , P 2,min , P 3,min , P 4,min , P 5,min , P 1,max , P 2,max , P 3,max , P 4,max , P 5,max They are the lower and upper limits of electricity price fluctuations during peak, peak, flat, valley and off-peak periods respectively.

[0101] In order to prevent the peak-valley electricity price difference from being too large or too small, resulting in peak-valley inversion or weak response, it is necessary to limit the peak-valley electricity price ratio. Correspondingly, the peak-valley electricity price ratio is:

[0102]

[0103] Among them, K1 and K2 are electricity price ratio limit constants. At the same time, according to national standards, the peak-to-valley electricity price ratio is generally between 2 and 5, so K1=2, K2=5.

[0104] Considering that autonomous reinforcement learning is suitable for processing sequential decision-making problems, it can learn the optimal strategy through interaction with the environment in an uncertain environment. In the problem of time-of-use electricity price difference optimization, reinforcement learning can learn how to dynamically dispatch power resources based on environmental information such as electricity price fluctuations, energy demand, and load forecasts to achieve optimal utilization of electricity price differences. Therefore, this embodiment proposes an autonomous reinforcement learning functional architecture based on the selected optimization algorithm to optimize the user response model and obtain the peak and valley time-of-use electricity price optimization strategy for the target area.

[0105] Accordingly, in an optional embodiment, step S140 includes:

[0106] Under the constraints, taking the minimum peak-valley load difference as the goal, the user response model is optimized according to the selected corresponding optimization algorithm to obtain the first peak-valley time-of-use electricity price optimization strategy.

[0107] A multi-agent deep reinforcement algorithm is used to perform reinforcement learning training on the first peak-valley time-of-use electricity price optimization strategy to obtain the second peak-valley time-of-use electricity price optimization strategy.

[0108] Determine whether the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution. If not, return to the constraints, with the goal of minimizing the peak-valley load difference, optimize the user response model according to the corresponding optimization algorithm selected, and obtain the steps of the first peak-valley time-of-use electricity price optimization strategy, until the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution, and then obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

[0109] In this embodiment, the autonomous reinforcement learning functional architecture includes the selection of an optimization algorithm, and the optimization solution based on the selected optimization algorithm and the multi-agent deep reinforcement algorithm. The selection of the optimization algorithm has been described in the above-mentioned related embodiments, so it will not be repeated here.

[0110] See also Figure 4 , the multi-agent deep reinforcement algorithm is described below. In traditional reinforcement learning, the optimal strategy is considered to be a parameterized update strategy π0. For a policy gradient with a local minimum J(π), the parameter Therefore, the updated electricity price strategy Given by:

[0111]

[0112] Among them, the policy gradient formula is:

[0113]

[0114] in, is the update gradient; π(S|a) is the distribution of electricity prices in the queuing model, γ t is a discount factor, which represents the weight of future rewards. The policy gradient criticism algorithm using the proposed queuing model provides the maximum expected reward and satisfies the condition of minimum peak-to-valley difference. The policy π is updated by searching the constraint c and The optimized value is given by:

[0115]

[0116] Therefore, this constrained Markov decision process can be solved by the proposed reinforcement learning memory-enhanced network (Multi-Attention Convolutional Neural Network, MADNN). In the proposed MADNN, The Gaussian distribution of is approximated by a multilayer perceptron network.

[0117]

[0118] Among them, logγ is the queue model The standard deviation of μ , are bias and weight respectively. The features are presented to the proposed MADNN through the rectified linear unit.

[0119]

[0120]

[0121] Among them, n = 1, 2, 3, ... (n-1) is used for training and proposes a Gaussian distribution The MADNN is used to explore the pricing strategy of peak and valley time-of-use electricity prices. For policy updates, the reward functions Π(S|a) and μ(S|a) are used to extract deep features through linear combinations. The updated value function is given by:

[0122]

[0123] Among them, μ n (S|a) refers to the updated strategy, which is optimized by the ReLu neuron:

[0124]

[0125] Among them, γ is the convergence gradient coefficient of updating the strategy rules to formulate the optimal peak-valley time-of-use electricity price. The updated constraint strategy is:

[0126] μ k+1 = argmaxE μ [A(S|a)],k∈n

[0127]

[0128]

[0129] In order to guide the agents in the multi-agent deep reinforcement learning algorithm to choose the appropriate time to perform tasks by dynamically adjusting prices, this embodiment performs queuing standard pricing according to factors such as each agent's electricity consumption, load size, and current electricity price, and allocates different electricity prices to each agent. In the peak-valley time-of-use electricity price, queuing standard pricing means that during peak electricity consumption, the system guides users to postpone electricity consumption to the valley time when the electricity price is lower through a pricing mechanism.

[0130] For simple queue models without constrained Markov decision processes, the shortest queue model is considered as the policy. But for constrained Markov decision processes with dynamic pricing, another approach is needed. Let N queue numbers and M power consumption periods be allocated, two DNNs are used for policy update and value update. For policy update, the participant network adopts Parameterize and estimate J(π μ ). The actor network in MADNN undertakes the task of policy improvement and updates J(π μ ) is an estimated value.

[0131] In order to provide the best scheduling, two classes of objects are created using separate nodes: the minimum peak-valley load difference and the constant electricity cost. These two priority classes are used to build the queuing model. By adopting the proposed MADNN and the optimal allocation strategy, the electricity price optimization formula is:

[0132]

[0133] In the formula, To control the weight of the queue, α is the recommended learning rate of MADNN.

[0134] This example focuses on describing the optimization solution process. Figure 5 In this embodiment, after searching for a feasible and better initial solution in the solution space through the selected corresponding optimization algorithm, the initial solution generated by the optimization algorithm is used as the starting point of reinforcement learning, and interaction with the environment begins to gradually improve the electricity price strategy. Then, a multi-agent deep reinforcement learning algorithm is used, which together with the optimization algorithm can provide a flexible and efficient solution for the optimization of the time-of-use electricity price difference.

[0135] Specifically, according to the corresponding optimization algorithm, under the constraints and objective function, the first peak-valley time-of-use electricity price optimization strategy is obtained. The multi-agent deep reinforcement algorithm is adopted, the first peak-valley time-of-use electricity price optimization strategy is used to interact with the environment, and the first peak-valley time-of-use electricity price optimization strategy is updated based on the reward signal to obtain the second peak-valley time-of-use electricity price optimization strategy.

[0136] Determine whether the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution; if not, cyclically execute the step of outputting the first peak-valley time-of-use electricity price optimization strategy using the selected optimization algorithm until the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution.

[0137] If so, determine whether the second peak-valley time-of-use electricity price optimization strategy converges; if it converges, use the second peak-valley time-of-use electricity price optimization strategy as the peak-valley time-of-use electricity price optimization strategy for the target area; if it does not converge, re-adopt the multi-agent deep reinforcement algorithm to perform reinforcement learning training on the second peak-valley time-of-use electricity price optimization strategy to obtain the third peak-valley time-of-use electricity price optimization strategy, until the third peak-valley time-of-use electricity price optimization strategy converges, and then use the third peak-valley time-of-use electricity price optimization strategy as the peak-valley time-of-use electricity price optimization strategy for the target area.

[0138] like Figure 6 As shown, in step S110, the division of the peak and valley time periods of the target area is determined by the following method:

[0139] Based on the power data, the net load corresponding to each time point of each day in each season of the target area is calculated.

[0140] For any season in the target area, the net average load curve data of the season is calculated according to the net load corresponding to each time point of each day in the season, the weight coefficient and the total number of days in the season; wherein the weight coefficient is determined based on the point density of each time point;

[0141] The improved fuzzy C-means clustering algorithm IFCM is used to cluster the net average load curve data of this season to determine the peak and valley periods of this season; wherein, IFCM is obtained by improving the original membership in the fuzzy C-means clustering algorithm FCM into an exponential membership; the peak and valley periods include peak period, peak period, normal period, valley period and low period.

[0142] Specifically, Figure 6 As shown, the division of the peak and valley periods of the target area can be divided into a data preprocessing process and a division process; the data preprocessing process is:

[0143] In this embodiment, considering that the current time-of-use electricity prices are mostly divided into peak and valley periods based on the net load curve, the net load curve is mainly determined by statistical distribution method or measurement method. This method is insufficient in precision and cannot flexibly respond to changes in load characteristics.

[0144] In addition, existing studies generally use a net load-based clustering algorithm in terms of time period division, taking into account load transfer characteristics, but these methods usually ignore the role of energy storage facilities in power load regulation. At the same time, most of the existing time period divisions are based on historical data and fail to achieve real-time adjustment during load changes, resulting in a delayed response to dynamic load characteristics and difficulty in adapting to instantaneous fluctuations in the power system. Based on this, an embodiment of the present invention proposes a new peak-valley time period division method, which is described in detail as follows:

[0145] Since the traditional net load calculation method is to add the negative value of the new energy output to the centralized load. However, due to the balance of electricity demand or supply in some areas, it is necessary to purchase electricity or transmit electricity outside the region. In this case, the traditional net load calculation method shows a certain difference from the actual situation, resulting in inaccurate net load calculation. To solve this technical problem, this embodiment optimizes the calculation of net load and proposes a general framework formula for net load calculation, namely: Net load = centralized load - new energy output - purchase of electricity outside the region + transmission of electricity outside the region. This definition not only takes into account the impact of new energy output on the load within the region, but also combines the factors of cross-regional power allocation, and more comprehensively reflects the power supply and demand conditions in different regions.

[0146] Taking into account the power supply and demand conditions in various regions, in order to reduce the amount of calculation, the net load corresponding to each time point of each day in any season in the target area can select the corresponding net load calculation formula based on the net load scenario corresponding to the target area, specifically:

[0147] The net load calculation formula corresponding to the first net load scenario is:

[0148] Q Si =L Si -R Si

[0149] Among them, Q Si is the net load corresponding to each time point of each day in any season of the target area; L Si is the coordinated load corresponding to each time point of the day in the seasonal cycle; R Si The new energy output corresponding to each time point of each day in this seasonal cycle.

[0150] The net load calculation formula corresponding to the second net load scenario is:

[0151]

[0152] in, It is the amount of purchased electricity corresponding to each time point of each day in this seasonal cycle.

[0153] The net load calculation formula corresponding to the third net load scenario is:

[0154]

[0155] in, It is the amount of electricity transmitted at each time point of each day in this seasonal cycle.

[0156] The net load calculation formula corresponding to the fourth net load scenario is:

[0157]

[0158] In this embodiment, it is considered that there are holidays and rest days in each season, and the electricity consumption on these rest days and holidays is different from that on weekdays, that is, the net load corresponding to the time points on these rest days and holidays is the singular point in the load data. If the net load data at each time point of each day in the integrated season is directly averaged, the obtained seasonal average net load curve will lead to deviations in the classification results, so the concept of point density is introduced.

[0159] Specifically, a smaller weight is given to the larger net load data that appear at the same time point in the same season, and a larger weight is given to the smaller data, so as to improve the accuracy of the seasonal net average load curve. For each sample point (that is, the net load corresponding to each time point), its point density is defined as:

[0160]

[0161] Among them, f i represents the estimated density of net load sample points; n represents the total number of net load sample points; d ij Represents the distance between sample points i and j.

[0162] For each sample point f i Normalize them separately to get the weight coefficient of each sample point:

[0163]

[0164] The weight coefficient can reflect the influence of the i-th sample on the subsequent division of peak and valley time periods. Small weights are given to singular points to reduce their influence on the final result of the subsequent division of peak and valley time periods and improve accuracy.

[0165] Since the method provided in this embodiment is based on seasonal net average load curve data, it is necessary to divide the net load data according to seasons to determine the number of time points corresponding to each season and the specific net load data corresponding to each time point. Accordingly, the net average load curve data of each season can be calculated by the following formula:

[0166]

[0167] Where S is the total number of days for constructing the seasonal average load curve; Q si is the net load corresponding to each time point of each day in the entire seasonal cycle; Q i is the load corresponding to each time point on the seasonal average net load curve; w i is the weight of each time point. The time point load set corresponding to the seasonal average net load curve is D = {Q1, Q2, ..., Q 24}, the weight set of each time point on the seasonal average net load curve is W = {W1, W2, ..., W 24}.

[0168] After determining the seasonal net average load curve data, the division process is described below:

[0169] When dividing the peak and valley time periods, it is necessary to consider the peak and valley time period division principles, namely:

[0170] (1) Since the electricity price is lower during off-peak hours, the power supplier operates at a loss during this period. In order to ensure the basic income of the power supplier, the time division between peak hours and normal hours should not be too short. According to the calculation results, the peak hours should not be less than 6 hours per day, and the normal hours should not be less than 3 hours per day;

[0171] (2) The division of peak and valley periods should clearly reflect the characteristics of the actual seasonal load curve;

[0172] (3) To facilitate implementation and calculation, the peak and valley periods should not be divided too short, and the shortest period should not be less than 1 hour;

[0173] (4) The total electricity consumption before and after the peak-valley electricity price policy remains basically unchanged (the maximum error cannot exceed ±5%)

[0174] Considering that under normal circumstances, the peak and valley periods of daily electricity consumption in the same season are relatively fixed, and the periods close to the peak period are mostly peak periods or peak periods; the periods close to the valley period are mostly valley periods or low valley periods. However, the boundary between the peak period and the valley period is relatively fuzzy, and it is necessary to select an appropriate division method for division. This embodiment uses the Improved Fuzzy C-means (IFCM) clustering algorithm for division.

[0175] The IFCM algorithm is essentially the same as the FCM algorithm. Both are soft partitioning methods. The membership function is used to calculate the degree to which the electricity load of each hour belongs to the peak, peak, flat, valley, and off-peak periods. The size of the membership is used to determine which period the hour should belong to: peak, peak, flat, valley, or off-peak.

[0176] Both the FCM algorithm and the IFCM algorithm use an objective function to classify the load data into peak and valley periods. For the FCM algorithm, its objective function can be expressed as:

[0177]

[0178] Among them, U represents the membership matrix; V represents the matrix composed of cluster centers of each time period; c represents the number of time periods to be divided, that is, the number of clusters; n represents the number of samples; m represents the degree of fuzzification; u ij Indicates the degree to which the jth data in the load data belongs to the ith time period; d ij Represents the distance between the jth data and the i-th cluster center in the load data.

[0179] The purpose of clustering is to classify each group of loads into the corresponding peak, flat and valley periods, which is achieved mathematically by minimizing the objective function. Combining the objective function and its constraints, we can know that the corresponding Lagrangian function is:

[0180]

[0181] Find u for F ij and v i Partial derivative, get the membership function and cluster center function of the FCM algorithm:

[0182]

[0183] Where x j represents the jth sample in the load data; v i Represents the cluster center of the i-th category.

[0184] The traditional FCM algorithm has a weak global search capability due to the existence of local extreme values ​​in its membership function, and sometimes it is impossible to obtain the optimal solution during the clustering process. To address this problem, LI Enwen et al. proposed a new improved algorithm, the IFCM algorithm. The IFCM algorithm replaces the original membership with the exponential form of the membership based on the FCM algorithm, which enhances the global search capability of the algorithm, effectively avoids the occurrence of local extreme values, and improves the accuracy of clustering. The objective function is defined as:

[0185]

[0186] Where r is the adjustment similarity factor; P ij Represents the sample x in the load data i With cluster center v i The similarity of . Generally, the size of J value is used as the criterion for judging the clustering effect. When the J value reaches the minimum, the clustering effect is also optimal.

[0187] The following is an explanation of how to determine the peak and valley periods of the season:

[0188] Step 1: Since the seasonal average net load curve needs to be divided into five periods: peak, peak, flat, valley, and low valley, c = 5. According to the input load data sample set D, the randomly generated membership matrix U and the determined fuzzification degree m, the formula is used Determine ij By formula Generate cluster centers V1(1), V2(1), V3(1), V4(1), and V5(1). The number in brackets is the current iteration number.

[0189] Step 2: Calculate the distance from each sample point in the load data sample set D to any cluster center V j (k)(j=1,2,3,4,5); at the same time, according to the minimum Euclidean distance principle, for a certain sample point x i In the kth iteration, the distance from a cluster center V j (k) is closest, x i (k) belongs to D j (k), j = (1, 2, 3, 4, 5), then D j (k) is the jth classification situation formed after the kth iteration.

[0190] Step 3: Perform k+1 iterations according to the formula in step 1. Obtain the k+1th membership degree u of each sample ij (k+1) and the k+1th cluster center V j (k+1). According to the division criteria in step 2, output the classification situation D after the k+1th selection j (k+1). The final classification result is obtained based on the membership of each sample. The final division of peak, flat and valley time periods is completed.

[0191] Step 4: When the objective function J(U,V) meets the iteration termination condition, clustering is terminated and the membership matrix is ​​output. If the iteration termination condition is not met, return to step 2 to continue the iterative calculation. The iteration termination condition is:

[0192] |J(K+1)-J(K)|≤θ

[0193]

[0194] The membership matrix is:

[0195]

[0196] After the peak-valley time period division is performed, in order to ensure the rationality of the peak-valley time period division, this embodiment introduces two clustering evaluation indicators to quantitatively analyze the time period division results based on the clustering evaluation indicators. Among them, the proposed clustering evaluation indicators are the partition coefficient (PC) and the classification entropy (CE). Correspondingly, the expressions of these two clustering evaluation indicators are:

[0197]

[0198]

[0199] Among them, V pc Indicates the intra-class compactness of the data set. The larger the value, the better the clustering effect. ce Indicates the separability between classes. The smaller its value, the better the clustering effect. pc and V ce The value range of V is between 0 and 1. pc The closer the value of is to 1, the better the clustering effect. ce The closer the value is to 0, the better the clustering effect is.

[0200] It should be understood that the peak-valley time division method provided in this embodiment is only used as an implementation method. In actual application, the peak-valley time-sharing method provided in this embodiment is not adopted. Instead, the general method at this stage can also solve the technical problem to be solved by this application, that is, on the basis of fully meeting the needs of users and the scheduling requirements of the system, the peak-valley time-sharing electricity price optimization strategy for the target area is obtained.

[0201] In summary, the embodiments of the present invention take into account the differences in electricity demand patterns and power generation resource endowments in different regions, resulting in different electricity production and transportation costs in different regions. If a unified electricity price optimization strategy is used, it is difficult to meet the scheduling needs of the system. Based on this, the embodiments of the present invention determine the corresponding net load scenario through the power data of the target area; in order to ensure that an accurate electricity price optimization strategy is obtained, a corresponding optimization algorithm is selected for the net load scenario corresponding to the target area; finally, considering the user's wishes and needs, the peak-valley load difference is minimized as the objective function, and the user response model that can reflect the user's wishes is optimized according to the selected corresponding optimization algorithm. On the basis of fully meeting the user's needs and the system's scheduling requirements, the peak-valley time-of-use electricity price optimization strategy for the target area is obtained.

[0202] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0203] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.

[0204] Figure 7 The schematic diagram of the structure of the peak-valley time-of-use electricity price optimization device provided by the embodiment of the present invention is shown. For the convenience of explanation, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0205] The data processing device for optimizing the peak-valley time-of-use electricity price strategy provided by the embodiment of the present invention includes:

[0206] The acquisition module is used to obtain the power data of the target area during the peak and valley periods;

[0207] A partitioning module is used to determine the net load scenarios during peak and valley periods in the target area based on power data;

[0208] A selection module is used to select the corresponding optimization algorithm based on the net load scenario during peak and valley periods;

[0209] The optimization module is used to optimize the user response model under the constraints and with the goal of minimizing the peak-valley load difference, according to the selected corresponding optimization algorithm, to obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

[0210] In an optional embodiment, the optimization module is specifically used to:

[0211] Under the constraints, the user response model is optimized with the goal of minimizing the peak-valley load difference according to the selected corresponding optimization algorithm to obtain the first peak-valley time-of-use electricity price optimization strategy;

[0212] A multi-agent deep reinforcement algorithm is used to conduct reinforcement learning training on the first peak-valley time-of-use electricity price optimization strategy to obtain the second peak-valley time-of-use electricity price optimization strategy;

[0213] Determine whether the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution. If not, return to the constraints, with the goal of minimizing the peak-valley load difference, optimize the user response model according to the corresponding optimization algorithm selected, and obtain the steps of the first peak-valley time-of-use electricity price optimization strategy, until the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution, and then obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

[0214] In an optional embodiment, the multi-agent deep reinforcement algorithm includes a queuing model; the queuing model is constructed with the minimum peak-valley load difference and unchanged electricity cost as priorities.

[0215] In an optional embodiment, the user response model is:

[0216]

[0217] Among them, L i is the fitted load at time i after implementing the peak-valley time-of-use electricity price optimization strategy in the target area; L i0 The actual load at time i before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; The average value of the total load in the peak period, peak period, normal period and valley period in their corresponding periods before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; t2 , 13 , 14 , 15 , 23 , 24 , 25 , 34 , 35 , 45 It is the load transfer rate of peak-peak period, peak-normal period, peak-valley period, peak-valley period, peak-normal period, peak-valley period, peak-valley period, normal-valley period, normal-valley period, and valley-valley period; T1, T2, T3, T4, and T5 are the five periods of peak, peak, normal, valley, and valley.

[0218] In an optional embodiment, the selection module is specifically used for:

[0219] In the preset adaptive algorithm system, the convergence time, solution speed, solution accuracy and stability are used as indicators to match the corresponding algorithm for the net load scenarios in the peak and valley periods;

[0220] Among them, the preset adaptive algorithm system includes an algorithm library and a scenario type library; the algorithm library includes taboo search algorithm, simulated annealing algorithm, genetic algorithm, particle swarm optimization algorithm and differential optimization algorithm; the scenario type library includes a variety of net load scenarios during peak and valley periods.

[0221] In an optional embodiment, the partitioning module is specifically used for:

[0222] The power data is compared with a plurality of pre-determined net load scenarios to determine the net load scenarios of the target area during the peak and valley periods; wherein the plurality of pre-determined net load scenarios are determined based on clustering results of the power data of the peak and valley periods of the plurality of target areas.

[0223] In an optional embodiment, the peak-valley time-of-use electricity price optimization device further includes a clustering module, which is used to:

[0224] Based on the power data, calculate the net load corresponding to each time point of each day in each season of the target area;

[0225] For any season in the target area, the net average load curve data of the season is calculated according to the net load corresponding to each time point of each day in the season, the weight coefficient and the total number of days in the season, wherein the weight coefficient is determined based on the point density of each time point;

[0226] The improved fuzzy C-means clustering algorithm IFCM is used to cluster the net average load curve data of this season to determine the peak and valley periods of this season; wherein, IFCM is obtained by improving the original membership in the fuzzy C-means clustering algorithm FCM into an exponential membership; the peak and valley periods include peak period, peak period, normal period, valley period and low period.

[0227] The present invention further discloses a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are executed. The present invention also discloses a computer device, including a memory and a processor connected to each other, on which a computer program is stored, and when the computer program is executed by the processor, the steps of the above method are executed. The medium and device of the present invention correspond to the above method and also have the advantages described in the above method.

[0228] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiment when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. The memory is used to store computer programs and / or modules, and the processor implements various functions by running or executing computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0229] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A data processing method for optimizing peak-valley time-of-use electricity price strategy, characterized in that: Includes steps: S110, obtaining power data of the target area during peak and valley periods; S120, determining the net load scenario of the target area during the peak and valley periods according to the power data; S130, selecting a corresponding optimization algorithm based on the net load scenario during peak and valley periods; S140. Under the constraints, with the goal of minimizing the peak-valley load difference, the user response model is optimized according to the selected corresponding optimization algorithm to obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

2. The data processing method for optimizing the peak-valley time-of-use electricity price strategy according to claim 1 is characterized in that: The specific process of step S140 is: S1401. Under the constraints, with the goal of minimizing the peak-valley load difference, the user response model is optimized according to the selected corresponding optimization algorithm to obtain the first peak-valley time-of-use electricity price optimization strategy; S1402, using a multi-agent deep reinforcement algorithm to perform reinforcement learning training on the first peak-valley time-of-use electricity price optimization strategy to obtain a second peak-valley time-of-use electricity price optimization strategy; S1403. Determine whether the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution; if so, the optimization ends; if not, return to execute steps S1401-S1402 until the second peak-valley time-of-use electricity price optimization strategy is the global optimal solution, and obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

3. The data processing method for optimizing the peak-valley time-of-use electricity price strategy according to claim 2 is characterized in that: In step S1402, the multi-agent deep reinforcement algorithm includes a queuing model constructed with the minimum peak-valley load difference and the unchanged electricity cost as the priority class.

4. The data processing method for optimizing the peak-valley time-of-use electricity price strategy according to claim 1, 2 or 3, characterized in that: In step S140, the user response model is: Among them, L i is the fitted load at time i after implementing the peak-valley time-of-use electricity price optimization strategy in the target area; L i0 The actual load at time i before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; The average value of the total load in the peak period, peak period, normal period and valley period in their corresponding periods before implementing the peak-valley time-of-use electricity price optimization strategy in the target area; 12 , 13 , 14 , 15 , 23 , 24 , 25 , 34 , 35 , 45 It is the load transfer rate of peak-peak period, peak-normal period, peak-valley period, peak-valley period, peak-normal period, peak-valley period, peak-valley period, normal-valley period, normal-valley period, and valley-valley period; T1, T2, T3, T4, and T5 are the five periods of peak, peak, normal, valley, and valley.

5. The data processing method for optimizing the peak-valley time-of-use electricity price strategy according to claim 1, 2 or 3, characterized in that: The specific process of step S120 is: The power data is compared with a plurality of pre-determined net load scenarios to determine the net load scenarios of the target area during the peak and valley periods; wherein the plurality of pre-determined net load scenarios are determined based on clustering results of the power data of the peak and valley periods of the plurality of target areas.

6. The data processing method for optimizing the peak-valley time-of-use electricity price strategy according to claim 1, 2 or 3, characterized in that: In step S120, the peak and valley periods of the target area are determined by: Based on the power data, calculate the net load corresponding to each time point of each day in each season of the target area; For any season in the target area, the net average load curve data of the season is calculated according to the net load corresponding to each time point of each day in the season, the weight coefficient and the total number of days in the season; wherein the weight coefficient is determined based on the point density of each time point; The net average load curve data of the season is clustered to determine the peak and valley periods of the season; the peak and valley periods include peak period, peak period, normal period, valley period and low period.

7. The data processing method for optimizing the peak-valley time-of-use electricity price strategy according to claim 1, 2 or 3, characterized in that: The specific process of step S130 is: In the preset adaptive algorithm system, the convergence time, solution speed, solution accuracy and stability are used as indicators to match the corresponding algorithm for the net load scenarios in the peak and valley periods; Among them, the preset adaptive algorithm system includes an algorithm library and a scenario type library; the algorithm library includes taboo search algorithm, simulated annealing algorithm, genetic algorithm, particle swarm optimization algorithm and differential optimization algorithm; the scenario type library includes a variety of net load scenarios during peak and valley periods.

8. A data processing device for optimizing peak-valley time-of-use electricity price strategy, characterized in that: include: The acquisition module is used to obtain the power data of the target area during the peak and valley periods; A partitioning module is used to determine the net load scenarios during peak and valley periods in the target area based on power data; A selection module is used to select the corresponding optimization algorithm based on the net load scenario during peak and valley periods; The optimization module is used to optimize the user response model under the constraints and with the goal of minimizing the peak-valley load difference, according to the selected corresponding optimization algorithm, to obtain the peak-valley time-of-use electricity price optimization strategy for the target area.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

Images