Flexible load regulation and control method and system based on global optimization and edge collaboration

Abstract

The invention discloses a flexible load regulation and control method and system based on global optimization and edge collaboration, and belongs to the field of power systems, and the method comprises the steps: predicting a power grid flexible load prediction value corresponding to a target regulation and control day according to the flexible load data of a current power grid; iteratively optimizing the initial load regulation and control solution according to the power grid flexible load predicted value and a preset random swarm intelligent optimization algorithm to obtain an optimal load regulation and control solution; according to the adjustable power range corresponding to the flexible load of each edge device, decomposing the optimal load regulation and control solution to obtain a first load regulation and control solution corresponding to each edge device; according to the difference value between the actual value and the predicted value of the flexible load of each edge device in the target regulation and control day and a preset distribution rule, adjusting each first load regulation and control solution to obtain the compensation load capacity of each edge device; and according to each compensation load capacity, each edge device is controlled to execute flexible load regulation and control. According to the invention, the problem of poor flexible load regulation and control effect in the prior art can be solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of power systems, and relates to a flexible load regulation method and system based on global optimization and edge collaboration. BACKGROUND

[0002] High proportion of wind power, photovoltaic and other new energy grid-connected has become an important feature of new power systems. However, new energy power generation has significant randomness, intermittency and low disturbance resistance, which makes it difficult to balance the supply and demand of the power grid. At the same time, the load characteristics are more complex due to the diversified demand of industrial, commercial and residential electricity and the access of new types of loads such as electric vehicles, and the load fluctuation and time-varying nature are significantly enhanced. The traditional power dispatching relies on centralized management, which is difficult to adapt to regional differences and real-time dynamic regulation requirements, especially in the collaborative control of special variable loads (such as industrial parks, factories and mines, etc.), there is a lack of effective hierarchical collaborative strategy for flexible loads (transferable, convertible and reducible loads), which leads to insufficient suppression of load peak-valley difference and limited optimization of power grid loss, and cannot meet the requirements of new power systems for flexibility and stability.

[0003] However, the traditional centralized regulation method ignores regional differences and edge device collaboration capabilities, resulting in slow load distribution response speed and insufficient global optimization. When each flexible load is independently regulated, there is a lack of collaborative strategy, which cannot effectively balance the load demand and power generation fluctuation between regions, resulting in poor regulation effect, large load peak-valley difference, high power grid loss, and difficulty in meeting the requirements of new power systems for flexibility and stability. SUMMARY

[0004] The present application provides a flexible load regulation method and system based on global optimization and edge collaboration, which can solve the problem of poor flexible load regulation effect in the prior art, reduce the suppression of power grid load peak-valley difference and load fluctuation, and improve the power grid operation efficiency and stability.

[0005] To achieve the above-mentioned purpose, in a first aspect, the present application provides a flexible load regulation method based on global optimization and edge collaboration, comprising:

[0006] According to the flexible load data of the current power grid, the power grid flexible load prediction value corresponding to the target regulation day is predicted;

[0007] According to the power grid flexible load prediction value and the preset random swarm intelligence optimization algorithm, the initial load regulation solution is iteratively optimized to obtain the optimal load regulation solution;

[0008] According to the adjustable power range of the flexible load corresponding to each edge device, the optimal load regulation solution is decomposed to obtain the first load regulation solution corresponding to each edge device;

[0009] According to the difference between the actual value and the predicted value of the flexible load of each edge device on the target regulation day, and a preset allocation rule, the first load regulation solution of each edge device is adjusted to obtain a compensation load amount of each edge device;

[0010] According to the compensation load amount of each edge device, the corresponding edge device is controlled to perform regulation, and the regulation of the flexible load is completed.

[0011] Compared with the prior art, the embodiments of the application have the following beneficial effects: by predicting the power grid flexible load data of the target regulation day, data basis is provided for optimization; by using a random swarm intelligence optimization algorithm to iteratively generate an optimal solution, the rationality of global load distribution is ensured; by decomposing the optimal solution according to the adjustable power range of each edge device, local overload is avoided; by adjusting the regulation solution according to the dynamic difference between the actual value and the predicted value, load deviation is compensated in real time; finally, by executing the compensation amount through the edge device, precise regulation of the flexible load is realized, and overall load fluctuation and peak-valley difference are reduced.

[0012] In some embodiments of the first aspect of the application, the iterative optimization of the initial load regulation solution according to the power grid flexible load predicted value and the preset random swarm intelligence optimization algorithm to obtain the optimal load regulation solution comprises:

[0013] An initial temperature value is initialized, and the power grid flexible load predicted value is represented as an initial load regulation solution;

[0014] According to a preset simulated annealing optimization algorithm, the temperature value is iteratively updated until the current temperature value reaches a preset critical point, and the current load regulation solution is output as the optimal load regulation solution;

[0015] Wherein, after updating the temperature value each time, a first load regulation solution is iteratively generated according to the preset simulated annealing optimization algorithm and the constraint condition, a loss function difference value between the first load regulation solution and the current load regulation solution is calculated according to a preset loss function, and the current load regulation solution is iteratively updated according to the loss function difference value until a preset iteration number is reached; or when the loss difference between the load regulation solutions obtained by adjacent two iterations meets a preset threshold, the current load regulation solution is output as the optimal load regulation solution and the iteration is stopped.

[0016] Compared with the prior art, the above-mentioned embodiment has the following beneficial effects: initializing the temperature value and characterizing the power grid flexible load prediction value as an initial load regulation solution, providing a starting point for the simulated annealing optimization algorithm, iteratively updating the temperature value until a critical point is reached, outputting the current load regulation solution as an optimal load regulation solution, which can effectively avoid local optimization problems in a complex and variable environment and ensure that the best solution is found; after updating the temperature value each time, the first load regulation solution is iteratively generated based on the preset simulated annealing optimization algorithm and constraint condition, and the loss function difference is calculated for iterative updating, which ensures that each step moves in the optimal direction while meeting all necessary constraints, thereby improving the robustness and effectiveness of the entire process.

[0017] In some embodiments of the first aspect of the application, the iterative generation of the first load regulation solution according to the preset simulated annealing optimization algorithm and constraint condition comprises:

[0018] The first load regulation solution is iteratively generated according to the simulated annealing optimization algorithm and constraint condition; wherein the constraint condition comprises a maximum power constraint, a power balance constraint, a minimum cost and minimum loss constraint.

[0019] Compared with the prior art, the above-mentioned embodiment has the following beneficial effects: the maximum power constraint prevents the flexible load regulation amount from exceeding the equipment bearing limit, avoids the risk of equipment overload, and ensures the safety of power grid operation; the power balance constraint ensures that the total load remains stable within the regulation period, avoiding power grid fluctuations caused by supply-demand imbalance; the minimum cost and loss constraint limits the excessive reduction of cost or loss in the optimization process, preventing power grid shock or operation instability caused by regulation deviation, thereby improving the economy while maintaining the safety margin of the power grid.

[0020] In some embodiments of the first aspect of the application, the iterative optimization of the initial load regulation solution according to the power grid flexible load prediction value and the preset random swarm intelligence optimization algorithm to obtain the optimal load regulation solution further comprises:

[0021] The initial load regulation solution is iteratively optimized to obtain the optimal load regulation solution, with the minimization of the preset objective function as the optimization target of the random swarm intelligence optimization algorithm; wherein the optimization target is:

[0022] Wherein, min(f) represents the optimization target, ∑ 24h represents statistical optimization in units of days, P i is the i-th data of the power grid flexible load prediction value of the target day, P target is the average value of the power grid flexible load prediction value of the target day; Loss i is the power loss of the power grid at time i of the target day, Loss target,iacceptable target grid loss at a target time of a target day, and a and b are power balance factor and grid loss factor respectively.

[0023] Compared with the prior art, the above embodiment has the following beneficial effects: taking a dual target function of power balance and grid loss as the core, reducing the overall load fluctuation within the day through 24-hour statistical optimization; dynamically adjusting the weights of the power balance factor and the loss factor to flexibly adapt to different period demands (such as focusing on peak shaving and valley filling on the peak side, and focusing on loss optimization on the flat section); setting the target value based on the predicted average value to make the regulation scheme closer to the actual operation scenario, and improve the practicality and adaptability of the optimization result.

[0024] In some embodiments of the first aspect of the application, the adjustment of the first load regulation solution of each edge device to obtain the compensation load of each edge device according to the difference between the actual value and the predicted value of the flexible load of each edge device in the target regulation day, and the preset allocation rule, comprises:

[0025] According to the difference between the actual value and the corresponding flexible load predicted value of each edge device, the total amount of the load to be compensated of each edge device and the corresponding adjacent device is calculated respectively;

[0026] Taking minimizing the preset loss function as the target, the dynamic allocation threshold is iteratively adjusted in the preset value interval according to the segmented search algorithm, and based on the dynamic allocation threshold, the proportion relationship between the target load average value of each edge device and the first load regulation solution, the first load regulation solution of the edge device and the adjacent device is adjusted, until the loss function reaches the minimum value, to obtain the compensation load of each edge device; wherein the sum of the compensation load of each edge device and the adjacent device is equal to the corresponding total amount of the load to be compensated.

[0027] Compared with the prior art, the above embodiment has the following beneficial effects: the total amount to be compensated is determined by calculating the difference between the actual value and the predicted value, and the deviation source is accurately located; the segmented search algorithm is used to dynamically adjust the allocation threshold, and the optimal compensation ratio is quickly locked; the compensation amount is allocated based on the proportion relationship between the target load average value and the regulation solution, to ensure that the load distribution is uniform after compensation, avoid local overload or underload, and further suppress the fluctuation.

[0028] In some embodiments of the first aspect of the application, taking minimizing the preset loss function as the target, iteratively adjusting the dynamic allocation threshold in the preset value interval according to the segmented search algorithm, comprises:

[0029] According to the preset loss function and the segmented search algorithm, the dynamic allocation threshold is iteratively adjusted in the preset value interval; wherein the loss function is:

[0030]

[0031] ; wherein f r (ΔPc j ) represents a loss function, ΔPc j represents the compensation load of the edge device j, and m represents the number of the edge device j and the corresponding adjacent device, represents the first load regulation solution of the edge device j, Lave j represents the target load average of the edge device j, and ΔPcom represents the total amount of the load to be compensated for by the edge device j and the corresponding adjacent device.

[0032] Compared with the prior art, the above embodiment has the following beneficial effects: the difference between the load before and after compensation and the target average value is nonlinearly amplified by an exponential loss function, the deviation in the high fluctuation region is preferentially inhibited, the algorithm is forced to preferentially eliminate the fluctuation in the large deviation region, and the defect that the local small deviation is excessively optimized and the key high fluctuation region is ignored by the traditional square function is avoided; the allocation threshold is dynamically adjusted by combining the segmented search algorithm, the efficiency and precision balance of compensation allocation is realized under the constraint that the total amount of compensation and the actual difference are consistent, and finally the power grid operation stability is improved by inhibiting the global load fluctuation and the local peak-valley difference.

[0033] In some embodiments of the first aspect of the application, the first load regulation solution of the edge device and the adjacent device is adjusted based on the proportion relationship between the target load average and the first load regulation solution of each edge device and the dynamic allocation threshold, until the loss function reaches the minimum value, to obtain the compensation load of each edge device, including:

[0034] According to the dynamic allocation threshold, the proportion relationship between the target load average and the first load regulation solution of each edge device and the corresponding adjacent device, and the positive and negative of the corresponding total amount of load to be compensated for, the device to be compensated for is screened;

[0035] The first load regulation solution of the device to be compensated for is adjusted according to the dynamic allocation threshold, to obtain the compensation load of each edge device; wherein the adjustment algorithm is as follows:

[0036] Wherein λ represents the dynamic allocation threshold.

[0037] Compared with the prior art, the above embodiment has the following beneficial effects: high-deviation devices are screened according to the dynamic allocation threshold, resources are concentrated to compensate for key regions, and the low regulation efficiency caused by the dispersion of resources is avoided; the regulation solution is corrected according to the target load average, to ensure that the load of each device after compensation is closer to the target average value, to further smooth the load curve, reduce the local peak-valley difference, and improve the power grid operation stability.

[0038] In some embodiments of the first aspect of the application, the screening of the to-be-compensated devices according to the dynamic allocation threshold, the proportional relationship between the target load average of each edge device and the corresponding adjacent device and the first load regulation solution, and the positive or negative of the total amount of to-be-compensated load of the corresponding adjacent device comprises:

[0039] When the total amount of to-be-compensated load of each edge device and the corresponding adjacent device is greater than zero, the proportion of the first load regulation solution of each edge device and the adjacent device to the target load average is sorted from small to large, and the device with a proportion value lower than the dynamic allocation threshold is screened as the to-be-compensated device;

[0040] When the total amount of to-be-compensated load of each edge device and the corresponding adjacent device is less than zero, the proportion of the first load regulation solution of each edge device and the adjacent device to the target load average is sorted from large to small, and the device with a proportion value higher than the dynamic allocation threshold is screened as the to-be-compensated device.

[0041] Compared with the prior art, the above-mentioned embodiments have the following beneficial effects: the devices are sorted according to the positive or negative of the total amount of compensation, the area with the largest deviation is preferentially processed (for example, when positive compensation, the low-proportion device is preferentially allocated, and when negative compensation, the high-proportion device is preferentially reduced), and the resource utilization efficiency is improved; the key devices are screened through the dynamic threshold, the main deviation is concentratedly solved, the redundant regulation is avoided, and the adaptability and response speed of the regulation scheme are enhanced.

[0042] In a second aspect, the application further provides a flexible load regulation system based on global optimization and edge coordination, comprising a data acquisition module, a regulation solution acquisition module, a decomposition module, a compensation load calculation module and a control execution module.

[0043] The data acquisition module is configured to predict the predicted value of the flexible load of the power grid corresponding to the target regulation day according to the flexible load data of the current power grid.

[0044] The regulation solution acquisition module is configured to iteratively optimize the initial load regulation solution according to the predicted value of the flexible load of the power grid and a preset random swarm intelligent optimization algorithm, to obtain an optimal load regulation solution.

[0045] The decomposition module is configured to decompose the optimal load regulation solution according to the adjustable power range corresponding to the flexible load of each edge device, to obtain the first load regulation solution corresponding to each edge device.

[0046] The compensation load calculation module is configured to adjust each first load regulation solution according to the difference between the actual value and the predicted value of the flexible load of each edge device in the target regulation day, and a preset allocation rule, to obtain the compensation load amount of each edge device.

[0047] The control execution module is configured to control each edge device to perform regulation and control according to the compensation load amount of each edge device, so as to complete regulation and control of the flexible load.

[0048] Compared with the prior art, the above embodiments have the following beneficial effects: the power grid flexible load data of the target regulation and control day is predicted to provide data basis for optimization; the random swarm intelligence optimization algorithm is used to iteratively generate an optimal solution to ensure the rationality of global load distribution; the optimal solution is decomposed according to the adjustable power range of each edge device to avoid local overload; the regulation and control solution is adjusted according to the dynamic difference between the actual value and the predicted value to compensate for load deviation in real time; finally, the edge device executes the compensation amount to realize accurate regulation and control of the flexible load and reduce overall load fluctuation and peak-valley difference.

[0049] In some embodiments of the second aspect of the application, the regulation and control solution acquisition module comprises an initialization unit and a regulation and control solution iteration unit.

[0050] The initialization unit is configured to initialize a temperature value and represent the power grid flexible load predicted value as an initial load regulation and control solution.

[0051] The regulation and control solution iteration unit is configured to iteratively update the temperature value according to a preset simulated annealing optimization algorithm until the current temperature value reaches a preset critical point, and output the current load regulation and control solution as an optimal load regulation and control solution.

[0052] After each update of the temperature value, a first load regulation and control solution is iteratively generated according to the preset simulated annealing optimization algorithm and the constraint condition, a loss function difference value of the first load regulation and control solution and the current load regulation and control solution is calculated according to a preset loss function, and the current load regulation and control solution is iteratively updated according to the loss function difference value until a preset iteration number is reached; or when the loss difference between the load regulation and control solutions obtained by adjacent two iterations meets a preset threshold, the current load regulation and control solution is output as the optimal load regulation and control solution and the iteration is stopped.

[0053] Compared with the prior art, the above embodiments have the following beneficial effects: the temperature value is initialized and the power grid flexible load predicted value is represented as an initial load regulation and control solution to provide a starting point for the simulated annealing optimization algorithm; the temperature value is iteratively updated using the simulated annealing optimization algorithm until the critical point is reached, and the current load regulation and control solution is output as the optimal load regulation and control solution, which can effectively avoid local optimal problems in a complex and changeable environment and ensure that the best solution is found; after each update of the temperature value, the first load regulation and control solution is iteratively generated based on the preset simulated annealing optimization algorithm and the constraint condition, and the loss function difference value is calculated for iterative updating, which ensures that each step moves in the optimal direction and meets all necessary constraints, thereby improving the robustness and effectiveness of the whole process. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 Figure 1 is a flowchart of a flexible load regulation method based on global optimization and edge coordination provided in some embodiments of the present application.

[0055] Figure 2 Figure 2 is a structural diagram of a flexible load regulation system based on global optimization and edge coordination provided in some embodiments of the present application.

[0056] Figure 3 Figure 3 is a diagram of total load curve and flexible load operation curve of a special transformer when not regulated provided in some embodiments of the present application.

[0057] Figure 4 Figure 4 is a diagram of 30-day load peak-valley difference comparison under different regulation modes provided in some embodiments of the present application.

[0058] Figure 5 Figure 5 is a 30-day load fluctuation comparison table provided in some embodiments of the present application.

[0059] Figure 6 Figure 6 is a diagram of load fluctuation reduction amount of global optimization and edge coordination scheme compared with global optimization scheme provided in some embodiments of the present application. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0061] Embodiment one:

[0062] Please refer to Figure 1 To solve the problem of poor flexible load regulation effect in the prior art, an embodiment of the present application provides a flexible load regulation method based on global optimization and edge coordination, which comprises the following steps S1 to S4:

[0063] Step S1: According to the flexible load data of the current power grid, the power grid flexible load prediction value corresponding to the target regulation day is predicted.

[0064] In specific implementation, a prediction model can be pre-trained using historical data, and then the prediction model is used to predict the current data. After collecting the current data, the flexible load of the power grid as a whole can be summarized, which is represented as: L f =∑L fi ; Wherein L fL is the total amount of flexible load fi L i is the value of the i-th flexible load.

[0065] In the embodiment, step S1 provides data basis for optimization by predicting the grid flexible load data of the target regulation day.

[0066] Step S2: iteratively optimizes the initial load regulation solution according to the grid flexible load prediction value and the preset random swarm intelligence optimization algorithm to obtain the optimal load regulation solution.

[0067] Further, the step S2 can be implemented by the following preferred embodiment, including steps S21-S22, specifically as follows:

[0068] S21: initialize the temperature value, and represent the grid flexible load prediction value as the initial load regulation solution;

[0069] S22: iteratively update the temperature value according to the preset simulated annealing optimization algorithm until the current temperature value reaches the preset critical point, and output the current load regulation solution as the optimal load regulation solution;

[0070] Wherein, after updating the temperature value each time, according to the preset simulated annealing optimization algorithm and the constraint condition, iteratively generate the first load regulation solution, and according to the preset loss function, calculate the loss function difference value of the first load regulation solution and the current load regulation solution, and according to the loss function difference value, iteratively update the current load regulation solution until the preset iteration number is reached; or when the loss difference between the load regulation solutions obtained by adjacent two iterations meets the preset threshold, output the current load regulation solution as the optimal load regulation solution and stop iteration.

[0071] The simulated annealing optimization algorithm is an algorithm that simulates the annealing process of metal, which randomly walks in the search area, that is, randomly takes values of flexible load in the regulation range, and then uses Metropolis sampling criterion to gradually approach the local optimal solution. Metropolis sampling is an effective importance sampling method, and its algorithm is:

[0072] When the system changes from one energy state to another, the corresponding energy changes from E1 to E2, and the probability is: Where T is the annealing temperature. If E2 < E1, the system directly accepts this state as a better solution; otherwise, it accepts or discards this state with a random probability, where the probability of state 2 being accepted is: After a certain number of iterations, the system will gradually tend to a stable distribution state.

[0073] In this application, the above-mentioned simulated annealing optimization algorithm is introduced as a random swarm intelligence optimization algorithm for implementation, specifically as follows:

[0074] One: initialization, set the initial temperature T, set the initial load control solution X as the data of the flexible load at 288 time points of the target day (if the target day is the current day, use the actual value, if the target day is a future date, use the predicted data) and the iteration number L of each T value = 500;

[0075] Two: repeat iteration L times, and reduce the temperature T for reiteration after each iteration of L times. In each iteration, a random new solution X * (that is, the first load control solution described above) is first generated by using a random function, and then the loss function difference Δf = f(X * )-f(X) is calculated; if Δf < 0, then X * is updated as a better solution, otherwise, X is accepted as a new optimal solution with a probability * . In specific implementation, the termination condition can be that the loss difference between two consecutive better solutions is lower than a tolerance (0.01) or the temperature is reduced to a critical point. If the termination condition is met, the iteration program is ended, and the current solution is output as the optimal solution (that is, the optimal load control solution described above).

[0076] In the preferred embodiment, step S21 initializes the temperature value and characterizes the predicted value of the grid flexible load as the initial load control solution, which provides a starting point for the simulated annealing optimization algorithm. Step S22 iteratively updates the temperature value until the critical point is reached, and outputs the current load control solution as the optimal load control solution. This can effectively avoid local optimal problems in complex and variable environments and ensure that the best solution is found. After updating the temperature value each time, the first load control solution is iteratively generated based on the preset simulated annealing optimization algorithm and constraint conditions, and the loss function difference is calculated for iterative update. This mechanism ensures that each step moves in the optimal direction while satisfying all necessary constraints, thereby improving the robustness and effectiveness of the entire process.

[0077] Further, the iteration to generate the first load control solution in step S22 can be implemented by the following preferred embodiment, specifically:

[0078] According to the simulated annealing optimization algorithm and the constraint conditions, the first load control solution is iteratively generated. The constraint conditions include maximum power constraint, power balance constraint, minimum cost and minimum loss constraint.

[0079] In specific implementation, the above four constraints can be represented as:

[0080] ∑P jmin <△P i <∑P jmax ;∑ 24h △P i= ε; C i ≥ C min ; Loss i ≥ Loss min ;

[0081] wherein, △P i is the regulation amount of the i-th flexible load, P jmin is the lower limit value of the j-th flexible load regulation, P jmax is the upper limit value of the j-th flexible load regulation; ε is a number close to 0; C i represents the regulation cost corresponding to the flexible load i, C min represents the minimum regulation cost; Loss represents the power loss of the power grid.

[0082] In the preferred embodiment, the maximum power constraint is used to prevent the flexible load regulation amount from exceeding the equipment bearing limit, avoiding the risk of equipment overload and ensuring the safe operation of the power grid; the power balance constraint ensures that the total load remains stable within the regulation period, avoiding power fluctuations caused by supply and demand imbalance; the minimum cost and loss constraint limits the excessive reduction of cost or loss in the optimization process, preventing power grid shocks or operation instability caused by regulation deviation, thereby improving the economy while maintaining the safety margin of the power grid.

[0083] Further, the step S2 can also be implemented by one of the following preferred embodiments, specifically:

[0084] The preset objective function is minimized as the optimization target of the random swarm intelligence optimization algorithm, and the initial load regulation solution is iteratively optimized to obtain the optimal load regulation solution; wherein the optimization target is:

[0085] wherein, min(f) represents the optimization target, ∑ 24h represents the statistical optimization in units of day, P i is the i-th data of the predicted value of the grid flexible load on the target day, P target is the average value of the predicted value of the grid flexible load on the target day; Loss i is the power loss of the power grid at i time on the target day, Loss target,i is the acceptable target power loss at i time on the target day, and a and b are the power balance factor and the power loss factor, respectively.

[0086] In specific implementation, as with the simulated annealing optimization algorithm described above, if the target day is the current day, the actual value of the flexible load is used for P i , P target and Loss i , and if the target day is a future day, the predicted value is used.

[0087] In the preferred embodiment, the dual objective function of power balance and grid loss is used as the core to reduce the overall load fluctuation within the day through 24-hour statistics optimization; the weights of the power balance factor and the loss factor are dynamically adjusted to flexibly adapt to different period demands (such as focusing on peak shaving and valley filling on the peak side, and focusing on loss optimization on the flat section); the target value is set based on the predicted average value to make the control scheme closer to the actual operation scenario and improve the practicality and adaptability of the optimization results.

[0088] Step S3: According to the adjustable power range corresponding to the flexible load of each edge device, the optimal load control solution is decomposed to obtain the first load control solution corresponding to each edge device.

[0089] In specific implementation, the decomposition of the optimal load control solution can be represented as:

[0090] ΔPmax j =P jmax -P jmin ; wherein, is the optimal load control curve (i.e. the first load control solution) of the jth edge device, and ΔPmax j is the flexible load adjustment range of the jth edge device.

[0091] In the embodiment, by decomposing the optimal solution according to the adjustable power range of each edge device, the problem of local overload can be effectively avoided.

[0092] Step S4: According to the difference between the actual value and the predicted value of the flexible load of each edge device in the target control day, and the preset allocation rule, the first load control solution of each edge device is adjusted to obtain the compensation load of each edge device.

[0093] Further, step S4 can be implemented through the following preferred embodiment, including steps S41-S42, as follows:

[0094] S41: According to the difference between the actual value and the corresponding predicted value of the flexible load of each edge device, the total amount of the load to be compensated of each edge device and the corresponding adjacent device is calculated respectively.

[0095] In specific implementation, the difference and the total amount of the load to be compensated are calculated as follows:

[0096]

[0097] wherein, ΔPcom j (i) is the load to be compensated of the jth flexible load at the ith moment, Preal j (n) is the actual load value corresponding to the nth moment, and Ppred j(n) is the predicted value of the jth flexible load at the nth moment; ΔPcom(i) is the total amount of load to be compensated for between the edge device and the adjacent device at the ith moment, and m is the total number of the edge device and the adjacent device.

[0098] S42: iteratively adjusting the dynamically allocated threshold value in the preset value interval according to the segmentation search algorithm, and adjusting the first load control solution of the edge device and the adjacent device based on the dynamically allocated threshold value, the average value of the target load of each edge device, and the proportional relationship of the first load control solution, until the loss function reaches the minimum value, to obtain the compensation load of each edge device, with the minimum preset loss function as the target; wherein the sum of the compensation load of each edge device and the adjacent device is equal to the corresponding total amount of load to be compensated.

[0099] In the preferred embodiment, steps S41-S42 determine the total amount of load to be compensated by calculating the difference between the actual and predicted values, accurately locate the source of deviation; use the segmentation search algorithm to dynamically adjust the allocation threshold, quickly lock the optimal compensation ratio; based on the proportional relationship between the average value of the target load and the control solution, the compensation amount is allocated to ensure that the load distribution is uniform after compensation, avoid local overload or underload, and further suppress fluctuations.

[0100] Further, the dynamically allocated threshold value in step S42 can be iterated by the following preferred implementation, specifically:

[0101] According to the preset loss function and the segmentation search algorithm, the dynamically allocated threshold value is iteratively adjusted in the preset value interval; wherein the loss function is:

[0102]

[0103] ; wherein f r (ΔPc j ) represents the loss function, ΔPc j represents the compensation load of the edge device j, m represents the number of edge device j and the corresponding adjacent device, represents the first load control solution of the edge device j, Lave j represents the average value of the target load of the edge device j, and ΔPcom represents the total amount of load to be compensated for between the edge device j and the corresponding adjacent device.

[0104] In the preferred embodiment, the difference between the load before and after compensation deviating from the target average value is nonlinearly amplified by an exponential loss function, the deviation in the high fluctuation region is preferentially suppressed, the algorithm is forced to preferentially eliminate the fluctuation in the large deviation region, and the defect of the traditional square function of over-optimizing local small deviation and ignoring the key high fluctuation region is avoided; the threshold is dynamically adjusted by combining the segmented search algorithm, the efficiency and accuracy of compensation distribution are balanced under the constraint of strictly meeting the consistency of the total compensation amount and the actual difference, and finally the stability of power grid operation is improved by suppressing the global load fluctuation and local peak-valley difference.

[0105] Further, in step S42, the first load control solution of the edge device and the adjacent device is adjusted based on the dynamic allocation threshold, the proportional relationship between the target load average value of each edge device and the first load control solution, until the loss function reaches the minimum value, and the compensation load of each edge device is obtained, which can be realized by the following preferred implementation, including steps S421-S422, as follows:

[0106] S421: According to the dynamic allocation threshold, the proportional relationship between the target load average value of each edge device and the first load control solution, and the positive and negative of the corresponding total amount of load to be compensated, the device to be compensated is selected;

[0107] S422: According to the dynamic allocation threshold, the first load control solution of the device to be compensated is adjusted, and the compensation load of each edge device is obtained; wherein the adjustment algorithm is as follows:

[0108] Where λ represents the dynamic allocation threshold.

[0109] In specific implementation, the optimal value of the dynamic allocation threshold λ can be iteratively searched and adjusted by the segmented search method such as binary search, for example, the value space of λ, i.e. [0, 1], is divided into n (n>2) segments, n+1 critical points are obtained, f r (ΔPc j ) at each critical point is calculated, the critical point corresponding to the minimum value of f r (ΔPc j ) is found, and the adjacent two critical points are taken as the value space for the next search, then they are divided into n (n>2) segments for search, and so on, until the value space is relatively small or the change of f r (ΔPc j ) is relatively small, and λ at this time is taken as the optimal value.

[0110] In the preferred embodiment, steps S421-S422 screen high-deviation devices according to the dynamic allocation threshold, concentrate resources to compensate for key areas, avoid low regulation efficiency caused by scattered resources, and correct the regulation solution in combination with the target load average to ensure that the load of each device after compensation is closer to the target average, further smooth the load curve, reduce local peak-valley difference, and improve power grid operation stability.

[0111] Further, step S421 can be implemented through the following preferred embodiments, specifically:

[0112] When the total amount of load to be compensated for each edge device and the corresponding adjacent device is greater than zero, the proportion of the first load regulation solution of each edge device and adjacent device to the target load average is sorted from small to large, and the devices with a proportion value lower than the dynamic allocation threshold are screened as the devices to be compensated for;

[0113] When the total amount of load to be compensated for each edge device and the corresponding adjacent device is less than zero, the proportion of the first load regulation solution of each edge device and adjacent device to the target load average is sorted from large to small, and the devices with a proportion value higher than the dynamic allocation threshold are screened as the devices to be compensated for.

[0114] In specific implementation, the proportion of the first load regulation solution to the target load average is expressed as: If the total amount of load to be compensated for is greater than zero, the load compensation is preferentially allocated to small, and if the total amount of load to be compensated for is less than zero, the load compensation is preferentially allocated to large.

[0115] In the preferred embodiment, step S421 sorts the devices according to the positive and negative nature of the compensation total amount, preferentially processes the area with the largest deviation (such as preferentially allocating to low-proportion devices when positive compensation and preferentially reducing high-proportion devices when negative compensation), improves resource utilization efficiency, screens key devices through a dynamic threshold, concentrates on solving major deviations, avoids redundant regulation, and enhances the adaptability and response speed of the regulation scheme.

[0116] Step S5: According to each of the compensation load, control the corresponding each edge device to execute regulation, complete the regulation of flexible load.

[0117] In the embodiment, step S5 finally executes the compensation amount through the edge device to realize accurate regulation of flexible load and reduce overall load fluctuation and peak-valley difference.

[0118] For example, as Figure 3The diagram shows the overall load curve and the flexible load operation curve of a special transformer when it is not regulated. Taking a special transformer load that includes photovoltaic power generation, electric vehicle charging, elevators, air conditioning, etc. as an example, the load is divided into three parts according to the geographical distribution of the flexible load, and each part is regulated by edge devices.

[0119] in Figure 3 The blue line represents the overall load curve of the dedicated transformer, while the other curves represent the operating curves of the controllable flexible loads when uncontrolled. These curves are composed of uniformly sampled data at 5-minute intervals, totaling 288 load data points per day and 8460 data points over 30 days. This 30-day data is used for daily intelligent control of the flexible loads to reduce load fluctuations on each day. Considering the degree of control over the flexible loads, the maximum control power for each flexible load is set as: ΔP Lmax1 =ΔP Lmax2 =100; ΔP Lmax3 =30; where ΔP Lmax1 ΔP Lmax2 and ΔP Lmax3 These represent the adjustment ranges for three flexible loads.

[0120] When using this solution to flexibly regulate the three-part load, such as... Figure 4 The diagram illustrates a comparison of 30-day load peak-valley differences under different control methods. The blue line represents the load peak-valley difference without any load control. The red line represents the peak-valley difference obtained by using the proposed random swarm intelligence optimization algorithm to globally optimize the three loads and send the control rules to three edge devices, which then independently control the loads according to their respective instructions (i.e., without edge collaboration). The green line represents the 30-day load peak-valley difference under the proposed control method that includes both global optimization and edge collaboration. It is evident that this invention effectively reduces the load peak-valley difference, achieving better control performance.

[0121] To evaluate the control effect of the optimization algorithm, a daily load fluctuation W is defined. L for:

[0122] Where L j Ave represents the actual load value at time point j. L This represents the average load at all points in time on that day.

[0123] like Figure 5 A 30-day load fluctuation comparison table is shown in one embodiment, and Figure 6 The diagram illustrates the reduction in load fluctuation between the global optimization and edge collaboration schemes in one embodiment. As can be seen from the curve data, the edge collaboration scheme can significantly further reduce load fluctuation.

[0124] In summary, compared with the prior art, the above embodiments of the present application have the following beneficial effects: by predicting the flexible load data of the target regulation day, data basis is provided for optimization; by using a random swarm intelligence optimization algorithm to iteratively generate an optimal solution, the rationality of global load distribution is ensured; by decomposing the optimal solution according to the adjustable power range of each edge device, local overload is avoided; by adjusting the regulation solution according to the dynamic difference between the actual value and the predicted value, load deviation is compensated in real time; finally, by executing the compensation amount through the edge device, precise regulation of the flexible load is realized, and overall load fluctuation and peak-valley difference are reduced.

[0125] Embodiment Two:

[0126] Please refer to Figure 2 Based on the same inventive concept, the flexible load regulation system based on global optimization and edge coordination disclosed in the embodiments of the present application comprises a data acquisition module M1, a regulation solution acquisition module M2, a decomposition module M3, a compensation load calculation module M4, and a control execution module M5.

[0127] The data acquisition module M1 is configured to predict the predicted value of the flexible load of the target regulation day according to the flexible load data of the current power grid.

[0128] In this embodiment, the data acquisition module M1 provides data basis for optimization by predicting the flexible load data of the target regulation day.

[0129] The regulation solution acquisition module M2 is configured to iteratively optimize an initial load regulation solution according to the predicted value of the flexible load of the power grid and a preset random swarm intelligence optimization algorithm, and obtain an optimal load regulation solution.

[0130] Further, the regulation solution acquisition module M2 comprises an initialization unit and a regulation solution iteration unit.

[0131] The initialization unit is configured to initialize a temperature value and represent the predicted value of the flexible load of the power grid as an initial load regulation solution.

[0132] The regulation solution iteration unit is configured to iteratively update the temperature value according to a preset simulated annealing optimization algorithm until the current temperature value reaches a preset critical point, and output the current load regulation solution as the optimal load regulation solution.

[0133] Wherein, after updating the temperature value each time, the first load regulation solution is iteratively generated according to the preset simulated annealing optimization algorithm and constraint conditions, the loss function difference between the first load regulation solution and the current load regulation solution is calculated according to the preset loss function, and the current load regulation solution is iteratively updated according to the loss function difference until a preset iteration number is reached; or when the loss difference between the load regulation solutions obtained by adjacent two iterations meets a preset threshold, the current load regulation solution is output as the optimal load regulation solution and the iteration is stopped.

[0134] In the preferred embodiment, the regulation solution obtaining module M2 provides a starting point for the simulated annealing optimization algorithm by initializing the temperature value and representing the predicted value of the grid flexible load as the initial load regulation solution, iteratively updates the temperature value until a critical point is reached, and outputs the current load regulation solution as the optimal load regulation solution, which can effectively avoid local optimization problems in a complex and variable environment and ensure that the best solution is found; after updating the temperature value each time, the first load regulation solution is iteratively generated based on the preset simulated annealing optimization algorithm and constraint conditions, and the loss function difference is calculated for iterative updating, which ensures that each step moves in the optimal direction while meeting all necessary constraints, thereby improving the robustness and effectiveness of the entire process.

[0135] Further, the regulation solution iteration unit comprises a first load regulation solution generation subunit.

[0136] The first load regulation solution generation subunit is configured to iteratively generate a first load regulation solution according to the simulated annealing optimization algorithm and constraint conditions; wherein the constraint conditions comprise maximum power constraint, power balance constraint, minimum cost and minimum loss constraint.

[0137] In the preferred embodiment, the first load regulation solution generation subunit prevents the flexible load regulation amount from exceeding the device bearing limit through the maximum power constraint, avoids the risk of device overload, and ensures the safety of grid operation; the power balance constraint ensures that the total load remains stable within the regulation period, avoiding power grid fluctuations caused by supply-demand imbalance; and the minimum cost and loss constraint limits the excessive reduction of cost or loss in the optimization process, preventing power grid shock or operation instability caused by regulation deviation, thereby improving economy while maintaining grid safety margin.

[0138] Further, the regulation solution obtaining module M2 further comprises a regulation solution iteration control unit.

[0139] The regulation solution iteration control unit is configured to minimize the preset target function as the optimization target of the random swarm intelligence optimization algorithm, iteratively optimize the initial load regulation solution, and obtain the optimal load regulation solution; wherein the optimization target is:

[0140] wherein min(f) represents an optimization objective, ∑ 24h () represents a statistical optimization in a day unit, P i is the i-th data of the predicted value of the flexible load of the power grid on the target day, P target is the average value of the predicted value of the flexible load of the power grid on the target day; Loss i is the power loss of the power grid at the i-th moment of the target day, Loss target,i is the acceptable target power loss at the i-th moment of the target day, and a and b are respectively a power balance factor and a power loss factor.

[0141] In the preferred embodiment, the regulation and decomposition iteration control unit takes the dual objective function of power balance and power loss as the core, reduces the overall load fluctuation within a day through 24-hour statistical optimization, dynamically adjusts the weights of the power balance factor and the loss factor to flexibly adapt to different period demands (such as focusing on peak shaving and valley filling on the peak side and focusing on loss optimization on the flat section), sets the target value based on the predicted average value to make the regulation scheme closer to the actual operation scenario, and improves the practicality and adaptability of the optimization result.

[0142] The decomposition module M3 is configured to decompose the optimal load regulation solution according to the adjustable power range corresponding to the flexible load of each edge device to obtain a first load regulation solution corresponding to each edge device.

[0143] In the embodiment, the decomposition module M3 can effectively avoid the problem of local overload by decomposing the optimal solution according to the adjustable power range of each edge device.

[0144] The compensation load calculation module M4 is configured to adjust the first load regulation solution of each edge device according to the difference between the actual value and the predicted value of the flexible load of each edge device in the target regulation day and a preset distribution rule to obtain the compensation load amount of each edge device.

[0145] Further, the compensation load calculation module M4 includes a total amount calculation unit and an iterative distribution unit.

[0146] The total amount calculation unit is configured to calculate the total amount of the compensation load of each edge device and the corresponding adjacent device according to the difference between the actual value and the corresponding predicted value of the flexible load of each edge device.

[0147] The iterative distribution unit is configured to iteratively adjust a dynamic distribution threshold in a preset value interval according to a segment search algorithm with a preset loss function as a target, and adjust the first load control solution of the edge device and the adjacent device based on the dynamic distribution threshold, a target load average of each edge device, and a proportional relationship of the first load control solution, until the loss function reaches a minimum value, so as to obtain a compensation load of each edge device; wherein a sum of the compensation loads of each edge device and the adjacent device is equal to a total amount of the load to be compensated.

[0148] In the preferred embodiment, the compensation load calculation module M4 determines the total amount to be compensated by calculating the difference between the actual and predicted values, accurately locates the deviation source, dynamically adjusts the distribution threshold by using the segment search algorithm, quickly locks the optimal compensation ratio, and distributes the compensation amount based on the proportional relationship of the target load average and the control solution, so as to ensure the uniform distribution of the compensated load, avoid overloading or underloading in local areas, and further suppress fluctuations.

[0149] Further, the iterative distribution unit comprises a threshold adjustment subunit.

[0150] The threshold adjustment subunit is configured to iteratively adjust the dynamic distribution threshold in a preset value interval according to a preset loss function and a segment search algorithm; wherein the loss function is:

[0151]

[0152] ; wherein f r (ΔPc j ) represents a loss function, ΔPc j represents a compensation load of the edge device j, m represents the number of the edge device j and the corresponding adjacent device, represents a first load control solution of the edge device j, Lave j represents a target load average of the edge device j, and ΔPcom represents a total amount of the load to be compensated of the edge device j and the corresponding adjacent device.

[0153] In the preferred embodiment, the threshold adjustment subunit nonlinearly amplifies the difference between the load before and after compensation and the target average value by using an exponential loss function, preferentially suppresses the deviation in the high fluctuation area, forces the algorithm to preferentially eliminate the fluctuation in the large deviation area, and avoids the defect that the traditional square function excessively optimizes the local small deviation and ignores the key high fluctuation area; the distribution threshold is dynamically adjusted by using the segment search algorithm, the efficiency and accuracy balance of the compensation distribution is realized under the constraint that the total amount of compensation is consistent with the actual difference, and finally the power grid operation stability is improved by suppressing the global load fluctuation and the local peak-valley difference.

[0154] Further, the iterative allocation unit further comprises a screening subunit and a regulation solution adjustment subunit.

[0155] The screening subunit is configured to screen the to-be-compensated devices according to the dynamic allocation threshold, the proportional relationship between the target load average value of each edge device and the corresponding adjacent device and the first load regulation solution, and the positive or negative of the total amount of the corresponding to-be-compensated load.

[0156] The regulation solution adjustment subunit is configured to adjust the first load regulation solution of the to-be-compensated devices according to the dynamic allocation threshold to obtain the compensation load amount of each edge device, wherein the adjustment algorithm is as follows: Wherein λ represents the dynamic allocation threshold.

[0157] In the preferred embodiment, the screening subunit screens the high-deviation devices according to the dynamic allocation threshold, concentrates resources to compensate the key areas, and avoids low regulation efficiency caused by scattered resources; and the regulation solution adjustment subunit corrects the regulation solution in combination with the target load average value, ensures that the load of each device after compensation is closer to the target average value, further smooths the load curve, reduces the local peak-valley difference, and improves the stability of power grid operation.

[0158] Further, the screening subunit can be implemented through the following preferred embodiments, specifically as follows:

[0159] When the total amount of the to-be-compensated load of each edge device and the corresponding adjacent device is greater than zero, the proportion of the first load regulation solution of each edge device and the adjacent device to the target load average value is sorted from small to large, and the devices with a proportion value lower than the dynamic allocation threshold are screened as the to-be-compensated devices.

[0160] When the total amount of the to-be-compensated load of each edge device and the corresponding adjacent device is less than zero, the proportion of the first load regulation solution of each edge device and the adjacent device to the target load average value is sorted from large to small, and the devices with a proportion value higher than the dynamic allocation threshold are screened as the to-be-compensated devices.

[0161] In the preferred embodiment, the screening subunit sorts the devices according to the positive and negative of the total compensation amount, preferentially processes the areas with the largest deviation (such as preferentially allocating to the low-proportion devices in positive compensation, and preferentially reducing the high-proportion devices in negative compensation), and improves the resource utilization efficiency; the dynamic threshold is used to screen the key devices, the main deviation is concentrated to be solved, the redundant regulation is avoided, and the adaptability and response speed of the regulation scheme are enhanced.

[0162] The control execution module M5 is configured to control each edge device to perform regulation according to the compensation load amount, and complete the regulation of the flexible load.

[0163] In the embodiment, the control execution module M5 finally executes the compensation amount through the edge device, realizes accurate regulation and control of the flexible load, and reduces overall load fluctuation and peak-valley difference.

[0164] To sum up, compared with the prior art, the embodiment has the following beneficial effects: the flexible load data of the power grid on the target regulation day is predicted to provide data basis for optimization; a random swarm intelligence optimization algorithm is used to iteratively generate an optimal solution to ensure the rationality of global load distribution; the optimal solution is decomposed according to the adjustable power range of each edge device to avoid local overload; the regulation solution is adjusted according to the dynamic difference between the actual value and the predicted value to compensate for load deviation in real time; finally, the compensation amount is executed through the edge device to realize accurate regulation and control of the flexible load, and reduce overall load fluctuation and peak-valley difference.

[0165] The specific working processes of the modules described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described here. The division of the modules is only a logical functional division, and another division mode can be used in actual implementation, for example, multiple modules can be combined or integrated into another system.

[0166] The above specific embodiments further specifically describe the purposes, technical solutions and beneficial effects of the present application. It should be understood that the above description is only for specific embodiments of the present application and is not used to limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A flexible load regulation method based on global optimization and edge collaboration, characterized in that, include: Based on the current flexible load data of the power grid, predict the flexible load forecast value of the power grid corresponding to the target control day; Based on the predicted flexible load of the power grid and the preset stochastic swarm intelligent optimization algorithm, the initial load control solution is iteratively optimized to obtain the optimal load control solution; Based on the adjustable power range corresponding to the flexible load of each edge device, the optimal load control solution is decomposed to obtain the first load control solution corresponding to each edge device. Based on the difference between the actual and predicted values ​​of the flexible load of each edge device during the target control day, and the preset allocation rules, adjust each of the first load control solutions to obtain the compensation load amount of each edge device; Based on the compensation load amount, control the corresponding edge devices to perform regulation and control, and complete the regulation of flexible load.

2. The flexible load regulation method based on global optimization and edge collaboration as described in claim 1, characterized in that, The step of iteratively optimizing the initial load control solution based on the power grid flexible load forecast and a preset stochastic swarm intelligence optimization algorithm to obtain the optimal load control solution includes: Initialize the temperature value and characterize the power grid flexible load prediction value as the initial load regulation solution; According to the preset simulated annealing optimization algorithm, the temperature value is iteratively updated until the current temperature value reaches the preset critical point, and the current load control solution is output as the optimal load control solution. Specifically, after each temperature value update, a first load control solution is iteratively generated based on a preset simulated annealing optimization algorithm and constraints. The difference in loss function between the first load control solution and the current load control solution is calculated based on a preset loss function. The current load control solution is iteratively updated based on the difference in loss function until a preset number of iterations is reached. Alternatively, when the loss difference between the load control solutions obtained from two adjacent iterations meets a preset threshold, the current load control solution is output as the optimal load control solution and the iteration stops.

3. The flexible load regulation method based on global optimization and edge collaboration as described in claim 2, characterized in that, The step of iteratively generating the first load control solution based on a preset simulated annealing optimization algorithm and constraints includes: Based on the simulated annealing optimization algorithm and constraints, the first load regulation solution is generated iteratively; wherein the constraints include maximum power constraint, power balance constraint, minimum cost constraint, and minimum loss constraint.

4. The flexible load regulation method based on global optimization and edge collaboration as described in claim 1, characterized in that, The step of iteratively optimizing the initial load control solution based on the power grid flexible load forecast value and a preset stochastic swarm intelligence optimization algorithm to obtain the optimal load control solution further includes: Minimizing a preset objective function is used as the optimization objective of the stochastic swarm intelligence optimization algorithm. The initial load control solution is iteratively optimized to obtain the optimal load control solution; wherein the optimization objective is: Where min(f) represents the optimization objective, ∑ 24h () indicates that statistical optimization is performed on a daily basis, P i For the i-th data point of the power grid flexible load forecast for the target date, P target The average value of the grid flexible load forecast for the target date; Loss i Loss represents the grid power loss at time i on the target day. target,i Let be the acceptable target grid loss at time i on the target day, and let a and b be the power balance factor and grid loss factor, respectively.

5. The flexible load regulation method based on global optimization and edge collaboration as described in claim 1, characterized in that, The step of adjusting each of the first load control solutions based on the difference between the actual and predicted values ​​of the flexible load of each edge device during the target control day, and according to a preset allocation rule, to obtain the compensation load amount of each edge device includes: Based on the difference between the actual value of the flexible load of each edge device and the corresponding predicted value of the flexible load, calculate the total amount of load to be compensated for each edge device and its corresponding adjacent devices. With the goal of minimizing a preset loss function, a dynamic allocation threshold is iteratively adjusted within a preset value range using a piecewise search algorithm. Based on the dynamic allocation threshold and the proportional relationship between the target load average value of each edge device and the first load control solution, the first load control solution of the edge device and its neighboring devices is adjusted until the loss function reaches its minimum value, thereby obtaining the compensation load amount of each edge device. The sum of the compensation load amounts of each edge device and its neighboring devices is equal to the corresponding total load to be compensated.

6. The flexible load regulation method based on global optimization and edge collaboration as described in claim 5, characterized in that, The step of iteratively adjusting the dynamically allocated threshold within a preset value range based on a piecewise search algorithm, with the objective of minimizing a preset loss function, includes: The dynamically allocated threshold is iteratively adjusted within a preset value range based on a preset loss function and a segmented search algorithm; wherein, the loss function is: Among them, f r (ΔPc j ) represents the loss function, ΔPc j Let m represent the compensation load of edge device j, and m represent the number of edge device j and its corresponding neighboring devices. Lave represents the first load regulation solution for edge device j. j ΔPcom represents the target average load of edge device j, and ΔPcom represents the total load to be compensated for edge device j and its corresponding neighboring devices.

7. The flexible load regulation method based on global optimization and edge collaboration as described in claim 6, characterized in that, The adjustment of the first load control solution for the edge devices and neighboring devices based on the dynamic allocation threshold and the proportional relationship between the target load average of each edge device and the first load control solution, until the loss function reaches its minimum value, yields the compensated load amount for each edge device, including: Based on the dynamic allocation threshold, the ratio of the target load average of each edge device and its corresponding neighboring device to the first load control solution, and the sign of the corresponding total load to be compensated, the devices to be compensated are selected. Based on the dynamic allocation threshold, the first load control solution of the equipment to be compensated is adjusted to obtain the compensation load amount of each edge equipment; wherein the adjustment algorithm is as follows: Where λ represents the dynamically allocated threshold.

8. The flexible load regulation method based on global optimization and edge collaboration as described in claim 7, characterized in that, The step of screening devices to be compensated based on the dynamic allocation threshold, the ratio of the target load average of each edge device and its corresponding neighboring device to the first load control solution, and the positive or negative sign of the corresponding total load to be compensated, includes: When the total load to be compensated for each edge device and its corresponding neighboring device is greater than zero, the ratio of the first load control solution of each edge device and its neighboring device to the average target load is sorted from small to large and devices with a ratio value lower than the dynamic allocation threshold are selected as devices to be compensated. When the total load to be compensated for each edge device and its corresponding neighboring device is less than zero, the ratio of the first load control solution of each edge device and its neighboring device to the average value of the target load is sorted from largest to smallest, and devices with a ratio value higher than the dynamic allocation threshold are selected as devices to be compensated.

9. A flexible load control system based on global optimization and edge collaboration, characterized in that, include: The module includes a data acquisition module, a control solution acquisition module, a decomposition module, a compensation load calculation module, and a control execution module. The data acquisition module is used to predict the power grid flexible load forecast value corresponding to the target control day based on the current power grid flexible load data. The control solution acquisition module is used to iteratively optimize the initial load control solution based on the power grid flexible load forecast value and a preset stochastic swarm intelligent optimization algorithm to obtain the optimal load control solution. The decomposition module is used to decompose the optimal load control solution according to the adjustable power range corresponding to the flexible load of each edge device, and obtain the first load control solution corresponding to each edge device. The compensation load calculation module is used to adjust each of the first load control solutions based on the difference between the actual value and the predicted value of the flexible load of each edge device during the target control day, as well as the preset allocation rules, to obtain the compensation load amount of each edge device. The control execution module is used to control the corresponding edge devices to perform regulation according to the compensation load amount, so as to complete the regulation of flexible load.

10. A flexible load control system based on global optimization and edge collaboration as described in claim 9, characterized in that, The control solution acquisition module includes an initialization unit and a control solution iteration unit; The initialization unit is used to initialize the temperature value and characterize the power grid flexible load prediction value as an initial load control solution. The load control solution iteration unit is used to iteratively update the temperature value according to a preset simulated annealing optimization algorithm until the current temperature value reaches a preset critical point, and output the current load control solution as the optimal load control solution. Specifically, after each temperature value update, a first load control solution is iteratively generated based on a preset simulated annealing optimization algorithm and constraints. The difference in loss function between the first load control solution and the current load control solution is calculated based on a preset loss function. The current load control solution is iteratively updated based on the difference in loss function until a preset number of iterations is reached. Alternatively, when the loss difference between the load control solutions obtained from two adjacent iterations meets a preset threshold, the current load control solution is output as the optimal load control solution and the iteration stops.

Images