Power optimization distribution method, system, device and medium based on active regulation capability of distributed photovoltaic cluster

Abstract

The application discloses a power optimization distribution method, system, device and medium based on active regulation capacity of a distributed photovoltaic cluster, and the method comprises the following steps: extracting historical data, selecting a sample inverter, and giving a calculation formula of the active regulation capacity value of the distributed photovoltaic cluster; training a time convolution network by using the historical data, and constructing a function mapping relationship among the active regulation capacity value of the sample inverter, the active output value, the target power value and a plurality of environmental factors; inputting experimental data into the time convolution network, calculating the active regulation capacity value of the inverter, and establishing a cluster power distribution model based on equal proportion distribution of the sample inverter; generating an active regulation capacity value sorting sequence based on the active regulation capacity value of a subordinate area of the distributed photovoltaic inverter cluster, and dividing three levels to which each inverter belongs; and solving the model by using a maximum power point estimation algorithm to obtain a power optimization distribution scheme. The application optimizes power distribution among the distributed photovoltaic clusters.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of new energy grid-connected power generation, and particularly relates to a power optimization distribution method, system, device and medium based on active regulation capability of a distributed photovoltaic cluster. BACKGROUND

[0002] With the vigorous promotion of renewable energy such as photovoltaic and wind power in China, the grid-connected power generation capacity of new energy is continuously improved, and the proportion of new energy in some regions has reached 30%, which reduces the proportion of spinning reserve power generation capacity of conventional frequency modulation resources such as hydraulic and thermal power plants in the region, and the structural contradiction of power supply is increasingly prominent. Moreover, due to the intermittency and volatility of renewable energy output, the peak-valley difference of the power grid increases, which brings difficulties to making a reasonable power generation plan. And with the commissioning of multiple large-capacity and ultra-high voltage direct current transmission projects, the system frequency stability risk caused by ultra-high voltage direct current blocking or power sudden drop is highlighted, which makes the power grid also put forward the requirements of fast speed and high precision of frequency modulation control response to new energy clusters. Therefore, it is necessary to optimize the power distribution scheme in the station when the new energy cluster participates in dispatching response, and to promote the participation of new energy distributed clusters in the work of fast frequency response of the power grid.

[0003] At present, there are few literatures about the active power optimization distribution scheme of distributed photovoltaic clusters, and the distribution scheme is relatively direct. The traditional power distribution scheme fails to fully utilize the massive historical data in the cluster, and also fails to consider the actual power regulation level of the photovoltaic cluster after the inverter receives the power regulation instruction. There are also few literatures about the calculation method of the active regulation capability value of the distributed photovoltaic. Implementing new energy fast frequency modulation may encounter the following three substandard situations: (1) the first round of power regulation is accurately executed, but the response time is over-standard; (2) the first round of response is time-standard, but the deviation is large and does not meet the standard, and there is no extra time for secondary fine-tuning; (3) both the time and accuracy of the first round of regulation do not meet the standard. No matter which of the above situations, to optimize and make it meet the standard depends on whether the first round of power distribution is appropriate and whether the fast and accurate execution of the inverter is obtained. Among them, whether the power distribution between clusters can be reasonable needs the active regulation capability data of each inverter or each party array inverter group as distribution data support.

[0004] Therefore, in view of the fast frequency response demand of the photovoltaic cluster, a data-driven method is adopted, and the power distribution of the photovoltaic cluster is researched, which has important significance for improving the power distribution between clusters and the response speed of the conventional inverter power execution of the photovoltaic cluster. SUMMARY

[0005] The present invention addresses the shortcomings of the existing average power allocation algorithm, which suffers from low accuracy, and the slow power execution response speed of conventional inverters. The present invention provides a method, system, device, and medium for optimizing power allocation based on the active power regulation capability of distributed photovoltaic clusters. A temporal convolutional network is introduced to establish a proportional allocation model based on the active power regulation capability values ​​of constructed sample inverters. This model generates a sorted sequence of cluster inverter power generation capability values, and divides each inverter into its current level to optimize the power allocation scheme when participating in the dispatch response of new energy distributed clusters, thereby promoting the participation of new energy clusters in the rapid frequency response of the power grid.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A power optimization allocation method based on the active power regulation capability of a distributed photovoltaic cluster includes the following steps:

[0008] S1. Select sample inverters from each area of ​​the distributed photovoltaic cluster to define a calculation formula for the active power regulation capability value of the distributed photovoltaic cluster;

[0009] S2. Extract historical data of distributed photovoltaic clusters, perform feature screening on the historical data of distributed photovoltaic clusters, and select data related to active power regulation capability values ​​to construct training feature sets and test feature sets;

[0010] S3. Training a temporal convolutional network using the training feature set, so that the temporal convolutional network fits a functional mapping relationship between the active power regulation capability value and the active power output value, the target power value, and multiple environmental factors;

[0011] S4. Input the test feature set into the trained temporal convolutional network to predict the inverter active regulation capability value; establish a cluster power allocation model based on equal proportion allocation of sample inverters;

[0012] S5. Sort the predicted active power regulation capability values ​​of the inverters in each area of ​​the distributed photovoltaic cluster, and divide the inverters in each area into three levels;

[0013] S6. Determine the inverter level involved in power regulation based on the ratio of the scheduling instruction adjustment amount of the cluster's total target power and the cluster's total target power execution value at the current moment. Take the hierarchical relationship of each inverter as a constraint condition, and use the maximum power point estimation algorithm to solve the cluster power distribution model based on proportional distribution of inverters to obtain the active power optimization distribution plan of the distributed photovoltaic cluster.

[0014] To optimize the above technical solutions, specific measures taken also include:

[0015] Furthermore, in step S1, the calculation formula for defining the active power regulation capability value of the distributed photovoltaic cluster is specifically as follows:

[0016] The calculation formula for the current active power regulation capability value of each area of ​​the distributed photovoltaic cluster is expressed as:

[0017]

[0018] Where, P n_f_max Indicates the current active power regulation capability value of the nth area. represents the average natural power generation capacity of m sample inverters in the area, m represents the number of sample inverters selected in the nth area, m * Indicates the number of power generation units that can operate normally in an area;

[0019] Among them, the average natural power generation capacity of m sample inverters The expression is as follows:

[0020]

[0021] Where, P mi represents the natural maximum power generation of the i-th sample inverter.

[0022] Furthermore, in step S2, the feature screening of the historical data of the distributed photovoltaic cluster is specifically performed as follows:

[0023] The power output value, target power value and environmental factors of each inverter are constructed into a feature set, where the environmental factors include ambient temperature, light intensity, ambient humidity and weather conditions.

[0024] Furthermore, in step S3, the structure of the temporal convolutional network includes:

[0025] Causal convolution for processing sequence information and dilated causal convolution and residual modules with memory for historical data.

[0026] Furthermore, in step S4, the establishment of a cluster power distribution model based on equal-proportion distribution of sample inverters is specifically as follows:

[0027] Based on the inverter active regulation capability value predicted by the time convolutional network, the active regulation capability value of each region in the current distributed photovoltaic cluster is calculated, and the active regulation capability value of each region is proportionally distributed. The expression of the cluster power distribution model based on the proportional distribution of sample inverters is as follows:

[0028]

[0029] Where, P n_f_set is the power to be allocated to the nth area, Pcluster_k P is the cluster total target power execution value at the current k moment n_f_max P is the active regulation capacity value of the nth region, and N is the number of regions divided by the distributed photovoltaic cluster;

[0030] The cluster total target power execution value P cluster_k The expression is as follows:

[0031] P cluster_k = P cluster_k-2s + DP fast_k

[0032] In the formula, P cluster_k-2s is the actual power generation of the photovoltaic cluster grid-connected point 2s before the trigger k moment, and DP fast_k is the scheduling instruction adjustment amount of the cluster total target power; the value of DP fast_k is calculated based on the cluster grid-connected point frequency f at the current k moment through the P-f frequency modulation curve function provided by the local power grid scheduling department, and the formula is:

[0033]

[0034] Wherein, f is the cluster grid-connected point frequency, f d is the fast frequency response action threshold, P N is the cluster rated total power, f N is the rated frequency, and d% is the regulation difference rate.

[0035] Further, in step S6, the ratio of the scheduling instruction adjustment amount of the cluster total target power and the cluster total target power execution value at the current moment is used to determine the inverter level participating in power regulation, and the ratio is k.

[0036] When the ratio k of the scheduling instruction adjustment amount DP fast_k of the cluster total target power and the cluster total target power execution value P cluster_k at the current k moment is less than the first threshold K1, only the first layer of inverters in the region is used to participate in power regulation; when k is greater than or equal to the first threshold K1 and less than the second threshold K2, the first and second layers of inverters in the region are used to participate in power regulation; when k is greater than or equal to the second threshold K2, all the first, second and third layers of inverters participate in the power regulation task; in the level not participating in power regulation, the inverter will maintain the existing output.

[0037] Further, in step S6, the maximum power point estimation algorithm is used to solve the cluster power distribution model based on the equal proportion distribution of the inverter.

[0038] S61, set the total number of inverters M and the natural maximum power generation P max,i of each inverter; determine the cluster total target power Ptarget ; confirm the hierarchical relationship and power regulation capability value of each inverter;

[0039] S62, according to the power regulation capability value of each inverter, calculate its initial power allocation P i , ensure Adjust the power allocated to each inverter to ensure that it does not exceed the natural maximum power P max,i of each inverter;

[0040] S63, adopt iterative method to optimize power allocation; in each iteration, according to the current power allocation and the real-time efficiency of each inverter, adjust the power allocation P i To approach the natural maximum power generation;

[0041] S64, check whether the power allocation reaches or approaches the predetermined P target , while the output of each inverter approaches its natural maximum power generation; ensure that the adjusted power allocation still satisfies the constraint condition of the hierarchical relationship of each inverter; once the algorithm converges or reaches the preset number of iterations, output the final power allocation scheme, otherwise, continue to adjust the power allocation.

[0042] The application also provides a power optimization distribution system based on the active regulation capability of a distributed photovoltaic cluster, comprising:

[0043] A distributed photovoltaic cluster active regulation capability value definition module is used to select sample inverters from each area of the distributed photovoltaic cluster, and is used to define the calculation formula of the distributed photovoltaic cluster active regulation capability value.

[0044] A data set establishment module is used to extract historical data of the distributed photovoltaic cluster, perform feature screening on the historical data of the distributed photovoltaic cluster, select data related to the active regulation capability value to construct a training feature set and a test feature set.

[0045] A training module is used to train a time convolution network by using the training feature set, so that the time convolution network fits the functional mapping relationship between the active regulation capability value and the active output value, the target power value, and multiple environmental factors.

[0046] A cluster power distribution model establishment module is used to input the test feature set into the trained time convolution network to predict the active regulation capability value of the inverter; and establish a cluster power distribution model based on the sample inverter proportional distribution.

[0047] A hierarchical module is used to sort the predicted active regulation capability value of the inverters in each area of the distributed photovoltaic cluster, and divide the inverters in each area into three levels.

[0048] The solving module is used for determining an inverter level participating in power adjustment according to a ratio of a scheduling instruction adjustment amount of the total target power of the cluster and a current time execution value of the total target power of the cluster, taking a hierarchical relationship of each inverter as a constraint condition, and solving a cluster power distribution model based on inverter equal proportion distribution by using a maximum power point estimation algorithm to obtain an active power optimization distribution scheme of the distributed photovoltaic cluster.

[0049] The application further provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements the power optimization distribution method based on the active regulation capability of the distributed photovoltaic cluster when running the computer program.

[0050] The application further provides a computer readable storage medium, which stores a computer program, and the computer program enables a computer to execute the power optimization distribution method based on the active regulation capability of the distributed photovoltaic cluster.

[0051] The application has the advantages that: in the application, firstly, historical data is extracted, and a calculation method of the active regulation capability value of the distributed photovoltaic cluster is given by selecting a sample inverter; then, a time convolution network is introduced to perform deep learning on the historical data, and a function mapping relationship between the active regulation capability value of the sample inverter and the active output value, the target power value, and multiple environmental factors is constructed; experimental data is input into the time convolution network, the active regulation capability value of the inverter is calculated, and a cluster power distribution model based on equal proportion distribution of the sample inverter is established; then, based on the active regulation capability value of the subordinate region of the distributed photovoltaic inverter cluster, an active regulation capability value sorting sequence is generated, and three levels to which each inverter belongs are divided; finally, a maximum power point estimation algorithm is used to solve the model, and the active power optimization distribution of the distributed photovoltaic inverter cluster is completed.

[0052] The application can optimize power distribution among the distributed photovoltaic cluster, has high power distribution accuracy, and has fast inverter power execution response speed. The system frequency recovery is more optimal. The method of inverter hierarchical sorting effectively reduces the number of optimization variables, improves the solving speed of the model, and uses the maximum power point estimation method to solve the model, so that the photovoltaic output can be directly controlled at the target power point. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 A flowchart of the power optimization distribution method based on the active regulation capability of the photovoltaic cluster according to the application;

[0054] Figure 2 A flowchart of the inverter hierarchical sorting in the embodiment of the application;

[0055] Figure 3Flow chart of MPPE algorithm in the embodiment of the present application

[0056] Figure 4 Photovoltaic unit structure in the embodiment of the present application

[0057] Figure 5 Frequency recovery curves of different methods DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0059] Embodiment one

[0060] The present application provides a power optimization distribution method based on active regulation capacity of distributed photovoltaic cluster, the flow of the method is as shown in Figure 1 The method comprises the following steps:

[0061] S1, selecting sample inverters from each region of the distributed photovoltaic cluster, for defining a calculation formula of the active regulation capacity value of the distributed photovoltaic cluster; the calculation formula of the active regulation capacity value of the distributed photovoltaic cluster is defined as follows:

[0062] The calculation formula of the current active regulation capacity value of each region of the distributed photovoltaic cluster is expressed as:

[0063]

[0064] In the formula, P n_f_max represents the current active regulation capacity value of the nth region, represents the average natural power generation capacity of the m sample inverters in the region, m represents the number of sample inverters selected in the nth region, and m * represents the number of power generation units that can normally operate in a region.

[0065] The expression of the average natural power generation capacity of the m sample inverters is as follows:

[0066]

[0067] In the formula, P mi represents the natural maximum power generation of the ith sample inverter.

[0068] ​S2, extract the historical data of the distributed photovoltaic cluster, perform feature screening on the historical data of the distributed photovoltaic cluster, select data related to the active regulation capacity value to construct a training feature set and a test feature set; the feature screening on the historical data of the distributed photovoltaic cluster is specifically:

[0069] The power output value P of each inverter at j time points output,j , the target power value P target,j , and four environmental factors (environmental temperature T j , light intensity I j , environmental humidity H j , and weather W j ) are constructed as a feature set, input to the TCN for supervised learning, and the learning goal is the active regulation capacity value of the inverter;

[0070] The input matrix A is:

[0071]

[0072] The data is organized in chronological order, with each row representing a feature and each column representing an observation at a time point. The weather needs to be converted into a numerical value (sunny = 0, cloudy = 1, rainy = 2).

[0073] S3, train the time convolution network (TCN) using the training feature set, so that the time convolution network fits the functional mapping relationship between the active regulation capacity value and the active power output value, the target power value, and multiple environmental factors; the structure of the time convolution network includes:

[0074] Causal convolution for processing sequence information and expansion causal convolution with memory for historical data and residual module.

[0075] S4, input the test feature set into the trained time convolution network to predict the active regulation capacity value of the inverter; establish a cluster power allocation model based on sample inverters in proportion; the establishment of the cluster power allocation model based on sample inverters in proportion is specifically:

[0076] Based on the active regulation capacity value of the inverter predicted by the time convolution network, the active regulation capacity value of each region in the current distributed photovoltaic cluster is calculated, and the active regulation capacity value of each region is allocated in proportion to obtain the expression of the cluster power allocation model based on sample inverters in proportion as follows:

[0077]

[0078] In the formula, P n_f_set is the power to be allocated to the nth region, P cluster_k is the total target power execution value (actual value) of the cluster at the current k time, and Pn_f_max is the active power regulation capability value of the nth region, and N is the number of regions divided by the distributed photovoltaic cluster;

[0079] Cluster total target power execution value P cluster_k The expression is as follows:

[0080] P cluster_k =P cluster_k-2s +DP fast_k

[0081] Where, P cluster_k-2s is the actual power generation of the photovoltaic cluster grid-connected point 2s before triggering time k, DP fast_k is the scheduling instruction adjustment amount of the total target power of the cluster; DP fast_k The value of is calculated at the current time k based on the cluster grid connection point frequency f, using the Pf frequency regulation curve function provided by the local power grid dispatching department. The formula is:

[0082]

[0083] Among them, f is the frequency of the cluster grid connection point, f d is the fast frequency response action threshold, P N is the total rated power of the cluster, f N is the rated frequency, and d% is the modulation rate.

[0084] S5. Sort the predicted active power regulation capability values ​​of the inverters in each area of ​​the distributed photovoltaic cluster, and divide the inverters in each area into three levels;

[0085] S6. Determine the inverter level involved in power regulation based on the ratio of the scheduling instruction adjustment amount of the cluster total target power and the current cluster total target power execution value, such as Figure 2 As shown, specifically: when the scheduling instruction adjustment amount DP of the total target power of the cluster fast_k and the total target power execution value P of the cluster at the current k moment cluster_k When the ratio k is less than the first threshold K1, only the inverters in the first layer of the area are used to participate in power regulation; when k is greater than or equal to the first threshold K1 and less than the second threshold K2, the inverters in the first and second layers of the area are used to participate in power regulation; when k is greater than or equal to the second threshold K2, all inverters in the first, second and third layers participate in the power regulation task; in the layers that do not participate in power regulation, the inverters will maintain the existing output.

[0086] Taking the hierarchical relationship of each inverter as a constraint condition, the maximum power point estimation (MPPE) algorithm is used to solve the cluster power distribution model based on the proportional distribution of inverters, and the active power optimization distribution scheme of the distributed photovoltaic cluster is obtained. Figure 3 As shown, specifically:

[0087] S61, set the total number of inverters M and the natural maximum power generation P of each inverter max,i ; determine the total target power P of the cluster target ; confirm the hierarchical relationship and power regulation capability value of each inverter; here, the inverter refers to all inverters in the distributed photovoltaic cluster, not only the sample inverters selected in step S1.

[0088] S62, calculate the initial power allocation P of each inverter according to the power regulation capability value of each inverter i , to ensure adjust the power allocated to each inverter to ensure that it does not exceed the natural maximum power generation P of each inverter max,i ;

[0089] S63, use an iterative method to optimize the power allocation; in each iteration, adjust the power allocation P according to the current power allocation and the real-time efficiency of each inverter i to approach the natural maximum power generation;

[0090] S64, check whether the power allocation reaches or approaches the predetermined P target , while the output of each inverter approaches its natural maximum power generation; ensure that the adjusted power allocation still satisfies the constraint conditions of the hierarchical relationship of each inverter; once the algorithm converges or reaches the preset number of iterations, output the final power allocation scheme, otherwise, continue to adjust the power allocation.

[0091] Embodiment two

[0092] The present application proposes a power optimization distribution system based on the active regulation capability of a distributed photovoltaic cluster corresponding to the method of embodiment one, comprising:

[0093] A distributed photovoltaic cluster active regulation capability value definition module is used to select sample inverters from each region of the distributed photovoltaic cluster, and is used to define the calculation formula of the active regulation capability value of the distributed photovoltaic cluster;

[0094] A data set establishment module is used to extract historical data of the distributed photovoltaic cluster, perform feature screening on the historical data of the distributed photovoltaic cluster, and select data related to the active regulation capability value to construct a training feature set and a test feature set;

[0095] A training module is used to train a time convolution network using the training feature set, so that the time convolution network fits the functional mapping relationship between the active regulation capability value and the active power value, the target power value, and multiple environmental factors;

[0096] The cluster power distribution model establishing module is configured to input the test feature set into the trained time convolution network to predict the active regulation capability value of the inverter.

[0097] The hierarchical module is configured to sort the active regulation capability values of the inverters in each region of the predicted distributed photovoltaic cluster, and divide the inverters in each region into three levels.

[0098] The solving module is configured to determine the inverter level participating in power regulation according to the ratio of the scheduling instruction adjustment amount of the cluster total target power to the current time cluster total target power execution value, take the level relationship of each inverter as a constraint condition, and solve the cluster power distribution model based on inverter equal proportion distribution by using a maximum power point estimation algorithm to obtain an active power optimization distribution scheme of the distributed photovoltaic cluster.

[0099] The implementation manners of the modules and the functions of the modules in the system are completely consistent with the method steps of the first embodiment, and thus will not be described here.

[0100] Embodiment three

[0101] The present application provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the power optimization distribution method based on the active regulation capability of the distributed photovoltaic cluster is realized.

[0102] Embodiment four

[0103] The present application provides a computer readable storage medium, which stores a computer program, wherein the computer program enables a computer to execute the power optimization distribution method based on the active regulation capability of the distributed photovoltaic cluster.

[0104] The effects of the present application will be illustrated below according to experiments.

[0105] The experimental data set is from a distributed photovoltaic with an installed capacity of 3.26 MW in a certain city, which is composed of four inverters with a rated capacity of 500 kW and two inverters with a capacity of 630 kW. The dispatching center will take the high-voltage side of the main transformer as the active power examination point, which is mainly composed of the network loss of the box-type transformer, the collection line and the main transformer, and the structure of the photovoltaic unit is as shown in Figure 4 .

[0106] In order to verify the optimization effect, the commonly used equal proportion allocation strategy, the equal margin allocation strategy and the power optimization allocation strategy of the application are used respectively for simulation analysis of the whole day scheduling response. During the period of 12:30-15:40, the photovoltaic power station is in the scheduling response state, and needs to be adjusted according to the instruction value issued by the scheduling center every minute, and the rest of the time the inverter is in the MPPT operation state and does not perform power limitation. After the power allocation method proposed by the application, the average frequency deviation percentage is 0.63%, which is reduced by 1.48% and 1.38% compared with the equal proportion allocation strategy and the equal margin allocation strategy respectively. The system frequency recovery curves of different methods are shown in Figure 5 Table 1. Compared with other methods, the power optimization allocation method of the application has better system frequency recovery, smaller frequency fluctuation, faster frequency recovery speed, and verifies the superiority of the application.

[0107] Table 1. Compared with other methods, the power optimization allocation method of the application has better system frequency recovery, smaller frequency fluctuation, faster frequency recovery speed, and verifies the superiority of the application.

[0108]

[0109] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium which can contain or store programs for use by or in connection with an instruction execution system, device or apparatus. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or apparatus, or any suitable combination of the above. More specific examples of computer storage media can include one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0110] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed in the present application can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0111] The above are only preferred embodiments of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical scheme falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled in the art, some improvements and refinements without departing from the principles of the present application shall be considered as the protection scope of the present application.

Claims

1. A power optimization allocation method based on the active power regulation capability of a distributed photovoltaic cluster, characterized in that: The steps include: S1. Select sample inverters from each area of ​​the distributed photovoltaic cluster to define a calculation formula for the active power regulation capability value of the distributed photovoltaic cluster. In step S1, the calculation formula for defining the active power regulation capability value of the distributed photovoltaic cluster is as follows: The calculation formula for the current active power regulation capability value of each area of ​​the distributed photovoltaic cluster is expressed as: Where, P n_f_max Indicates the current active power regulation capability value of the nth area. represents the average natural power generation capacity of m sample inverters in the area, m represents the number of sample inverters selected in the nth area, m * Indicates the number of power generation units that can operate normally in an area; Among them, the average natural power generation capacity of m sample inverters The expression is as follows: Where, P mi represents the natural maximum power generation of the i-th sample inverter; S2. Extract historical data of distributed photovoltaic clusters, perform feature screening on the historical data of distributed photovoltaic clusters, and select data related to active power regulation capability values ​​to construct training feature sets and test feature sets; S3. Training a temporal convolutional network using the training feature set, so that the temporal convolutional network fits a functional mapping relationship between the active power regulation capability value and the active power output value, the target power value, and multiple environmental factors; S4. Input the test feature set into the trained temporal convolutional network to predict the inverter active regulation capability value; establish a cluster power distribution model based on equal proportion distribution of sample inverters; in step S4, the establishment of the cluster power distribution model based on equal proportion distribution of sample inverters is specifically as follows: Based on the inverter active regulation capability value predicted by the time convolutional network, the active regulation capability value of each region in the current distributed photovoltaic cluster is calculated, and the active regulation capability value of each region is proportionally distributed. The expression of the cluster power distribution model based on the proportional distribution of sample inverters is as follows: Where, P n_f_set is the power to be allocated to the nth area, P cluster_k is the total target power execution value of the cluster at the current k moment, P n_f_max is the active power regulation capability value of the nth region, and N is the number of regions divided by the distributed photovoltaic cluster; Cluster total target power execution value P cluster_k The expression is as follows: P cluster_k =P cluster_k-2s +ΔP fast_k Where, P cluster_k-2s is the actual power generation of the photovoltaic cluster grid-connected point 2s before triggering time k, ΔP fast_k is the scheduling instruction adjustment amount of the total target power of the cluster; ΔP fast_k The value of is calculated at the current time k based on the cluster grid connection point frequency f, using the Pf frequency regulation curve function provided by the local power grid dispatching department. The formula is: Among them, f is the frequency of the cluster grid connection point, f d is the fast frequency response action threshold, P N is the total rated power of the cluster, f N is the rated frequency, δ% is the modulation rate; S5. Sort the predicted active power regulation capability values ​​of the inverters in each area of ​​the distributed photovoltaic cluster, and divide the inverters in each area into three levels; S6. Determine the inverter hierarchy involved in power regulation based on the ratio of the dispatch instruction adjustment amount of the cluster's total target power and the cluster's total target power execution value at the current moment. Using the hierarchy relationship of each inverter as a constraint, use the maximum power point estimation algorithm to solve the cluster power allocation model based on equal-proportional allocation of inverters, and obtain an optimized active power allocation scheme for the distributed photovoltaic cluster. In step S6, the inverter level participating in power regulation is determined according to the ratio of the scheduling instruction adjustment amount of the cluster total target power and the cluster total target power execution value at the current moment: When the scheduling instruction adjustment amount of the total target power of the cluster is ΔP fast_k and the total target power execution value P of the cluster at the current k moment cluster_k When the ratio k is less than the first threshold K1, only the inverters in the first layer of the area are used to participate in power regulation; when k is greater than or equal to the first threshold K1 and less than the second threshold K2, the inverters in the first and second layers of the area are used to participate in power regulation; when k is greater than or equal to the second threshold K2, all inverters in the first, second and third layers participate in the power regulation task; in the layers not participating in power regulation, the inverters will maintain the current output; In step S6, the maximum power point estimation algorithm is used to solve the cluster power allocation model based on inverter proportional allocation as follows: S61. Set the total number of inverters M and the natural maximum power generation power P of each inverter. max,i ; Determine the total target power P of the cluster target ;Confirm the hierarchical relationship and power regulation capability value of each inverter; S62, calculate the initial power allocation P according to the power regulation capability value of each inverter i ,make sure Adjust the power allocated to each inverter to ensure that it does not exceed the natural maximum power generation power P of each inverter. max,i ; S63, using an iterative method to optimize power distribution; in each iteration, adjust the power distribution P according to the current power distribution and the real-time efficiency of each inverter. i With a power close to the natural maximum power P max,i ; S64, check whether the power distribution reaches or is close to the predetermined P target , and the output of each inverter is close to its natural maximum power generation power P max,i ; Ensure that the adjusted power distribution still meets the constraints of the hierarchical relationship of each inverter; once the algorithm converges or reaches the preset number of iterations, output the final power distribution plan, otherwise, continue to adjust the power distribution.

2. The power optimization allocation method based on the active power regulation capability of a distributed photovoltaic cluster according to claim 1 is characterized in that: In step S2, the feature screening of the historical data of the distributed photovoltaic cluster is specifically as follows: The power output value, target power value and environmental factors of each inverter are constructed into a feature set, where the environmental factors include ambient temperature, light intensity, ambient humidity and weather conditions.

3. The power optimization allocation method based on the active power regulation capability of a distributed photovoltaic cluster according to claim 1, characterized in that: In step S3, the structure of the temporal convolutional network includes: Causal convolution for processing sequence information and dilated causal convolution and residual modules with memory for historical data.

4. A power optimization distribution system based on the active power regulation capability of a distributed photovoltaic cluster implementing the method according to claim 1, characterized in that: include: A distributed photovoltaic cluster active power regulation capability value definition module is used to select sample inverters from each area of ​​the distributed photovoltaic cluster and to define a calculation formula for the distributed photovoltaic cluster active power regulation capability value; The data set establishment module is used to extract historical data of distributed photovoltaic clusters, perform feature screening on the historical data of distributed photovoltaic clusters, and select data related to active power regulation capability values ​​to construct training feature sets and test feature sets; A training module is used to train a temporal convolutional network using a training feature set, so that the temporal convolutional network fits a functional mapping relationship between an active power regulation capability value and an active power output value, a target power value, and multiple environmental factors; The cluster power allocation model establishment module is used to input the test feature set into the trained temporal convolutional network to predict the inverter active power regulation capability value; and establish a cluster power allocation model based on the proportional allocation of sample inverters; A hierarchical module is used to sort the predicted active power regulation capability values ​​of the inverters in each area of ​​the distributed photovoltaic cluster and divide the inverters in each area into three levels; The solution module is used to determine the inverter hierarchy involved in power regulation based on the ratio of the scheduling instruction adjustment amount of the cluster's total target power and the cluster's total target power execution value at the current moment. The hierarchical relationship of each inverter is used as a constraint condition, and the maximum power point estimation algorithm is used to solve the cluster power distribution model based on the proportional distribution of inverters to obtain the active power optimization distribution plan of the distributed photovoltaic cluster.

5. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the power optimization allocation method based on the active power regulation capability of the distributed photovoltaic cluster as described in any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables a computer to execute the power optimization allocation method based on the active power regulation capability of a distributed photovoltaic cluster as described in any one of claims 1 to 3.

Images