Distribution Estimation Algorithm-Based Optimization Method and System for Distribution Network Protection Point Layout

Abstract

The present invention discloses an optimized algorithm for distribution network protection point layout based on the estimation of distribution algorithm. According to the topological structure of the distribution network and the layout of relay protection points, the initial population of the evolutionary population is generated. Then, the fitness of the initial population is evaluated, that is, the fitness of each individual is calculated according to the objective function, namely, the protection investment, maintenance cost, and power outage loss cost of the protection point layout scheme are calculated; and the individuals are screened with the power supply reliability as the constraint condition. Then, a new evolutionary population is generated according to the new probability model for the next iteration until the end condition is met. The present invention aims at the optimal economy, including protection investment and maintenance cost, taking into account the power outage loss cost caused by the formation of PV and energy storage islands, with power supply reliability as the constraint, selecting the system average power supply availability as the power supply reliability index, considering the influence of distributed power source access, and optimizing the distribution network protection point layout.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems and their automation, and relates to a method and system for optimizing the layout of distribution network protection points based on the estimation of distribution algorithm. Background Art

[0002] With the substantial increase in users' electricity consumption demands, users have higher and higher requirements for the reliability of power supply. In the power system, most power grid faults are caused by the distribution network, resulting in immeasurable economic and social losses due to user power outages, and also significantly reducing the satisfaction of power users with power supply enterprises. As the terminal network of the power system, distributed energy sources such as photovoltaic and energy storage devices are mostly connected to the distribution network, changing the topology of the distribution network and forming a multi-source operation mode. When a fault occurs, the protection acts reliably, cutting off the faulty branch, causing some branches to lose power. However, new energy sources such as photovoltaic and energy storage connected to the distribution network can continuously supply power to surrounding loads, forming an island power supply area and reducing the original power outage losses. Therefore, it is extremely urgent to optimize the layout of protection devices considering new energy sources such as photovoltaic and energy storage, which can not only improve the observability and power supply reliability of the distribution network, reduce the blind investment in power facilities and fault power outage losses, but also weaken the impact of high-proportion new energy access on relay protection to a certain extent.

[0003] Generally speaking, the more the number of relay protections, the higher the power supply reliability of the distribution network. Many domestic and foreign studies optimize the layout of protection points for the power supply reliability of the distribution network to meet the reliability requirements. However, with the increase in the number of relay protections, the investment cost will also increase accordingly. Therefore, on the premise of meeting the power supply reliability requirements of the distribution network, it is necessary to determine the optimal configuration quantity and optimal configuration location of relay protections in the distribution network to minimize the power outage loss and the full life cycle cost of relay protection configuration. Traditional optimization algorithms such as genetic algorithms optimize from the individual level, resulting in poor performance when facing high-dimensional problems.

[0004] A method and system for optimizing the protection configuration of a distribution network disclosed in Publication No. CN 112487710 A includes: obtaining the opening and closing states of each circuit breaker in the outgoing line unit and the protection configuration of each circuit breaker according to the real-time operation mode, and randomly generating an initial population of the outgoing line unit based on the power flow distribution; calculating the fitness function of each population and recording the optimal chromosome; selecting individuals for crossover and mutation to form a new generation of population; calculating the fitness function of the new population and recording the final individuals; determining whether the maximum number of iterations is reached. If so, end the loop and output the optimal individuals. Otherwise, perform the crossover and mutation operations again. The invention combines the real-time power flow to identify the main line and branch lines, improves the reliability of the setting values of the distribution network, reduces the power outage range of the distribution network, and ensures the safe and reliable operation of the distribution network. However, the optimization process of this method is based on individual optimization, performing selection, crossover, and mutation operations on individuals. For complex systems, the convergence is poor, the calculation amount is large, and the solution is difficult. Moreover, the rationality of the point layout is not considered from an economic perspective, resulting in uneconomical point layout. The present invention optimizes the population from a macroscopic perspective using a statistical probability model, with a small calculation amount and good convergence. With the goal of optimal economy, considering factors such as distributed power supply energy and fault recovery time on the premise of meeting the requirements of power supply reliability, the optimal protection point layout result of the distribution network is obtained. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to reasonably and efficiently configure the position and quantity of relay protection, and to make the economy of power supply enterprises optimal on the basis of meeting the power supply reliability requirements of the distribution network.

[0006] The present invention realizes the solution of the above technical problems through the following technical means:

[0007] A method for optimizing the protection point layout of a distribution network based on the estimation of distribution algorithm, characterized by including the following steps:

[0008] S1. Generate an initial population. According to the topological structure of the distribution network and the layout of relay protection, generate the initial population of the evolving population. Among them, the topological structure of the distribution network determines the length of the individuals in the initial population, and the layout of relay protection determines the binary coding of the individuals in the initial population. The number of individuals in the population is determined according to the length of the individuals.

[0009] S2. Evaluate the fitness of the population. Calculate the fitness of each individual according to the objective function, that is, calculate the protection investment, maintenance cost, and power outage loss cost of the protection point layout scheme.

[0010] S3. Selection operation. First, according to the power supply reliability constraint conditions, preliminarily screen the individuals in the population, select the individuals that meet the power supply reliability requirements, and then, on this basis, according to the individual fitness evaluation results, screen out the dominant individuals and store them in a new population.

[0011] S4. Construct a probability model. Based on the selected dominant population, calculate the probability model according to the statistical learning theory.

[0012] S5. Sample a new population according to the probability model. Generate a new evolutionary population according to the probability model and perform the next iteration.

[0013] S6. Determine whether the condition is satisfied after iteration. After each iteration, determine whether it meets the iteration end condition. To ensure obtaining the optimal solution, use the number of iterations as the judgment condition. If the end condition is met, the iterative evolution ends. If not, continue the iteration until the end condition is satisfied.

[0014] The basic principle of the present invention is as follows: First, according to the topological structure of the distribution network and the layout of relay protection, an initial population of the evolutionary population is generated. The topological structure of the distribution network determines the length of the individuals in the initial population, and the layout of relay protection determines the binary coding of the individuals in the initial population. Secondly, evaluate the fitness of the initial population, that is, calculate the fitness of each individual according to the objective function, and then, with the power supply reliability as the constraint condition, screen the individuals. Those that meet the reliability requirements enter the next screening, and those that do not meet the reliability requirements are eliminated, greatly improving the population evolution efficiency and accelerating the algorithm convergence rate. According to the fitness evaluation results, further screen out the dominant individuals in the initial population and store them in a new population. Based on the selected dominant population, calculate the probability model according to the statistical learning theory. And add a learning rate during evolution to further update the probability model. Then, according to the new probability model, generate a new evolutionary population and perform the next iteration. After each iteration, determine whether it meets the iteration end condition. If the end condition is met, the iterative evolution ends. If not, continue the iteration until the end condition is satisfied to obtain the optimal solution.

[0015] Further, in the step S1, the steps of generating the initial population are as follows:

[0016] S1.1. Initialize basic constants.

[0017] Each individual represents a protection layout plan. The individual length represents all possible places where protection can be installed in the distribution network, and is encoded in binary. 0 represents installing protection, and 1 represents not installing protection; thus, define the individual length, individual appearance, number of individuals, number of iterations, and learning rate.

[0018] S1.2. Randomly generate the initial population.

[0019] Define the matrix Binary_X to store the initial population. Use the rand function to generate random numbers, and use the round function to round the generated random numbers to 0 or 1, representing the installation situation of protection, so as to generate the initial population.

[0020] Furthermore, in the step S2, the steps for evaluating the fitness of the population are as follows:

[0021] S2.1. Calculate the protection investment cost;

[0022] The equivalent annual value conversion formula for the investment cost of the electromechanical protection switch is as follows:

[0023]

[0024] In the formula, M is the number of protection installation units, C S is the present value of the investment cost per unit of protection, i is the discount rate, and p is the service life of the protection equipment;

[0025] S2.2. Calculate the operation and maintenance cost;

[0026] The annual operation and maintenance cost of the electromechanical protection switch equipment is given as a percentage of its investment:

[0027] L COST = S COST ·η

[0028] In the formula, η is the proportional coefficient of the operation cost to the investment cost;

[0029] S2.3. Calculate the power outage loss;

[0030] The annual power outage loss cost C IP can be expressed as:

[0031]

[0032] In the formula, n is the total number of load points in the system, C Cj is the unit power outage loss of the jth load point, and W ENSj is the expected power supply shortage of the jth load point, and its calculation formula is as follows:

[0033]

[0034] In the formula, t ni is the expected power outage time of load point i, and P Li is the load demand of load point i;

[0035] And t ni is divided into two categories. One is the power outage time t ISi when the load is within the island division range, and the other is the power outage time t Ci when the load is not within the island division range, and its expression is as follows:

[0036] t ni = IP i t ISi+(1 - IP i )t Ci

[0037] Wherein, IP i is the probability of islanding formation,

[0038] Assume that within a certain short time period, the average output power and load of the distributed power source do not change significantly, that is, they are considered constant; then within the selected time period, the output power of the distributed power source can be equivalent to a normal distribution function with known mean and variance, and the parameters of the normal distribution function can be obtained by statistical analysis of historical data; then the probability of islanding formation can be transformed into the probability that the output power of the distributed power source is greater than the load demand within the island, which is:

[0039] IP i = IPB j,i

[0040] Wherein, IP i is the probability of islanding formation at load point i, and IPB j,i is the probability that the output power of the distributed power source is greater than the load demand within the island;

[0041] When load point i is within the island range that can be covered by the distributed power source, the calculation of its power outage time t ISi is divided into two cases: The first case is that load point i is in front of the distributed power source, and the calculation formula is as follows:

[0042]

[0043] Wherein, l i is the length of the i-th feeder, λ is the average line fault probability, t r is the average line fault repair time, t s is the switching operation time of the protection electromechanical protection switch, and x m is the state variable of the protection installed on the m-th section of the feeder;

[0044] The second case is that load point i is behind the distributed power source, and the calculation formula is as follows:

[0045]

[0046] Wherein, k is the feeder section where the distributed power source is located;

[0047] When load point i is not within the island range that can be covered by the distributed power source, the calculation formula of its power outage time t Ci is as follows:

[0048]

[0049] Furthermore, in the step S3, the steps of the selection operation are:

[0050] S3.1. Calculate power supply reliability;

[0051] The power supply reliability of the distribution network is expressed by the average power supply availability rate. The system average power supply availability index ASAI can be calculated by the following formula:

[0052]

[0053] In the formula, N i is the number of users at load point i;

[0054] S3.2. Select operations;

[0055] First, screen out the individuals that meet the power supply reliability requirements, and then select the relatively better individuals from these individuals. Generally, select the first half of the sorted individuals and store them in the new population;

[0056] Furthermore, in the step S4, the steps for constructing the probability model are as follows:

[0057] S4.1. Statistically calculate the probability model;

[0058] Based on the screened dominant population, calculate the probability model based on statistical learning theory; here, the probability model is a one-dimensional matrix with the same length as the individual length, and its calculation is as follows:

[0059] Assume that the dominant population is Q, and the calculation formula for the probability vector P[1,10] is as follows:

[0060]

[0061] In the formula, P i is the probability of the i-th column of the probability vector, q is the number of dominant population individuals in Q dominant populations, x i represents the i-th column in the binary coding of the individual. If it is 1, x i is 1. If it is 0, x i is 0;

[0062] S4.2. Update the probability model;

[0063] In the process of evolution, in addition to the innate natural selection, there is also the acquired autonomous learning that promotes evolution. The distribution algorithm takes this into account. Therefore, the learning rate needs to be added during evolution. Only the part of natural selection was calculated in the previous probability vector, so it needs to be updated here. The calculation formula is as follows:

[0064]

[0065] In the formula, P l+1 (I) is the updated probability vector, α is the learning rate, x kl The bit of 1 in the dominant individual; thus, the probability model in the first evolution is constructed;

[0066] Further, in step S5, the step of sampling a new population according to the probability model is as follows:

[0067] S5.1. Sample a new population according to the probability model;

[0068] Use the rand function to randomly generate a number between 0 and 1, and compare it with the probability vector P l+1 (I). If the random number is larger, the binary number stored in the matrix Binary_X is 0; if the random number is smaller, the stored binary number is 1, thereby updating the population matrix and generating a new population; thus, a complete iteration is completed; then it is judged whether the iteration ends until the iteration ends to obtain the optimal solution.

[0069] Corresponding to the above method, the present invention also provides a distribution estimation algorithm-based optimal distribution system for distribution network protection points, including:

[0070] An initial population generation module, configured to generate an initial population of an evolutionary population according to the topological structure of the distribution network and the distribution of relay protection points, where the topological structure of the distribution network determines the length of the individuals in the initial population, and the distribution of relay protection points determines the binary encoding of the individuals in the initial population, and the number of individuals in the population is determined according to the individual length;

[0071] A population fitness evaluation module, configured to calculate the fitness of each individual according to the objective function; that is, calculate the protection investment, maintenance cost, and power outage loss cost of the protection point scheme;

[0072] A selection operation module, configured to first preliminarily screen the population individuals according to the power supply reliability constraint conditions, select the individuals that meet the power supply reliability requirements, and then, on this basis, screen out the dominant individuals according to the individual fitness evaluation results and store them in a new population;

[0073] A probability model construction module, configured to calculate a probability model based on the statistical learning theory according to the selected dominant population;

[0074] A new population generation module, configured to generate a new evolutionary population according to the probability model for the next iteration;

[0075] A judgment module, configured to judge whether the condition is satisfied after the iteration. After each iteration is completed, judge whether it meets the iteration end condition. To ensure obtaining the optimal solution, the iteration number is used as the judgment condition. If the end condition is met, the iterative evolution ends; if not, continue the iteration until the end condition is met.

[0076] Further, in the initial population generation module, the steps for generating the initial population are as follows:

[0077] S1.1. Initialize basic constants;

[0078] Each individual represents a protection layout plan. The length of the individual represents all possible locations for installing protection in the distribution network, and it is encoded in binary. 0 represents installing protection, and 1 represents not installing protection; thus, the individual length, individual appearance, number of individuals, number of iterations, and learning rate are defined.

[0079] S1.2. Randomly generate the initial population;

[0080] Define the matrix Binary_X to store the initial population. Use the rand function to generate random numbers, and use the round function to round the generated random numbers to 0 or 1, representing the installation situation of protection, thereby generating the initial population.

[0081] Further, in the population fitness evaluation module, the steps for evaluating the population fitness are as follows:

[0082] S2.1. Calculate the protection investment cost;

[0083] The equivalent annual value conversion formula for the investment cost of the electromechanical protection switch is as follows:

[0084]

[0085] In the formula, M is the number of protection installation units, C S is the present value of the investment cost per unit of protection, i is the discount rate, and p is the service life of the protection equipment;

[0086] S2.2. Calculate the operation and maintenance cost;

[0087] The annual operation and maintenance cost of the electromechanical protection switch equipment is given as a percentage of its investment:

[0088] L COST = S COST ·η

[0089] In the formula, η is the proportional coefficient of the operation cost to the investment cost;

[0090] S2.3. Calculate the power outage loss;

[0091] The annual power outage loss cost C IP can be expressed as:

[0092]

[0093] In the formula, n is the total number of load points in the system, C Cj is the unit power outage loss of the jth load point, W ENSjis the expected unsupplied electricity of the j-th load point, and its calculation formula is as follows:

[0094]

[0095] In the formula, t ni is the expected power outage time of load point i, and P Li is the load demand of load point i;

[0096] And t ni is divided into two categories. One is the power outage time t ISi when the load is within the island division range, and the other is the power outage time t Ci when the load is not within the island division range. Their expressions are as follows:

[0097] t ni = IP i t ISi +(1 - IP i )t Ci

[0098] In the formula, IP i is the island formation probability.

[0099] Assume that within a certain short time period, the average output power and load of the distributed power source do not change much, that is, they are considered constant; then the output power of the distributed power source within the selected time period can be equivalent to a normal distribution function with known mean and variance, and the parameters of the normal distribution function can be obtained by statistical analysis of historical data; then the island formation probability can be transformed into the probability that the output power of the distributed power source is greater than the load demand within the island, which is:

[0100] IP i = IPB j,i

[0101] In the formula, IP i is the island formation probability of load point i, and IPB j,i is the probability that the output power of the distributed power source is greater than the load demand within the island;

[0102] When load point i is within the island range that can be covered by the distributed power source, the calculation of its power outage time t ISi is divided into two cases: The first case is when load point i is in front of the distributed power source, and the calculation formula is as follows:

[0103]

[0104] In the formula, l i is the length of the i-th feeder, λ is the average line fault probability, t r is the average line fault repair time, and t sTo protect the switching operation time of the electromechanical protection switch, x m is the state variable of the protection installed for the m-th section of the feeder;

[0105] The second case is when the load point i is behind the distributed power source, and the calculation formula is as follows:

[0106]

[0107] In the formula, k is the feeder section where the distributed power source is located;

[0108] When the load point i is not within the island range that can be covered by the distributed power source, its power outage time t Ci has the following calculation formula:

[0109]

[0110] Furthermore, in the selection operation module, the steps of the selection operation are as follows:

[0111] S3.1. Calculate the power supply reliability;

[0112] The power supply reliability of the distribution network is expressed by the average power supply availability rate. The system average power supply availability index ASAI can be calculated by the following formula:

[0113]

[0114] In the formula, N i is the number of users at the load point i;

[0115] S3.2. Select the operation;

[0116] First, filter out the individuals that meet the power supply reliability requirements, and then from these individuals, filter out the better individuals. Generally, select the first half of the sorted individuals and store them in the new population;

[0117] Furthermore, in the probability model construction module, the steps of constructing the probability model are as follows:

[0118] S4.1. Statistically calculate the probability model;

[0119] Based on the selected dominant population and statistical learning theory, calculate the probability model; here the probability model is a one-dimensional matrix with the same length as the individual length, and its calculation is as follows:

[0120] Assume that the dominant population is Q, and the calculation formula for the probability vector P[1,10] is as follows:

[0121]

[0122] In the formula, P iis the probability of the i-th column of the probability vector, q is the number of dominant population individuals in Q dominant populations, and x i represents the i-th column in the individual binary encoding. If it is 1, x i is 1. If it is 0, x i is 0;

[0123] S4.2. Update the probability model;

[0124] In the process of evolution, in addition to the innate natural selection, there is also the acquired autonomous learning that promotes evolution. The distribution algorithm takes this into account. Therefore, the learning rate needs to be added during evolution. Only the part of natural selection was calculated in the previous probability vector, so it needs to be updated here. The calculation formula is as follows:

[0125]

[0126] In the formula, P l+1 (I) is the updated probability vector, α is the learning rate, and x k l is the bit of 1 in the dominant individual; thus, the probability model in the first evolution is constructed;

[0127] Furthermore, in the new population generation module, the steps of sampling to generate a new population according to the probability model are as follows:

[0128] S5.1. Sample according to the probability model to generate a new population;

[0129] Use the rand function to randomly generate a number between 0 and 1, and compare it with the probability vector P l+1 (I). If the random number is large, the binary number stored in the matrix Binary_X is 0. If the random number is small, the stored binary number is 1, thereby updating the population matrix and generating a new population; thus, a complete iteration is completed; then it is judged whether the iteration ends until the iteration ends to obtain the optimal solution.

[0130] The advantages of the present invention are as follows:

[0131] This invention uses the estimation of distribution algorithm to optimize the protection layout of the distribution network. The estimation of distribution algorithm, with its binary encoding method, well adapts to the protection installation state and performs optimization iteration from a macroscopic perspective. Even with a high dimension, it can converge well. In this way, for a complex distribution system, the estimation of distribution algorithm can still effectively solve the problem of optimizing the protection layout.

[0132] The invention uses the system average power supply reliability rate as the power supply reliability index. In the calculation formula of the average power supply reliability rate, the annual expected power outage time of the system is used to replace the traditional annual average power outage time of users, taking into account the influence of different protection layout schemes on the power outage time. At the same time, the influence of distributed power source access on the power outage time is also considered, preventing the wrong selection or arrangement of advantageous schemes during the screening process, improving the iteration efficiency, ensuring that the finally obtained result is the optimal protection layout scheme, and thus providing effective suggestions for the protection layout planning of the distribution network.

[0133] The invention aims at the optimal economy, including protection investment, maintenance cost, and power outage loss cost caused by the formation of photovoltaic and energy storage islands. Constrained by power supply reliability, the system average power supply availability rate is selected as the power supply reliability index, considering the influence of distributed power source access, and optimizing the protection layout of the distribution network. Brief Description of the Drawings

[0134] Figure 1 It is the flowchart of the distribution network protection layout optimization algorithm based on the estimation of distribution algorithm in the embodiment of the invention;

[0135] Figure 2 It is the example given in the embodiment of the invention: the IEEE33-node diagram of the distribution network;

[0136] Figure 3 is Figure 2 the protection layout result diagram of the example in;

[0137] Figure 4 is Figure 2 the iterative simulation diagram of the example in. Detailed Embodiment

[0138] To make the objectives, technical solutions, and advantages of the embodiments of the invention clearer, the technical solutions in the embodiments of the invention will be clearly and completely described below in conjunction with the embodiments of the invention. Obviously, the described embodiments are part of the embodiments of the invention, rather than all of the embodiments. Based on the embodiments of the invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the invention.

[0139] The distribution network protection layout optimization algorithm based on the estimation of distribution algorithm is characterized by including the following steps:

[0140] S1. Generate an initial population. According to the topological structure of the distribution network and the layout of relay protection, generate the initial population of the evolutionary population, where the topological structure of the distribution network determines the length of the individuals in the initial population, the layout of relay protection determines the binary coding of the individuals in the initial population, and the number of individuals in the population is determined according to the length of the individuals.

[0141] S2. Evaluate the fitness of the population. Calculate the fitness of each individual according to the objective function, that is, calculate the economy of each protection point - setting scheme.

[0142] S3. Selection operation. First, according to the power supply reliability constraint conditions, preliminarily screen the individuals in the population, select the individuals that meet the power supply reliability requirements, and then, on this basis, according to the individual fitness evaluation results, screen out the dominant individuals and store them in a new population.

[0143] S4. Construct a probability model. According to the selected dominant population, calculate the probability model based on the statistical learning theory. At the same time, add the part of acquired autonomous learning to further update the probability model.

[0144] S5. Sample a new population according to the probability model. Generate a new evolutionary population according to the new probability model and perform the next iteration.

[0145] S6. Judge whether the condition is met after iteration. After each iteration, judge whether it meets the iteration end condition. To ensure obtaining the optimal solution, use the number of iterations as the judgment condition. If the end condition is met, the iterative evolution ends; if not, continue the iteration until the end condition is met.

[0146] Another technical solution of the present invention is that in the step S1, the steps of generating the initial population are as follows:

[0147] S1.1. Initialize basic constants.

[0148] Each individual represents a protection point - setting scheme. The length of the individual represents all possible places where protection can be installed in the distribution network, and it is encoded in binary. 0 represents installing protection, and 1 represents not installing protection. Thus, define the length of the individual, the appearance of the individual, the number of individuals, the number of iterations, and the learning rate.

[0149] S1.2. Randomly generate the initial population.

[0150] Define the matrix Binary_X to store the initial population. Use the rand function to generate random numbers, and use the round function to round the generated random numbers to 0 or 1, representing the installation situation of the protection, so as to generate the initial population.

[0151] Another technical solution of the present invention is that in the step S2, the steps of evaluating the fitness of the population are as follows:

[0152] S2.1. Calculate the protection investment cost.

[0153] The individual fitness refers to the value of the objective function corresponding to the protection point - setting scheme, that is, calculate the protection investment, maintenance cost, and power outage loss cost of the protection point - setting scheme.

[0154] The costs encountered in the economic analysis of optimizing the protection layout can be divided into two categories: one is the one-time payment costs, such as the investment cost of purchasing equipment; the other is the annual payment costs, such as operation and maintenance costs, equipment depreciation costs, capital recovery, loan interest, dividends and bonuses, various taxes, etc. Considering factors such as taxes, interest, and currency devaluation caused by inflation, these two types of costs must be converted before they can be compared. Moreover, for the same type of cost, even if the amounts are equal when occurring in different years, their actual values are different and cannot be simply calculated and compared.

[0155] The investment cost of the electromechanical protection switch belongs to the one-time payment cost. Considering the different service lives of various electromechanical protection switch devices, the present value of the investment in the electromechanical protection switch can be converted into the equivalent annual value by the equivalent annual value method. This can not only avoid the influence brought by the difference in the service life of various electromechanical protection switch devices, but also show the annual income of different schemes. The equivalent annual value conversion formula for the investment cost of the electromechanical protection switch is as follows:

[0156]

[0157] In the formula, M is the number of protection installation units, C S is the present value of the investment cost per unit of protection, i is the discount rate, and p is the service life of the protection equipment.

[0158] S2.2. Calculate the operation and maintenance costs.

[0159] The annual operation and maintenance costs of the electromechanical protection switch equipment are given as a percentage of its investment:

[0160] L COST = S COST ·η

[0161] In the formula, η is the proportional coefficient of the operation cost to the investment cost.

[0162] S2.3. Calculate the power outage loss.

[0163] The calculation of the user's power outage loss is relatively complex, and factors such as the time of power outage, the lack of power supply, the duration of power outage, the frequency of power outage, and the user type should be considered. Among them, the duration of the user's power outage and the user type have the greatest impact on the power outage loss. This includes the impact of line faults, self-faults of the electromechanical protection switch, and transformer faults on the duration of the load power outage.

[0164] Generally, the annual power outage loss cost C IP of the system can be expressed as:

[0165]

[0166] In the formula, n is the total number of load points in the system, C CjThe unit outage loss of the jth load point, W ENSj The expected unsupplied electricity of the jth load point, and its calculation formula is as follows:

[0167]

[0168] In the formula, t ni is the expected outage time of load point i, P Li is the load demand of load point i.

[0169] And t ni is divided into two categories. One is the outage time t ISi when the load is within the island division range, and the other is the outage time t Ci when the load is not within the island division range. Its expression is as follows:

[0170] t ni = IP i t ISi +(1 - IP i )t Ci

[0171] In the formula, IP i is the island formation probability. With the access of distributed power sources, its impact on outage losses cannot be ignored. After a distribution network fault, the distributed power source can supply power to the load alone, reducing its outage time and improving power supply reliability. Therefore, when calculating the outage loss cost, it is necessary to consider the island formation probability to account for the impact of distributed power sources such as photovoltaic and energy storage.

[0172] Assume that within a relatively short time period, the average output of the distributed power source and the load do not change much, that is, they are considered constant. Then the output power of the distributed power source within the selected time period can be equivalent to a normal distribution function with known mean and variance, and the normal distribution parameters can be obtained by statistical analysis of historical data. Then the island formation probability can be transformed into the probability that the output power of the distributed power source is greater than the load demand within the island, which is:

[0173] IP i = IPB j,i

[0174] In the formula, IP i is the island formation probability of load point i, and IPB j,i is the probability that the output power of the distributed power source is greater than the load demand within the island.

[0175] When load point i is within the island range that can be covered by the distributed power source, the calculation of its outage time t ISi is divided into two cases: The first is when load point i is in front of the distributed power source, and the calculation formula is as follows:

[0176]

[0177] Wherein, l i is the length of the i-th feeder, λ is the average line fault probability, and t r is the average repair time of line faults, and t s is the switching operation time of the protection electromechanical protection switch, and x m is the state variable of the protection installed on the m-th section of the feeder.

[0178] The second case is when the load point i is behind the distributed power source, and the calculation formula is as follows:

[0179]

[0180] Wherein, k is the feeder section where the distributed power source is located.

[0181] When the load point i is not within the island range that can be covered by the distributed power source, its power outage time t Ci is calculated as follows:

[0182]

[0183] Another technical solution of the present invention is that in the step S3, the steps of the selection operation are:

[0184] S3.1. Calculate the power supply reliability.

[0185] The power supply reliability of the distribution network is represented by the average power supply availability here, which is one of the common indicators of distribution reliability. The system average power supply availability index ASAI (Average Service Availability Index) refers to the ratio of the total power outage-free time obtained by users in a year to the total power supply time required by users. Then the system average power supply availability index ASAI can be calculated according to the following formula:

[0186]

[0187] Wherein, N i is the number of users at the load point i.

[0188] S3.2. Select the operation.

[0189] First, screen out the individuals that meet the power supply reliability requirements, and then from these individuals, screen out the better individuals. Generally, select the first half of the sorted individuals and store them in the new population.

[0190] Another technical solution of the present invention is that in the step S4, the steps of constructing the probability model are:

[0191] S4.1. Statistically calculate the probability model.

[0192] Based on the selected dominant population, a probability model is calculated based on statistical learning theory. Here, the probability model is a one-dimensional matrix with the same length as the individual length, and its calculation is as follows:

[0193] Assume that the dominant population is as shown in the following table:

[0194] Table 1

[0195] Individual 1 1001101010 Individual 2 0101011101 Individual 3 1011100010 Individual 4 0111010101

[0196] The calculation formula for the probability vector P[1,10] is as follows:

[0197]

[0198] In the formula, P i is the probability of the i-th column of the probability vector, q is the number of dominant individuals, and x i represents the i-th column in the binary encoding of the individual. If it is 1, x i is 1, and if it is 0, x i is 0.

[0199] S4.2. Update the probability model.

[0200] In the process of evolution, in addition to the innate natural selection, there is also the acquired autonomous learning that promotes evolution. The distributed algorithm takes this into account. Therefore, when evolving, the learning rate also needs to be added. Only the part of natural selection was calculated in the previous probability vector, so it needs to be updated here. The calculation formula is as follows:

[0201]

[0202] In the formula, P l+1 (I) is the updated probability vector, α is the learning rate, m is the number of individuals in the dominant population, and x k l is the bit of 1 in the dominant individual. Thus, the probability model in the first evolution is constructed.

[0203] Another technical solution of the present invention is that in step S5, the step of sampling to generate a new population according to the probability model is as follows:

[0204] S5.1. Sample to generate a new population according to the probability model.

[0205] Use the rand function to randomly generate a number between 0 and 1, and the probability vector P l+1(I) Compare. If the random number is large, the binary number stored in matrix Binary_X is 0. If the random number is small, the stored binary number is 1, thereby updating the population matrix to generate a new population. Thus, a complete iteration is completed. Then, it is judged whether the iteration ends until the iteration ends to obtain the optimal solution.

[0206] Based on Figure 2 Optimize the distribution points according to the IEEE 33-node diagram of the medium-voltage distribution network, including the topological structure of distributed energy sources such as photovoltaic and energy storage. The number of protection devices that need to be configured in the network is determined by the engineering budget. From the perspectives of network redundancy and power outage losses, it is hoped that the more protection devices are configured, the better. However, due to project cost limitations, there should be an upper limit to the number of protection devices.

[0207] First, according to the topological structure of the distribution network and the distribution points of relay protection, randomly generate the initial population of the evolutionary population. Determine the length of the individuals in the initial population according to the topological structure of the distribution network, and determine the binary encoding of the individuals in the initial population according to the distribution points of relay protection, forming different-length binary-encoded individuals defined by using 0 to represent the installation of protection and 1 to represent the non-installation of protection. Calculate the fitness of each individual in the initial population according to the set objective function, and calculate the economy of each protection distribution plan;

[0208] Secondly, define matrix Binary_X to store the initial population. Use the rand function to generate random numbers, and the round function rounds the generated random numbers to 0 or 1, representing the installation situation of protection, thereby generating the initial population. Initially screen out the individuals that meet the power supply reliability requirements from the initial population, and then according to the individual fitness evaluation results, screen out the dominant individuals, and then screen out the dominant individuals that meet the constraint conditions and have high fitness, and store them in the new population. Based on statistical learning theory, calculate the probability model, and at the same time add the part of acquired autonomous learning to further update the probability model;

[0209] Finally, sample to generate a new population according to the probability model for the next iteration. Use the rand function to randomly generate a number between 0 and 1, and compare it with the probability vector Pl+1(I). Through continuous iterative calculation based on the estimation of distribution algorithm, finally select the dominant individual with the highest fitness, that is, the optimal relay protection distribution plan.

[0210] The main purpose of the present invention is to rationally and efficiently configure the position and quantity of relay protection, and on the basis of meeting the power supply reliability requirements of the distribution network, optimize the economy of power supply enterprises, and adopt a distribution network protection point optimization algorithm based on the estimation of distribution algorithm.

[0211] Its basic principle is as follows: First, according to the topological structure of the distribution network and the layout of relay protection, the initial population of the evolutionary population is generated. Among them, the topological structure of the distribution network determines the length of the individuals in the initial population, and the layout of relay protection determines the binary coding of the individuals in the initial population. Secondly, the fitness of the initial population is evaluated, that is, the fitness of each individual is calculated according to the objective function, and then, with the power supply reliability as the constraint condition, the individuals are screened. Those that meet the reliability requirements enter the next screening, and those that do not meet the reliability requirements are excluded, which greatly improves the population evolution efficiency and speeds up the algorithm convergence rate. And according to the fitness evaluation results, the dominant individuals in the initial population are further screened out and stored in a new population.

[0212] The estimation of distribution algorithm is a global search algorithm. Different from traditional evolutionary algorithms that perform iterative optimization at the individual level, the estimation of distribution algorithm uses a probability model to describe the distribution of solutions in space. By selecting better solutions to update the probability vector describing the solutions, new samples are generated until the optimal solution or the set number of iterations is reached. It conducts optimization calculations from a macroscopic perspective to find the global optimal solution. Therefore, next, according to the selected dominant population, based on statistical learning theory, the probability model is calculated. In the process of population evolution, in addition to the innate natural selection, there is also the acquired autonomous learning that promotes its evolution. Therefore, the estimation of distribution algorithm takes this into account and adds a learning rate during evolution to further update the probability model. Then, according to the new probability model, a new evolutionary population is generated for the next iteration. The evolutionary process is a long process, and it may take multiple generations of evolution to obtain the optimal solution. After each iteration is completed, it is judged whether it meets the iteration end condition. Here, to ensure obtaining the optimal solution, the number of iterations is used as the judgment condition. If the end condition is met, the iterative evolution ends; if not, the iteration continues until the end condition is met.

[0213] Figure 2 For this embodiment, a distribution network IEEE 33-node graph is given. Taking this graph as an example, the above method is described as follows: According to Figure 2 the distribution network IEEE 33-node graph in Figure 2 to optimize the layout. The topological graph contains distributed energy sources such as photovoltaic and energy storage. Photovoltaic power generation systems are installed at the positions of nodes 10, 15, 18, 23, and 29, which can supply power to the load nodes in its island when a fault occurs in the system, reducing its power outage time. Energy storage systems are installed at nodes 3 and 8 to store the excess electric energy of the photovoltaic power generation system and supply power to the load nodes within its power supply range when a fault occurs in the system, which can also reduce its power outage time. Therefore, the position, quantity, and capacity of photovoltaic and energy storage directly affect the power supply reliability of the system and the calculation of power outage losses. When using the estimation of distribution algorithm to optimize the protection layout, their influence must be considered.

[0214] Since the number of protection devices required in the network is determined by the project budget, and from the perspective of network redundancy and power outage losses, the more protection devices are configured, the better. However, due to the limitation of project cost, there should be an upper limit on the number of protection devices.

[0215] First, according to the topological structure of the distribution network and the layout of relay protection, the initial population of the evolutionary group is randomly generated. The length of the individuals in the initial population is determined by the topological structure of the distribution network. Figure 2 The IEEE33 node topology has 33 nodes and 32 lines, so the individual length is 32. The binary codes of individuals in the initial population are determined by the distribution of relay protection. The rand function is used to generate random numbers. The round function rounds the generated random numbers to 0 or 1, forming binary coded individuals of different lengths defined by 0 representing installed protection and 1 representing not installed protection, thereby generating the initial population.

[0216] Secondly, define the matrix Binary_X to store the initial population, and use the objective function to evaluate the fitness of the individuals in the population, that is, calculate the economic efficiency of each scheme. The objective function includes protection investment costs, protection maintenance costs, and system power outage loss costs. The first two items are related to the number of protection installations, while the power outage loss costs are not only related to the number of protection installations, but also to the protection installation location, node load size, line length, photovoltaic and energy storage, etc. Among them, system parameters such as node load size and line length are known conditions, so the power outage loss mainly considers the protection point distribution and the impact of photovoltaic and energy storage. Since the generated initial population has given the protection point information, the protection investment cost and protection maintenance cost can be calculated. The power outage loss cost is affected by photovoltaics and energy storage, resulting in large fluctuations in node power outage time. It is necessary to use the breadth-first search algorithm to divide photovoltaics and energy storage into islands. For nodes within the island range, when a system failure occurs, if there is protection between the node and the fault point, the node can be isolated from the fault point, and photovoltaics or energy storage continue to supply power to it. The power outage time is the protection action time; if there is no protection between the node and the fault point, although it is within the island range, it cannot be isolated from the fault, and the power outage time is the fault repair time. For nodes outside the island range, when the system fails, if it is before the fault point and there is protection between it and the fault point, its power outage time is the protection action time, otherwise, the power outage time is the fault repair time. At this point, the power outage loss cost can also be calculated, and the individual fitness evaluation is completed.

[0217] Then, individuals that meet the power supply reliability requirements are initially screened from the initial population, and the dominant individuals are screened out based on the individual fitness evaluation results and stored in a new population. Based on statistical learning theory, the probability model is calculated, and the acquired autonomous learning part is added to further update the probability model.

[0218] Finally, a new population is sampled according to the probability model for the next iteration. The rand function is used to randomly generate a number between 0 and 1, which is compared with the probability vector Pl+1(I). Through continuous iterative calculation based on the estimation of distribution algorithm, the dominant individual with the highest fitness will be finally selected, that is, the optimal relay protection placement scheme.

[0219] Refer to Figure 1 the method flow chart, program on the Matlab platform, input Figure 2 the line data of each node in, set the iteration end condition to 1000 iterations, and perform the solution. Finally, the optimal protection placement result is obtained as Figure 3 shown. Protections are arranged at lines 1, 2, 7, 13, and 21 (shown as rectangular nodes in the figure). This protection placement result is the optimal individual selected during the evolution process according to the continuously updated probability model by continuously selecting the dominant population, that is, the most economical scheme, and it meets the power supply reliability requirements. The iteration process is as Figure 4 shown. Figure 4 In the figure, the abscissa is the number of iterations, and the ordinate is the optimal solution. From the final iteration results, it can be seen that using the estimation of distribution algorithm to optimize the relay protection placement of the distribution network can effectively solve the optimal solution, and the iteration process has good convergence and short iteration time, verifying the effectiveness of the optimization algorithm, which can provide effective suggestions for the relay protection placement planning of the distribution network.

[0220] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. The optimization method for the distribution point of distribution network protection based on the estimation of distribution algorithm is characterized in that It includes the following steps: S1. Generate the initial population. According to the topological structure of the distribution network and the layout of relay protection, generate the initial population of the evolutionary population. Among them, the topological structure of the distribution network determines the length of the individuals in the initial population, and the layout of relay protection determines the binary encoding of the individuals in the initial population. The number of individuals in the population is determined according to the length of the individuals; S2. Evaluate the fitness of the population. Calculate the fitness of each individual according to the objective function, that is, calculate the protection investment, maintenance cost, and power outage loss cost of the protection layout scheme; S3. Selection operation. First, according to the power supply reliability constraint conditions, preliminarily screen the individuals in the population, select the individuals that meet the power supply reliability requirements, and then, on this basis, according to the evaluation results of the individual fitness, screen out the dominant individuals and store them in a new population; S4. Construct a probability model. According to the selected dominant population, calculate the probability model based on the statistical learning theory; S5. Generate a new population by sampling according to the probability model. Generate a new evolutionary population according to the probability model and perform the next iteration; S6. Judge whether the condition is met after iteration. After each iteration is completed, judge whether it meets the iteration end condition. In order to ensure obtaining the optimal solution, use the number of iterations as the judgment condition. If the end condition is met, the iterative evolution ends. If not, continue the iteration until the end condition is met; In the step S2, the steps for evaluating the fitness of the population are: S2.

1. Calculate the protection investment cost; The equivalent annual value conversion formula for the investment cost of the electromechanical protection switch is as follows: Where M is the number of protection installation platforms, and C S is the present value of the investment cost for a single protection unit, i is the discount rate, and p is the service life of the protection equipment; S2.

2. Calculate the operation and maintenance cost; The annual operation and maintenance cost of the electromechanical protection switch equipment is given as a percentage of its investment: L COST = S COST · η In the formula, η is the proportional coefficient of the operation cost to the investment cost; S2.

3. Calculate the power outage loss; Annual power outage loss cost C IP Can be expressed as: where n is the total number of load points in the system, and C Cj is the unit outage loss of the j-th load point, and W ENSj is the expected power shortage of the j-th load point, and its calculation formula is as follows: where t ni is the expected outage time of load point i, and P Li is the load demand of load point i.

2. The distribution network protection point layout optimization method based on the estimation of distribution algorithm according to claim 1, characterized in that In the step S1, the steps for generating the initial population are: S1.

1. Initialize the basic constants; Each individual represents a protection layout scheme. The individual length represents all possible places where protection can be installed in the distribution network and is encoded in binary. 0 represents the installation of protection, and 1 represents the non-installation of protection; thus, define the individual length, individual appearance, number of individuals, number of iterations, and learning rate; S1.

2. Randomly generate the initial population; Define the matrix Binary_X to store the initial population. Use the rand function to generate random numbers, and use the round function to round the generated random numbers to 0 or 1, representing the installation situation of the protection, so as to generate the initial population.

3. The method for optimizing the distribution of protection points in a distribution network based on the estimation of distribution algorithm according to claim 1, characterized in that In the step S2, t ni is divided into two categories. One is the power outage time t ISi when the load is within the island division range, and the other is the power outage time t Ci when the load is not within the island division range. Their expressions are as follows: t ni = IP i t ISi +(1 - IP i )t Ci where IP i is the probability of island formation Assume that within a relatively short time period, the average output and load of the distributed power source do not change much, that is, it is considered constant; then the output power of the distributed power source within the selected time period can be equivalent to a normal distribution function with known mean and variance, and the parameters of the normal distribution function can be obtained by statistical analysis of historical data; then the probability of island formation can be transformed into the probability that the output power of the distributed power source is greater than the load demand within the island, which is: IP i = IPB j,i Where, IP i is the islanding formation probability of load point i, and IPB j,i is the probability that the output power of distributed generation is greater than the load demand within the island; When the load point i is within the island area that can be covered by distributed power sources, its power outage time t ISi is calculated in two cases: The first case is that the load point i is in front of the distributed power source, and the calculation formula is as follows: where, l i is the length of the i-th feeder, λ is the average line fault probability, t r is the average line fault repair time, t s is the switching operation time of the protection electromechanical protection switch, x m is the state variable of the protection installed on the m-th section of the feeder; The second case is that the load point i is behind the distributed power source, and the calculation formula is as follows: In the formula, k is the feeder section where the distributed power source is located; When the load point i is not within the islanding range that can be covered by distributed power sources, its power outage time t Ci is calculated as follows:

4. The method for optimizing the distribution point of the distribution network protection based on the estimation of distribution algorithm according to claim 1, characterized in that, In the step S3, the steps for the selection operation are: S3.

1. Calculate the power supply reliability; The reliability of distribution network power supply is expressed by the average power supply availability rate. The system average power supply availability index ASAI can be calculated by the following formula: where N i is the number of users at load point i; S3.

2. Selection operation; First, select the individuals that meet the power supply reliability requirements, and then from these individuals, select the better individuals. Select the first half of the individuals in the ranking and store them in the new population.

5. The method for optimizing the distribution point of the distribution network protection based on the estimation of distribution algorithm according to claim 1, characterized in that In the step S4, the steps for constructing the probability model are as follows: S4.

1. Statistical calculation of the probability model; Based on the selected dominant population, calculate the probability model based on statistical learning theory; here the probability model is a one-dimensional matrix with the same length as the individual length, and its calculation is as follows: Assume that the number of the dominant population is Q, and the calculation formula of the probability vector P[1,10] is as follows: Wherein, P i is the probability of the i-th column of the probability vector, q is the number of individuals in the dominant population among the Q dominant populations, and x i represents the i-th column in the binary encoding of the individual. If it is 1, x i is 1. If it is 0, x i is 0; S4.

2. Update the probability model; In the process of evolution, in addition to the innate natural selection, there is also the acquired autonomous learning that promotes evolution. The distributed algorithm takes this into account, so the learning rate should be added during evolution. Only the part of natural selection is calculated in the previous probability vector, so it needs to be updated here. The calculation formula is as follows: Where P l+1 (I) is the updated probability vector, α is the learning rate, and x k l is the bit of 1 in the dominant individual; thus, the probability model in the first evolution is constructed.

6. The method for optimizing the distribution of protection points in a distribution network based on the estimation of distribution algorithm according to claim 1, wherein In the step S5, the steps for sampling to generate a new population according to the probability model are as follows: S5.

1. Sample to generate a new population according to the probability model; Use the rand function to randomly generate a number between 0 and 1, and compare it with the probability vector P l+1 (I) If the random number is larger, the binary number stored in the matrix Binary_X is 0; if the random number is smaller, the stored binary number is 1, thereby updating the population matrix to generate a new population; So far, a complete iteration is over; Then judge whether the iteration ends until the iteration ends to obtain the optimal solution.

7. The optimal placement system for distribution network protection based on the estimation of distribution algorithm is characterized in that Including: An initial population generation module, which is used to generate the initial population of the evolutionary population according to the topological structure of the distribution network and the layout of relay protection. The topological structure of the distribution network determines the length of the individuals in the initial population, and the layout of relay protection determines the binary coding of the individuals in the initial population. The number of individuals in the population is determined according to the individual length; An evaluation of population fitness module, which is used to calculate the fitness of each individual according to the objective function, that is, calculate the protection investment, maintenance cost, and power outage loss cost of the protection layout plan; A selection operation module, which is used to first preliminarily screen the population individuals according to the power supply reliability constraint conditions, select the individuals that meet the power supply reliability requirements, and then on this basis, screen out the dominant individuals according to the individual fitness evaluation results and store them in the new population; A probability model construction module, which is used to calculate the probability model based on the selected dominant population and statistical learning theory; A new population generation module, which is used to generate a new evolutionary population according to the probability model for the next iteration; A judgment module, which is used to judge whether the conditions are met after iteration. After each iteration is completed, judge whether it meets the iteration end condition. In order to ensure obtaining the optimal solution, the number of iterations is used as the judgment condition. If the end condition is met, the iterative evolution ends. If not, continue the iteration until the end condition is met; In the evaluation of population fitness module, the steps for evaluating the population fitness are as follows: S2.

1. Calculate the protection investment cost; The equivalent annual value conversion formula for the investment cost of the electromechanical protection switch is as follows: where M is the number of protection installation platforms, C S is the present value of the investment cost for a single protection unit, i is the discount rate, and p is the service life of the protection equipment; S2.

2. Calculate the operation and maintenance cost; The annual operation and maintenance cost of the electromechanical protection switch equipment is given as a percentage of its investment: L COST = S COST · η In the formula, η is the proportional coefficient of the operation cost to the investment cost; S2.

3. Calculate the power outage loss; The annual power outage loss cost C IP can be expressed as: where n is the total number of load points in the system, and C Cj is the unit power outage loss of the j-th load point, and W ENSj is the expected power supply shortage of the j-th load point, and its calculation formula is as follows: where t ni is the expected outage time of load point i, and P Li is the load demand of load point i.

8. The distribution network protection point layout optimization system based on the estimation of distribution algorithm according to claim 7, characterized in that In the module for generating the initial population, the steps for generating the initial population are as follows: S1.

1. Initialize basic constants; Each individual represents a protection layout scheme. The length of an individual represents all possible locations for installing protection in the distribution network, and it is encoded in binary. 0 represents installing protection, and 1 represents not installing protection. Thus, the length of an individual, the appearance of an individual, the number of individuals, the number of iterations, and the learning rate are defined; S1.

2. Randomly generate the initial population; Define the matrix Binary_X to store the initial population. Use the rand function to generate random numbers, and use the round function to round the generated random numbers to 0 or 1, representing the installation situation of protection, thereby generating the initial population.

9. The distribution network protection point layout optimization system based on the estimation of distribution algorithm according to claim 7, characterized in that, In the step S2, t ni is divided into two categories. One is the power outage time t ISi when the load is within the island division range, and the other is the power outage time t Ci when the load is not within the island division range. Their expressions are as follows: t ni = IP i t ISi +(1 - IP i )t Ci where IP i is the probability of island formation, Assume that within a relatively short period of time, the average output power and load of the distributed power source do not change significantly, that is, it is considered constant; then the output power of the distributed power source within the selected time period can be equivalent to a normal distribution function with known mean and variance, and the parameters of the normal distribution function can be obtained by statistical analysis of historical data; then the probability of island formation can be transformed into the probability that the output power of the distributed power source is greater than the load demand within the island, which is: IP i = IPB j,i where, IP i is the islanding formation probability of load point i, and IPB j,i is the probability that the output power of distributed generation is greater than the load demand within the island; When the load point i is within the island range that can be covered by distributed power sources, its power outage time t ISi is calculated in two cases: The first case is that the load point i is in front of the distributed power source, and the calculation formula is as follows: where l i is the length of the i-th feeder, λ is the average line fault probability, t r is the average line fault repair time, t s is the switching operation time of the protection electromechanical protection switch, x m is the state variable of the protection installed on the m-th section of the feeder; The second case is that the load point i is behind the distributed power source, and the calculation formula is as follows: In the formula, k is the feeder segment where the distributed power source is located; When the load point i is not within the islanding range that can be covered by distributed power sources, its power outage time t Ci is calculated as follows:

10. The distribution network protection point layout optimization system based on the estimation of distribution algorithm according to claim 7, characterized in that, In the module for selection operation, the steps for selection operation are as follows: S3.

1. Calculate the power supply reliability; The power supply reliability of the distribution network is represented by the average power supply availability rate. The system average power supply availability index ASAI can be calculated according to the following formula: where N i is the number of users at load point i; S3.

2. Perform the selection operation; First, screen out the individuals that meet the power supply reliability requirements, and then select the relatively excellent individuals from these individuals. Select the first half of the sorted individuals and store them in the new population.

11. The distribution network protection point layout optimization system based on the estimation of distribution algorithm according to claim 7, characterized in that In the module for constructing the probability model, the steps for constructing the probability model are as follows: S4.

1. Statistically calculate the probability model; Based on the selected dominant population, calculate the probability model based on the statistical learning theory; here the probability model is a one-dimensional matrix with the same length as the individual length, and its calculation is as follows: Assume that the number of the dominant population is Q, and the calculation formula for the probability vector P[1,10] is as follows: where P i is the probability of the i-th column of the probability vector, q is the number of individuals in the dominant population among Q dominant populations, and x i represents the i-th column in the binary encoding of the individual. If it is 1, x i is 1. If it is 0, x i is 0; S4.

2. Update the probability model; In the process of evolution, in addition to the innate natural selection, there is also the acquired autonomous learning that promotes evolution. The distributed algorithm takes this into account, so the learning rate should be added during evolution. Only the part of natural selection was calculated in the previous probability vector, so it needs to be updated here, and the calculation formula is as follows: where P l+1 (I) is the updated probability vector, α is the learning rate, and x k l is the bit of 1 in the dominant individual; thus, the probability model in the first evolution is constructed.

12. The optimized system for distribution network protection point placement based on the estimation of distribution algorithm according to claim 7, characterized in that, In the module for generating the new population, the steps for sampling to generate the new population based on the probability model are as follows: S5.

1. Sample to generate the new population based on the probability model; Use the rand function to randomly generate a number between 0 and 1, and compare it with the probability vector P l+1 (I) If the random number is larger, the binary number stored in the matrix Binary_X is 0; if the random number is smaller, the stored binary number is 1, thereby updating the population matrix and generating a new population; So far, a complete iteration is considered to be over; Then it is judged whether the iteration is over, and until the iteration is over, the optimal solution is obtained.

Images