Power distribution network real-time voltage regulation method and device based on multi-strategy fusion algorithm

Abstract

The invention discloses a power distribution network real-time voltage regulation method and device based on a multi-strategy fusion algorithm, and relates to the field of power distribution network optimization scheduling, and the method comprises the steps: firstly collecting the historical operation data of a power distribution system, including the historical load data, the network frame data, the historical output data of a distributed power supply, and the like; clustering the historical operation data to obtain a typical source load state data set; solving a typical scene in the typical source load state data set offline based on a multi-strategy fusion algorithm, and constructing a knowledge base; carrying out ultra-short-term output prediction on the load and the distributed power supply, and carrying out system security check; and if a voltage problem occurs, combining historical experience in the knowledge base, and based on a multi-strategy fusion algorithm, carrying out power distribution network real-time voltage regulation strategy solution, so as to solve the voltage problem. According to the method, the operation challenge of the regional power distribution network under high-proportion new energy access is fully considered, the problem of insufficient efficiency of online voltage regulation is effectively relieved, and safe and economical operation of the power grid is ensured.

Description

Technical Field

[0001] The present application relates to the field of distribution network optimization and scheduling, and in particular to a distribution network real-time voltage regulation method and device based on a multi-strategy fusion algorithm. Background Art

[0002] To meet the requirements of a new, intelligent, controllable, eco-friendly, and sustainable power system, modern active distribution networks, with distributed renewable energy and increasingly diverse load demands, face unprecedented challenges. Reactive power optimization plays a vital role in distribution networks, reducing power flows along the network and maintaining the desired voltage profile under varying load demands. Reactive power optimization minimizes active power losses in the distribution network by optimizing the operation of various control devices. Traditional voltage regulation equipment, such as on-load tap-changing transformers and capacitor banks, is discrete and has slow regulation speeds and cannot be operated frequently, making it difficult to meet the control requirements of frequent voltage violations. While continuous compensation devices, such as static VAR compensators (SVCs), have a fast response time, their application is limited by their limited number of connections and high investment costs. In contrast, photovoltaic inverters offer the ability to quickly and continuously regulate reactive power at no additional cost, making them a crucial voltage regulation resource. Summary of the Invention

[0003] The purpose of this application is to provide a distribution network real-time voltage regulation method and device based on a multi-strategy fusion algorithm, which can improve the efficiency of online voltage regulation in the distribution network.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a real-time voltage regulation method for a distribution network based on a multi-strategy fusion algorithm, comprising:

[0006] Obtain historical operation data of the power distribution system;

[0007] Cluster the historical operation data to obtain a typical source-load status dataset;

[0008] Based on the multi-strategy fusion algorithm, we solve typical scenarios in typical source-load status datasets offline and build a transfer learning knowledge base.

[0009] Conduct ultra-short-term output forecasts for loads and distributed power sources, and perform system safety verification;

[0010] If a voltage problem occurs in the prediction results, the distribution network is regulated in real time based on the solution results of the multi-strategy fusion algorithm in combination with historical experience in the knowledge base.

[0011] Furthermore, K-meas clustering is used to cluster the historical operation data, and each group of data in the data set consists of the net load of each node in the distribution network.

[0012] Furthermore, the node voltage safety interval of the safety check is set to [0.93, 1.07]. In the power flow calculation results, nodes with a node voltage greater than 1.07 pu have an overvoltage problem, and nodes with a node voltage less than 0.93 pu have an undervoltage problem.

[0013] Furthermore, the multi-strategy fusion algorithm includes a snow ablation algorithm and a multi-strategy improvement of the snow ablation algorithm.

[0014] Furthermore, the snow melting algorithm includes an iterative process starting with a randomly generated population, represented by the following matrix:

[0015] ;

[0016] In the formula, N represents the population size, Dim represents the dimension of the solution space, L and U represent the lower bound and upper bound of the solution space respectively. Represents a randomly generated number in [0,1];

[0017] The solution space dimension is defined as 2N pv , where N p Represents the number of controllable PV inverters in the distribution network;

[0018] The randomness and irregularity of Brownian motion are used to describe the irregular movement of water vapor in space after snow or liquid water forms water vapor through sublimation or evaporation.

[0019] For standard Brownian motion, the step length of individual movement is obtained by using the probability density function of a normal distribution with a mean of 0 and a variance of 1; the mathematical formula of Brownian motion is as follows:

[0020] ;

[0021] The position update formula of the detection process in the snow melting algorithm is as follows:

[0022] ;

[0023] Where Z i (t) represents the i-th individual in the t-th iteration, RB i (t) is the Brownian motion random number vector; represents a random number in [0,1]; G(t) represents the optimal solution of the current population; Z i (t) represents the current position of the individual; It is an individual randomly selected from several elite groups in the population; is the center of mass position of the entire group in this update, and its mathematical expression is as follows:

[0024] ;

[0025] Where, and are the second best individual and the third best individual in the current population respectively, and the 50% individuals with the highest fitness in the population are called elite individuals. represents the centroid positions of these individuals;

[0026] ;

[0027] Where N1 represents the number of elite individuals, which is equal to half of N in value; It is a set consisting of the average of the best individual, the second best individual, the third best individual, and the elite individual in the group; during each position update during the exploration process, a random value is drawn from this set to assist in the position update;

[0028] When snow is converted into liquid water through melting, individuals are encouraged to develop high-quality solutions around the current best individual. The degree-day method is used to represent the snowmelt process:

[0029] ;

[0030] Where M is the snowmelt rate, DDF represents the degree-day factor, t is the number of iterations, and t max is the maximum number of iterations. T(t) is the temperature at time t;

[0031] The individual position update equation is as follows:

[0032] ;

[0033] Where, Represents a random number in [-1,1];

[0034] In the initial iteration of the snow melting algorithm, the entire population is randomly divided into two sub-populations P of equal size. a and P b , where P a Responsible for exploration, P b Responsible for development, the population size is N a and N b As the iterative process progresses, P b The scale of P gradually decreases, and a The scale increases accordingly.

[0035] Furthermore, the multi-strategy improvements to the snow melting algorithm include:

[0036] ;

[0037] The expression of parameter e is:

[0038] ;

[0039] Where, is the position of the random individual responsible for exploration in the tth iteration, e increases with the number of iterations, Is a random number between [-1,1];

[0040] For similar real-time voltage regulation problems in distribution networks, there is no need to repeatedly solve them. Transfer learning uses knowledge learned from other tasks to solve new problems.

[0041] The offline learning clusters historical source-load data to obtain a typical source-load state set, and solves it based on the HMSAO algorithm, and finally obtains the empirical knowledge of the corresponding source-load state; forms an RPO empirical knowledge base to store the knowledge of pre-learning convergence under different source-load states; finally, when solving online, the historical states in the knowledge base are matched based on the current system source-load state; after a successful match, the relevant knowledge is directly transferred for a quick solution; if the match fails, a reinforcement learning method without knowledge transfer is used to solve the problem, and the convergence result is used as new knowledge to supplement the knowledge base to achieve dynamic updating of the knowledge base.

[0042] The specific implementation method of knowledge transfer is as follows: when the source-load state of the new task of real-time voltage regulation of the distribution network has a certain similarity with a source task in the knowledge base, the knowledge matrix corresponding to the source task is directly transferred to the source task. As the initial knowledge matrix for online optimization tasks :

[0043] ;

[0044] Where, The similarity between the online source-load state of the real-time voltage regulation problem in the distribution network and the source-load state of the source task in the knowledge base is described. 、 are the number of loads and PV inverters, is the active power of the load and PV inverter, The threshold is set in advance. When there are multiple source tasks that meet the requirements, the most similar source task knowledge is adopted.

[0045] A second aspect of the present invention provides a real-time voltage regulation device for a distribution network based on a multi-strategy fusion algorithm, comprising:

[0046] Historical data collection module: used to obtain historical operation data of the power distribution system;

[0047] Source-load data set construction module: used to cluster historical operation data to obtain a typical source-load status data set;

[0048] Knowledge base construction module: used to solve typical scenarios in typical source-load status datasets offline based on a multi-strategy fusion algorithm and build a transfer learning knowledge base;

[0049] Safety verification module: used to make ultra-short-term output forecasts of loads and distributed power sources, and perform system safety verification;

[0050] Voltage regulation module: When voltage problems occur in the prediction results, it combines historical experience in the knowledge base and performs real-time voltage regulation on the distribution network based on the solution results of the multi-strategy fusion algorithm.

[0051] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for real-time voltage regulation of a distribution network.

[0052] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for real-time voltage regulation of a distribution network.

[0053] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the above-mentioned real-time voltage regulation method for the distribution network when executed by a processor.

[0054] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0055] This application provides a method and device for real-time voltage regulation in a distribution network based on a multi-strategy fusion algorithm. The optimization method, based on a snow melt algorithm, integrates a cold condensation strategy and a knowledge transfer learning strategy to improve the efficiency of online voltage control. This application fully considers the operational challenges of regional distribution networks under high-proportion renewable energy access, effectively alleviating the inefficiency of real-time voltage regulation and ensuring the safe and economical operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0057] Attachment Figure 1 This is a schematic diagram of the physical source of the SAO algorithm;

[0058] Attachment Figure 2 A map of ideas for knowledge transfer;

[0059] Attachment Figure 3 It is an IEEE33 node power distribution system with photovoltaic access;

[0060] Attachment Figure 4 It is the voltage distribution diagram of each node of the system before voltage regulation;

[0061] Attachment Figure 5 This is the voltage distribution diagram of each node in the system under real-time voltage regulation.

[0062] Attachment Figure 6 Schematic diagram of iterative convergence.

[0063] Attachment Figure 7 This is a functional module diagram of a real-time voltage regulation method for distribution networks based on a multi-strategy fusion algorithm;

[0064] Attachment Figure 8 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0065] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0066] The embodiment is carried out according to the steps of the present invention:

[0067] A method and device for real-time voltage regulation of a distribution network based on a multi-strategy fusion algorithm, comprising the following steps:

[0068] Step 1: Collect historical operation data of the distribution system, including historical load data, grid data, and historical output data of distributed power sources;

[0069] Step 2: Cluster the historical operation data to obtain a typical source-load status dataset;

[0070] Step 3: Based on the multi-strategy fusion algorithm, we solve typical scenarios in the typical source-load status dataset offline and build a transfer learning knowledge base.

[0071] Step 4: Conduct ultra-short-term output forecasts for loads and distributed generation, and perform system safety verification;

[0072] Step 5: If a voltage problem occurs, combine the historical experience in the knowledge base and solve the real-time voltage regulation strategy of the distribution network based on the multi-strategy fusion algorithm to solve the voltage problem.

[0073] Furthermore, the clustering method described in step 2 adopts K-meas clustering, and each set of data in the data set is composed of the net load of each node in the system;

[0074] Furthermore, the node voltage safety interval of the safety check in step 4 is set to [0.93, 1.07]. In the power flow calculation results, nodes with a node voltage greater than 1.07 pu have an overvoltage problem, and nodes with a node voltage less than 0.93 pu have an undervoltage problem.

[0075] Furthermore, the multi-strategy fusion algorithm described in step 5 is mainly based on the Snow Melting Optimization (SAO) algorithm and performs multi-strategy improvements. The SAO algorithm includes the following parts:

[0076] (1) SAO algorithm

[0077] The SAO algorithm mainly includes the initialization phase, exploration phase, development phase and dual population mechanism. Figure 2 A schematic diagram of the SAO algorithm is given.

[0078] 1) Initialization phase

[0079] In SAO, the iterative process starts with a randomly generated population, represented by the following matrix:

[0080] ;

[0081] In the formula, N represents the population size, Dim represents the dimension of the solution space, L and U represent the lower bound and upper bound of the solution space respectively. Represents a randomly generated number in [0,1].

[0082] In the online voltage regulation problem described in this application, what needs to be controlled is the active power and reactive power output of each photovoltaic inverter, so the solution space dimension in the above formula is defined as 2N pv , where N p Represents the number of controllable PV inverters in the system.

[0083] 2) Exploration phase

[0084] In the exploration phase, the SAO algorithm uses the randomness and irregularity of Brownian motion to describe the exploration of water vapor in space through irregular motion after snow or liquid water forms water vapor through sublimation or evaporation.

[0085] For standard Brownian motion, the SAO algorithm uses the probability density function of a normal distribution with a mean of 0 and a variance of 1 to obtain the step length of the individual movement. The mathematical formula of Brownian motion is as follows:

[0086] ;

[0087] The position update formula of the detection process in the SAO algorithm is as follows:

[0088] ;

[0089] Where Z i (t) represents the i-th individual in the t-th iteration, RB i (t) is the Brownian motion random number vector; represents a random number in [0,1]; G(t) represents the optimal solution of the current population; Z i (t) represents the current position of the individual; It is an individual randomly selected from several elite groups in the population; is the center of mass position of the entire group in this update, and its mathematical expression is as follows:

[0090] ;

[0091] Where, and They are the second best individual and the third best individual in the current population respectively. In the SAO algorithm, the 50% individuals with the highest fitness in the population are called elite individuals. represents the centroid positions of these individuals.

[0092] ;

[0093] In the formula, N1 represents the number of elite individuals, which is numerically equal to half of N. It is a set consisting of the average of the best individual, the second best individual, the third best individual, and the elite individual in the group. During each position update during the exploration process, a random value is drawn from this set to assist in the position update.

[0094] 3) Development phase

[0095] In the SAO algorithm, when snow is converted into liquid water through melting, individuals are encouraged to develop high-quality solutions around the current best individual rather than expanding highly dispersed features in the solution space. The degree-day method is used to represent the snowmelt process:

[0096] ;

[0097] Where M is the snowmelt rate, DDF is the degree-day factor, t is the number of iterations, and t max is the maximum number of iterations. T(t) is the temperature at time t.

[0098] During the SAO development phase, the individual position update equation is as follows:

[0099] ;

[0100] Where, Represents a random number in [-1,1].

[0101] 4) Dual population mechanism

[0102] The SAO algorithm introduces an innovative dual population mechanism. In the early stages of the algorithm iteration, the entire population is randomly divided into two sub-populations P of equal size. a and P b , where P a Responsible for exploration, P b Responsible for development, the population size is N a and N b As the iterative process progresses, P b The scale of P gradually decreases, and a This dynamic adjustment mechanism enables the algorithm to focus on local development in the early stages while gradually strengthening the exploration of the global solution space.

[0103] (2) Multi-strategy improvement

[0104] 1) Cold condensation strategy

[0105] The original SAO algorithm has the disadvantage of insufficient late-stage development capabilities. Inspired by the condensation phenomenon, this application proposes a condensation strategy. Considering that sublimation, melting, and evaporation are endothermic processes, the ambient temperature will drop, causing some water vapor to release heat and condense into water again, thereby maintaining the thermal cycle. In the Hybrid multi-strategy snow melting optimization (HMSAO) algorithm, this process is manifested as an increase in the number of subpopulations responsible for development and a decrease in the number of subpopulations responsible for exploration, thereby improving the algorithm's late-stage development capabilities and thus improving the convergence speed and solution quality. The specific formula is as follows:

[0106] ;

[0107] The expression of parameter e is:

[0108] ;

[0109] Where, is the position of the random individual responsible for exploration in the tth iteration. e increases with the number of iterations. is a random number between [-1,1].

[0110] 2) Knowledge Transfer Learning

[0111] For similar distribution network real-time voltage regulation problems, there is no need to repeatedly solve them. Transfer learning can flexibly use the knowledge learned in other tasks to solve new problems. Therefore, this application proposes a learning strategy that combines offline pre-learning with online fast solution, as shown in the attached Figure 3 shown.

[0112] First, offline learning involves clustering historical source-load data to obtain a set of typical source-load states. This is then solved using the HMSAO algorithm, ultimately yielding empirical knowledge for the corresponding source-load states. Secondly, an RPO empirical knowledge base is formed to store pre-learned convergence knowledge for different source-load states. Finally, during online solving, the current system source-load state is used to match historical states in the knowledge base. If a match is successful, the relevant knowledge is directly transferred for a faster solution. If a match fails, a reinforcement learning method without knowledge transfer is used to solve the problem, and the converged results are used as new knowledge to supplement the knowledge base, enabling dynamic updates.

[0113] The specific implementation method of knowledge transfer is as follows: when the source-load state of the new task of real-time voltage regulation of the distribution network has a certain similarity with a source task in the knowledge base, the knowledge matrix corresponding to the source task is directly transferred to the source task. As the initial knowledge matrix for online optimization tasks , thereby avoiding blind trial and error exploration to improve efficiency:

[0114] ;

[0115] Where, The similarity between the online source-load state of the real-time voltage regulation problem in the distribution network and the source-load state of the source task in the knowledge base is described. 、 are the number of loads and PV inverters, is the active power of the load and PV inverter, The threshold is set in advance. When there are multiple source tasks that meet the requirements, the most similar source task knowledge is adopted.

[0116] The following is the verification of the real-time voltage regulation method for distribution network proposed in this application:

[0117] The IEEE33 node system with PV access is selected as the simulation object to verify the effectiveness of the method proposed in this application. The system structure is shown in the attached figure. Figure 4 As shown in Figure 2, a 24-hour real-time voltage regulation simulation is performed with a 5-minute time granularity. PV panels are installed at nodes 7, 11, 18, 27, 30, and 33.

[0118] (1) Effectiveness analysis

[0119] Attachment Figure 5The voltage distribution of the entire power distribution system without voltage regulation is given. It can be seen that during the period of t=40-60, the sunlight intensity is high at noon, the photovoltaic power generation is large, and there are serious overvoltage problems at nodes 9-18 and nodes 27-33. Figure 5 From the system topology, nodes 9-18 are on the same branch, and their overvoltage is mainly caused by the power backflow caused by the large power generation of PV1, PV2 and PV3; nodes 27-33 are on the same branch, and their overvoltage is mainly caused by the power backflow caused by the large power generation of PV4, PV5 and PV6.

[0120] Based on the HMSAO optimization solution algorithm proposed in this application, the voltage distribution of each node in the system over time is as shown in the attached figure. Figure 6 It shows that the original voltage limit problem has been effectively solved, and the node voltages are all controlled within [0.93,,1.07]pu.

[0121] (2) Comparative analysis of advantages

[0122] In order to test the effectiveness of the rough knowledge transfer proposed in this application, taking the T=49 period as an example, a set of values ​​with certain similarity is generated by superimposing random disturbance signals on the basis of its source-load state, which is used as the source-load state for online simulation test. In order to compare and analyze the advantages of the proposed SAO algorithm, GA, PSO, SAO and HMSAO algorithms are used to solve the problem. Each algorithm is applied to the T=49 scenario for multiple solutions. Figure 7 The optimal iterative convergence result is given.

[0123] To further compare the performance of each algorithm, the results of 10 solutions are shown in Table 1.

[0124] Table 1 Statistics of solution results of each algorithm

[0125]

[0126] 1) From the perspective of the objective function, both PSO and HMSAO have the ability to converge to the optimal solution (1.22), but only the HMSAO algorithm proposed in this application can ensure that each iteration converges to the global optimal solution;

[0127] 2) The SAO algorithm has a fast convergence rate, but it fails to converge to the optimal solution. This is due to its insufficient late-stage exploration and development capabilities, making it prone to getting stuck in local optima. The HMSAO algorithm builds on this by incorporating a cold condensation mechanism. This process releases heat, condensing water vapor into water, maintaining the thermal cycle. This process increases the number of subpopulations responsible for development while decreasing the number of subpopulations responsible for exploration, thereby improving the algorithm's late-stage development capabilities. Therefore, the HMSAO algorithm possesses stronger optimization capabilities.

[0128] 3) In terms of convergence time, the HMSAO algorithm demonstrates significant efficiency advantages, achieving solutions at speeds up to 20 times faster than other algorithms. This performance advantage is primarily attributed to the introduction of a knowledge transfer mechanism, which enables the algorithm to obtain a relatively good initial solution in the initial stages. This allows the algorithm to obtain high-quality solutions with fewer iterations when faced with similar online tasks. Experimental results demonstrate that the HMSAO algorithm offers significant advantages in both convergence and computational efficiency.

[0129] In summary, the real-time voltage regulation method for distribution network based on multi-strategy fusion algorithm proposed in this application has good performance in actual optimization.

[0130] Based on the same inventive concept, the embodiment of the present application also provides a device for implementing the above-mentioned real-time voltage regulation of the distribution network. The solution provided by the device is similar to the solution described in the above-mentioned method. In an exemplary embodiment, as shown in the attached Figure 8 It is shown that a real-time voltage regulation device for a distribution network based on a multi-strategy fusion algorithm is provided, including:

[0131] Historical data collection module M1: collects historical operation data of the distribution system, including historical load data, grid data, historical output data of distributed power sources, etc.

[0132] Source-load data set construction module M2: clusters historical operation data to obtain a typical source-load status data set;

[0133] Knowledge base construction module M3: Based on the multi-strategy fusion algorithm, it solves typical scenarios in the typical source-load status dataset offline and builds a transfer learning knowledge base;

[0134] Safety verification module M4: performs ultra-short-term output forecasts for loads and distributed power sources, and performs system safety verification;

[0135] Voltage regulation module M5: If a voltage problem occurs, the real-time voltage regulation strategy of the distribution network is solved based on the multi-strategy fusion algorithm in combination with historical experience in the knowledge base to solve the voltage problem.

[0136] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as shown in the attached diagram. Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store power dispatch forecast data for provincial power grids. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a real-time voltage regulation method for a distribution network based on a multi-strategy fusion algorithm is implemented.

[0137] Those skilled in the art will understand that Figure 8 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0138] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0139] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

Claims

1. A real-time voltage regulation method for distribution network based on multi-strategy fusion algorithm, characterized in that: The following steps are involved: Obtain historical operation data of the power distribution system; Cluster the historical operation data to obtain a typical source-load status dataset; Based on the multi-strategy fusion algorithm, we solve typical scenarios in typical source-load status datasets offline and build a transfer learning knowledge base. Conduct ultra-short-term output forecasts for loads and distributed power sources, and perform system safety verification; If a voltage problem occurs in the prediction results, the distribution network is regulated in real time based on the solution results of the multi-strategy fusion algorithm in combination with historical experience in the knowledge base.

2. The method for real-time voltage regulation of distribution network based on multi-strategy fusion algorithm according to claim 1 is characterized in that: K-meas clustering is used to cluster the historical operation data. Each set of data in the data set consists of the net load of each node in the distribution network.

3. The method for real-time voltage regulation of distribution network based on multi-strategy fusion algorithm according to claim 1, characterized in that: The node voltage safety interval of the safety check is set to [0.93, 1.07]. In the power flow calculation results, nodes with a node voltage greater than 1.07 pu have an overvoltage problem, and nodes with a node voltage less than 0.93 pu have an undervoltage problem.

4. The method for real-time voltage regulation of distribution network based on multi-strategy fusion algorithm according to claim 1, characterized in that: The multi-strategy fusion algorithm includes a snow melting algorithm and a multi-strategy improvement of the snow melting algorithm.

5. The method for real-time voltage regulation of distribution network based on multi-strategy fusion algorithm according to claim 1, characterized in that: The snow melting algorithm includes an iterative process starting with a randomly generated population, represented by the following matrix: ； In the formula, N represents the population size, Dim represents the dimension of the solution space, L and U represent the lower bound and upper bound of the solution space respectively. Represents a randomly generated number in [0,1]; The solution space dimension is defined as 2N pv , where N p Represents the number of controllable PV inverters in the distribution network; The randomness and irregularity of Brownian motion are used to describe the irregular movement of water vapor in space after snow or liquid water forms water vapor through sublimation or evaporation. For standard Brownian motion, the step length of individual movement is obtained by using the probability density function of a normal distribution with a mean of 0 and a variance of 1; the mathematical formula of Brownian motion is as follows: ； The position update formula of the detection process in the snow melting algorithm is as follows: ； Where Z i (t) represents the i-th individual in the t-th iteration, RB i (t) is the Brownian motion random number vector; represents a random number in [0,1]; G(t) represents the optimal solution of the current population; Z i (t) represents the current position of the individual; It is an individual randomly selected from several elite groups in the population; is the center of mass position of the entire group in this update, and its mathematical expression is as follows: ； Where, and are the second best individual and the third best individual in the current population respectively, and the 50% individuals with the highest fitness in the population are called elite individuals. represents the centroid positions of these individuals; ； Where N1 represents the number of elite individuals, which is equal to half of N in value; It is a set consisting of the average of the best individual, the second best individual, the third best individual, and the elite individual in the group; during each position update during the exploration process, a random value is drawn from this set to assist in the position update; When snow is converted into liquid water through melting, individuals are encouraged to develop high-quality solutions around the current best individual. The degree-day method is used to represent the snowmelt process: ； The individual position update equation is as follows: Where, Represents a random number in [-1,1]; In the initial iteration of the snow melting algorithm, the entire population is randomly divided into two sub-populations P of equal size. a and P b , where P a Responsible for exploration, P b Responsible for development, the population size is N a and N b As the iterative process progresses, P b The scale of P gradually decreases, and a The scale increases accordingly.

6. The method for real-time voltage regulation of distribution network based on multi-strategy fusion algorithm according to claim 1, characterized in that: Multi-strategy improvements to the snow ablation algorithm include: ; The expression of parameter e is: Where, is the position of the random individual responsible for exploration in the tth iteration, e increases with the number of iterations, Is a random number between [-1,1]; For similar real-time voltage regulation problems in distribution networks, there is no need to repeatedly solve them. Transfer learning uses knowledge learned from other tasks to solve new problems. The offline learning is based on historical source-load data, clustering to obtain a typical source-load state set, and solving it based on the HMSAO algorithm, and finally obtaining the experience knowledge of the corresponding source-load state; forming an RPO experience knowledge base to store the knowledge of pre-learning convergence under different source-load states; finally, when solving online, the historical state in the knowledge base is matched based on the current system source-load state; after a successful match, the relevant knowledge is directly transferred to quickly solve; if the match fails, the reinforcement learning method without knowledge transfer is used to solve, and the convergence result is used as new knowledge to supplement the knowledge base to realize the dynamic update of the knowledge base; the specific implementation method of knowledge transfer is as follows: when the source-load state of the new task of real-time voltage regulation of the distribution network has a certain similarity with a source task in the knowledge base, the knowledge matrix corresponding to the source task is directly transferred. As the initial knowledge matrix for online optimization tasks : ； Where, The similarity between the online source-load state of the real-time voltage regulation problem in the distribution network and the source-load state of the source task in the knowledge base is described. 、 are the number of loads and PV inverters, is the active power of the load and PV inverter, The threshold is set in advance. When there are multiple source tasks that meet the requirements, the most similar source task knowledge is adopted.

7. A real-time voltage regulation device for distribution network based on multi-strategy fusion algorithm, characterized in that: include: Historical data collection module: used to obtain historical operation data of the power distribution system; Source-load data set construction module: used to cluster historical operation data to obtain a typical source-load status data set; Knowledge base construction module: used to solve typical scenarios in typical source-load status datasets offline based on a multi-strategy fusion algorithm and build a transfer learning knowledge base; Safety verification module: used to make ultra-short-term output forecasts of loads and distributed power sources, and perform system safety verification; Voltage regulation module: When voltage problems occur in the prediction results, it combines historical experience in the knowledge base and performs real-time voltage regulation on the distribution network based on the solution results of the multi-strategy fusion algorithm.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the real-time voltage regulation method for a distribution network based on a multi-strategy fusion algorithm according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the real-time voltage regulation method for a distribution network based on a multi-strategy fusion algorithm according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the real-time voltage regulation method for a distribution network based on a multi-strategy fusion algorithm according to any one of claims 1 to 6 is implemented.

Images