A distributed power generation planning method based on improved whale algorithm

Abstract

The present invention relates to a distributed power planning method based on an improved whale algorithm, including: using the LHS method to construct scenarios for wind, light and load to obtain an initial scenario, and reducing the initial scenario through a K-means clustering algorithm to obtain a typical scenario; establishing a distributed power planning model with the minimum improved voltage stability index and the minimum annual comprehensive cost as the goal; determining the nodes to be planned for the distributed power according to the numerical value sorting of the improved voltage stability index; using the improved whale algorithm to solve the distributed power planning model to obtain the planning results corresponding to the nodes to be planned. Compared with the prior art, the present invention improves the disadvantages of the original whale algorithm that it has a slow convergence speed and is easy to fall into a local optimal solution. Its Pareto frontier surface can improve the diversity and uniformity of the solution, and the solution is closer to the Pareto optimal frontier, thereby accelerating the solution speed and avoiding the solution from falling into the local optimum, so that the planning result can ensure the voltage stability after grid connection.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed power supply planning, and in particular to a distributed power supply planning method based on an improved whale algorithm. Background Art

[0002] With the advent of the electric age, people are increasingly demanding high-quality electricity. At present, the power systems of most countries in the world are powered by large power grids that supply power to loads over long distances. This power supply method has many disadvantages: due to the long transmission distance of power supply, it is easy to cause low voltage quality at the end of the line; local accidents in the interconnection of large power grids may cause the entire large power grid to fail or even paralyze; at the same time, most traditional large power grids use fossil energy for energy supply, which not only consumes energy quickly but also easily produces a large amount of harmful gases, which does not meet the current environmental protection policies. Therefore, improving the power supply mode, promoting energy reform, and building a low-carbon, environmentally friendly, efficient and safe distribution network power supply system are the new directions for the development of the current power grid.

[0003] At present, distributed generation (DG) has rapidly become a research hotspot in the field of power systems due to its advantages of various types, small investment, flexible application methods, clean and environmentally friendly. The International Council on Large Electric Systems (CIGRE) defines distributed generation (DG) as a small, decentralized, environmentally compatible, small independent power generation system with a power range of several kilowatts to tens of megawatts, which is arranged near the user and supplies power. It is generally connected to the distribution network and owned by the power department, power users or third-party investors. It can meet the specific needs of the power department and load users, such as avoiding peak load peaks, or supplying power to users in remote areas such as mountainous areas, saving transmission and transformation investment, and improving the safety and reliability of power supply in the power system. It can be independent of the traditional public distribution network and directly set up near the user to provide electricity for a small number of users with specific requirements, and can also be connected to the distribution system to supply power to the load together with the public power grid.

[0004] Distributed generation is an effective supplement to the traditional centralized power supply mode. Its mutual complementation and coordination with the traditional power grid is a better solution to provide users with safe, stable and reliable electricity with existing resources, technology and settings. The basic characteristics of the future power grid are environmental friendliness and sustainability, which requires the large-scale access and application of DG using renewable energy as the main energy source. DG planning is an extremely important preliminary work in the development planning of the power system. Since the random access of DG to the distribution network will have a wide range of impacts, mainly manifested in aggravating the deterioration of power quality, affecting the reliability of network power supply, increasing the complexity of relay protection strategies, improving the short-circuit capacity of the distribution network, and changing the voltage level of the distribution network, when planning one or more DGs to access the distribution network, it is necessary to reasonably select the site and size of DG. However, the current distributed power sources mainly adopt wind power generation and photovoltaic power generation with uncertain and random characteristics, which makes the solution process of the traditional planning algorithm more complicated, the solution speed is too slow, and it is easy to fall into the local optimum, which makes it difficult to deal with the impact of wind, light and load uncertainty and the impact of distributed power sources on voltage stability after grid connection. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a distributed power supply planning method based on the improved whale algorithm to speed up the solution and avoid the solution falling into local optimality, so that the planning result can ensure the voltage stability after grid connection.

[0006] The purpose of the present invention can be achieved by the following technical solution: A distributed power planning method based on an improved whale algorithm comprises the following steps:

[0007] S1. Obtain network topology and source-load information, use LHS (Latin hypercube sampling) method to construct scenarios for wind, light and load, obtain initial scenarios, and use K-means clustering algorithm to reduce the initial scenarios to obtain typical scenarios;

[0008] S2. Based on typical scenarios, a distributed power planning model is established with the goal of minimizing the improved voltage stability index and the annual comprehensive cost;

[0009] S3, determining the nodes to be planned for the distributed power source according to the numerical order of the improved voltage stability index;

[0010] S4. Use the improved whale algorithm to solve the distributed power planning model and obtain the planning results corresponding to the nodes to be planned.

[0011] Furthermore, the step S1 specifically includes the following steps:

[0012] S11. Probabilistic modeling is performed for wind power, photovoltaic power and load respectively, where the wind power probability model is represented by a two-parameter Weibull distribution, the photovoltaic probability model is represented by a Beta distribution, and the load probability model is represented by a normal distribution;

[0013] S12. Each random variable is sampled by the LHS method to generate a large number of initial scenarios, and then the initial scenarios are clustered by the improved K-means clustering algorithm to reduce the typical scenarios.

[0014] Furthermore, the wind power probability model in step S11 is specifically:

[0015] The two-parameter Weibull distribution is used to describe the wind speed, and its probability density function is:

[0016]

[0017] Among them, v is the actual wind speed, k and c are shape parameters and scale parameters respectively;

[0018] The actual output of the fan P w The relationship between it and wind speed v is expressed as:

[0019]

[0020] Among them, P wr is the rated power of the fan, v ci 、v co and v r They are the fan's cut-in, cut-out and rated wind speed respectively.

[0021] Furthermore, the photovoltaic probability model in step S11 is specifically:

[0022] Beta distribution is used to describe solar radiation, and its probability density function is:

[0023]

[0024] Among them, α and β are the two shape parameters of Beta distribution, I and I r are the actual value and maximum value of solar radiation respectively;

[0025] The actual output of photovoltaic power t The relationship with solar radiation I is as follows:

[0026]

[0027] Among them, P tr is the photovoltaic rated power.

[0028] Furthermore, the load probability model in step S11 is specifically:

[0029] The load size is represented by normal distribution, and its probability density function is:

[0030]

[0031]

[0032] Among them, P G is the active load, μ and σ are its expected value and standard deviation respectively, Q G is the reactive load; is the load power factor angle.

[0033] Furthermore, the specific process of clustering the initial scene by using the improved K-means clustering algorithm in step S12 is:

[0034] 1. The scene set to be clustered is:

[0035] X={x i ∈X p ,|i=1,2,...,n}

[0036] Among them, x i is the i-th initial scene, n is the total number of initial scenes in the scene set;

[0037] 2. Set the number of clusters to K and select the variance S(x c )The first K scenes with the smallest number are taken as the initial cluster centers, where c = 1, 2, ..., n, and the variance S(x c ) is calculated as follows:

[0038]

[0039] Where d is the Euclidean distance, is the mean of all scenes to be clustered in the scene set;

[0040] 3. Calculate the Euclidean distance between the remaining scenes and each cluster center, and classify them into the cluster with the closest cluster center, and re-obtain the cluster center of each cluster;

[0041] 4. Delete the scenes except the cluster center, and add the scene probability to the scene as the cluster center, so as to obtain K typical scenes and corresponding scene probabilities.

[0042] Furthermore, the distributed power planning model established in step S2 includes an objective function and constraints, and the objective function is specifically:

[0043] f=min[f1,f2]

[0044] Among them, f1 is the annual comprehensive cost, f2 is the improved voltage stability index;

[0045] The constraints include power balance constraints, branch capacity constraints, node voltage constraints and DG penetration rate constraints.

[0046] Furthermore, the annual comprehensive cost includes DG investment cost, operation and maintenance cost and network loss cost:

[0047]

[0048] Among them, P(s) is the probability of scene s, C I is the investment cost of DG, C OM is the DG operation and maintenance cost, C L Active power loss cost of distribution network;

[0049] The improved voltage stability index is specifically:

[0050]

[0051] L=max(L ij ),i,j=1,2,...,N

[0052] L ij =ω j H ij

[0053] Among them, L is the largest improved voltage stability index in the system, L ij is the improved voltage stability index of the branch between node i and node j, H ij is the original voltage stability index of the branch between node i and node j, ω j is the load level coefficient, which is used to indicate the importance of load point j. ij The larger the value, the worse the system stability. ij The smaller it is, the more stable the system is.

[0054] Furthermore, the step S3 specifically includes the following steps:

[0055] S31, arranging the improved voltage stability index corresponding to each branch in descending order of value;

[0056] S32, selecting the top Z improved voltage stability index values, and using the end nodes of the corresponding branches as the nodes to be planned for the distributed power generation.

[0057] Furthermore, the solution process of the improved whale algorithm in step S4 is specifically as follows:

[0058] First, the values ​​of each individual objective function are calculated, and then the TOPSIS method based on information entropy is used to select the solution with the highest comprehensive satisfaction;

[0059] In the first n iterations, the Nelder-Mead simplex method is used to generate better solutions on the Pareto optimal frontier to replace the original solutions to establish a moving strong Pareto frontier and accelerate convergence;

[0060] After n iterations, PAES is used to randomly select N solutions from the archive for local search every p iterations to find a better solution and add it to the archive to increase the coverage of the solution;

[0061] Then, in order to prevent falling into the local optimum and increase population diversity, the whale will follow a certain probability O R Search in the direction opposite to the chosen best solution until the opposite situation becomes worse.

[0062] Compared with the prior art, the present invention aims at the shortcomings of the existing whale algorithm in solving complex planning problems. By introducing a logarithmic weight distance control factor and the Nelder-Mead method, the convergence speed is accelerated, the Pareto archive evolution strategy is integrated to improve the diversity of the population, and the reverse learning strategy is applied in the search to prevent the algorithm from falling into the local optimum. This can speed up the solution speed and avoid the solution from falling into the local optimum, so that the planning result can ensure the voltage stability after grid connection.

[0063] The present invention adopts LHS technology to deal with the uncertainty of wind, light and load in the site selection and sizing planning of distributed power sources, and targets the defect of randomly selecting the initial clustering center by the K-means algorithm, and clusters the first K samples with the smallest variance as the initial clustering center scenes, which can effectively ensure the reliability of the generation of typical scenes.

[0064] The present invention improves the original voltage stability index and introduces a load level coefficient, so that branches with important load levels and prone to instability can be screened out, and the improved voltage stability index is sorted from high to low, and branches with larger improved voltage stability index values ​​are screened out, and then the nodes at the ends of the branches are selected as the nodes to be planned for distributed power sources, thereby reducing the calculation scale of distributed power source planning and further improving the accuracy of planning node selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0066] Figure 2 IEEE33 node configuration in the embodiment;

[0067] Figure 3 This is the scene clustering result in the embodiment;

[0068] Figure 4 This is a Pareto front comparison chart obtained by solving the embodiment. DETAILED DESCRIPTION

[0069] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0070] Example

[0071] like Figure 1 As shown, a distributed power planning method based on an improved whale algorithm includes the following steps:

[0072] (1) Latin hypercube sampling and improved scenario clustering method are used to obtain typical scenarios to solve the uncertainty of source load. In the clustering process, the first K samples with the smallest variance are set as the initial cluster centers, and the elbow method is used to select the number of clusters of the K-means algorithm, avoiding the subjectivity of K value selection;

[0073] (2) The traditional voltage stability index is improved based on the load level, and the objective function is constructed in combination with the annual comprehensive cost to establish a multi-objective planning model for distributed generation;

[0074] (3) The site selection process uses improved voltage stability indicators to determine the scope of distributed generation access;

[0075] (4) The improved whale algorithm is used to simulate the Pareto frontier, and the TOPSIS method based on information entropy is used to select the solution with the highest comprehensive satisfaction.

[0076] This embodiment applies the above method, and the specific process is as follows:

[0077] First, construct scenarios for wind, light and load, and then use the K-means clustering algorithm to reduce the initial scenarios to obtain typical scenarios.

[0078] 1.1. Scene Construction

[0079] (1) Wind power probability model

[0080] The two-parameter Weibull distribution is used to describe the wind speed, and its probability density function is:

[0081]

[0082] Where: v is the actual wind speed; k and c are the shape parameter and scale parameter respectively.

[0083] The actual output of the fan P w The relationship between the wind speed v can be approximately expressed as:

[0084]

[0085] Where: P wr Indicates the rated power of the fan; v ci 、v co and v r They are the fan's cut-in, cut-out and rated wind speed respectively.

[0086] (2) Photovoltaic probability model

[0087] Beta distribution is used to describe solar radiation, and its probability density function is:

[0088]

[0089] Where: α and β are the two shape parameters of the Beta distribution; I and I r are the actual value and maximum value of solar radiation respectively.

[0090] The actual output of photovoltaic power t The relationship with solar radiation I is as follows:

[0091]

[0092] Where: P tr is the photovoltaic rated power.

[0093] (3) Load probability model

[0094] The size of the load is usually expressed by normal distribution. Its probability density function is:

[0095]

[0096]

[0097] Where: P G is the active load; μ and σ are its expected value and standard deviation respectively; Q G is the reactive load; is the load power factor angle.

[0098] 1.2. Scene Generation

[0099] The present invention uses LHS technology to deal with the uncertainty of wind, light and load in the site selection and capacity planning of distributed power generation. Assume that m random variables are sampled, the sampling scale is N, y t =f t (x t ) represents the tth random variable x t The cumulative probability density function of , where t = 1, 2, ..., m. The specific sampling steps are as follows:

[0100] Step 1: Divide the sampling interval [0,1] into N equal parts, then the probability of any interval is equal to 1 / N;

[0101] Step 2: Randomly select sample values ​​in each interval (so that the correlation of random variables tends to be minimum). The cumulative probability density of interval i is

[0102]

[0103] n=1,2,……,N

[0104] Where r is a random number between 0 and 1.

[0105] Step 3: Use the inverse cumulative distribution transformation to obtain the corresponding sampling values. Each row represents the sampling value of a random variable. When sampling N times, the sampling matrix is ​​a T×N matrix.

[0106]

[0107] 1.3 Scene Clustering

[0108] After generating a large number of initial scenes through the LHS technique, in order to avoid the huge amount of calculation, the samples need to be clustered. The traditional K-means algorithm is subjective in selecting the number of clusters. This paper first uses the elbow method to determine the number of clusters K. Then, in view of the defect of the K-means algorithm in randomly selecting the initial cluster center, the first K samples with the smallest variance are clustered as the initial cluster center scenes, avoiding the subjectivity of the K value selection. The specific steps are as follows:

[0109] Step 1: Assume that the scene set to be clustered is:

[0110] X={x i ∈X p ,|i=1,2,...,n}

[0111] In the formula: the number of clusters is K, the selection variance S(x c )The first K smallest scenes are taken as the initial clustering centers, where c = 1, 2, …, n.

[0112]

[0113] Where: d is the Euclidean distance, is the mean of all scenes in the scene set.

[0114] Step 2: Calculate the Euclidean distance between the remaining scenes and each cluster center and classify them into the cluster with the closest cluster center, and then re-obtain the cluster center of each cluster.

[0115] Step 3: Then delete the scenes except the cluster center and add the scene probability to the scene as the cluster center, thus obtaining K typical scenes and corresponding scene probabilities.

[0116] 2. Improve the traditional voltage stability indicators based on load levels, and then establish a multi-objective planning model for distributed power sources based on annual comprehensive costs.

[0117] 1.1 Objective Function

[0118] With the goal of minimizing the annual comprehensive cost f1 and the voltage stability index f2, the comprehensive target f of DG planning is established.

[0119] f=min[f1,f2]

[0120] A. Minimum annual comprehensive cost

[0121] f1 includes DG investment costs, operation and maintenance costs, and network loss costs, and the specific forms are as follows:

[0122]

[0123] Where: P(s) is the probability of scene s.

[0124] For the sake of convenience, the scene number labeling is omitted in the subsequent formulas.

[0125] (1) DG investment cost C I

[0126]

[0127] Where: S(r0,n)=r0(1+r0) n / ((1+r0) n -1) is the cash conversion coefficient; r0 is the rate of return on investment; n x is the planning period; k represents the type of DG; N DG Indicates the set of nodes on which DG can be installed; represents the investment cost per unit capacity of the kth DG; P DGkj represents the kth DG capacity installed on node j.

[0128] (2) DG operation and maintenance costs C OM

[0129]

[0130] Where: is the operation and maintenance cost required for the k-th DG unit power generation; T DGkj is the annual power generation duration of the kth DG at node j.

[0131] (3) Active power loss cost of distribution network C L

[0132]

[0133] Where: m is the total number of distribution network branches; C e is the unit electricity price; T max is the annual maximum load loss duration; P j is the network active power loss of branch j under maximum load;

[0134] B. Voltage stability index is minimum

[0135] Original voltage stability index H ij for:

[0136]

[0137] H ij It represents the original voltage stability index of the branch between node i and node j. As the importance of load increases, the voltage stability of the branch where the load is located needs to be guaranteed more. Therefore, the load level coefficient ω is multiplied on the basis of the original voltage stability index. j Get the new voltage stability index L ij , the higher the level, the more important the load ω j The larger the L ij for:

[0138] L ij =ω j H ij

[0139] L ij The larger the value, the more unstable the branch, and vice versa. The overall voltage stability of the distribution system depends on the maximum value of the voltage stability index in all branches of the system, and this is used as the voltage stability index of the system as a whole. If L is the maximum voltage index value in the system, the mathematical expression of the objective function f2 is:

[0140]

[0141] L=max(L ij ),i,j=1,2,...,N

[0142] L is the maximum voltage index in the system. The voltage stability of this branch is the worst and is most susceptible to disturbances. Therefore, the difference between L and 1 can reflect the voltage stability margin of the entire system.

[0143] 1.2 Constraints

[0144] (1) Power balance constraints

[0145]

[0146] Where: P i and Q i Active and reactive power injection for node i; U i is the voltage amplitude of node i; δ is the phase angle difference; G ij , B ij is the branch admittance.

[0147] (2) Branch capacity constraints

[0148] S ij ≤S ijmax

[0149] Where: Sijmax is the upper limit of the transmission capacity of branch ij.

[0150] (3) Node voltage constraints

[0151] U imin ≤U i ≤U imax

[0152] Where: U imax and U imin are the upper and lower limits of the node i voltage respectively.

[0153] (4) DG penetration rate constraint

[0154]

[0155] Where: P DGi and P DGimax They represent the DG capacity installed at node i and the upper limit of allowed installation respectively; μ is the penetration rate; Ω represents the set of nodes where DG is allowed to be installed; P Ltotal Indicates the total active load of the distribution network.

[0156] 3. Select the access scope of distributed power generation according to voltage stability indicators

[0157] In order to reduce the calculation scale of distributed power generation planning, the voltage stability index is sorted from high to low, and L ij Then select the node at the end of the branch as the node to be planned for distributed generation.

[0158] 4. Solve the model using the improved whale algorithm

[0159] This technical solution aims at the shortcomings of the existing whale algorithm (WOA) in solving complex planning problems. It introduces the logarithmic weight distance control factor and the Nelder-Mead method to speed up the convergence speed, integrates the Pareto archive evolution strategy to improve the diversity of the population, and applies the reverse learning strategy in the search to prevent the algorithm from falling into the local optimum. This technical solution uses the improved whale algorithm simulation to obtain the Pareto frontier. The specific steps are as follows:

[0160] (1) Calculate the values ​​of each individual objective function and use the TOPSIS method based on information entropy to select the solution with the highest overall satisfaction.

[0161] (2) In the first n iterations, the Nelder-Mead simplex method is used to generate better solutions on the Pareto optimal front to replace the original solutions in order to establish a moving strong Pareto front and accelerate convergence.

[0162] (3) After n iterations, PAES is used to randomly select N solutions from the archive for local search every p iterations to find a better solution and add it to the archive to increase the coverage of the solution.

[0163] (4) In order to avoid falling into the local optimum and increase population diversity, the whales will follow a certain probability O R Search in the direction opposite to the chosen best solution until the opposite situation becomes worse.

[0164] This embodiment takes the IEEE 33-node standard system as an example for verification and analysis. The distribution network system structure is as follows: Figure 2 The node load levels and their weight coefficients are shown in Table 1.

[0165] Table 1 Node load levels

[0166]

[0167] The DG power factor is 0.9 and the maximum penetration rate is 30%. The planning period is set to 20 years and the discount rate is 0.1. The maximum installed capacity of DG for each candidate node is 1,000MVA. The wind turbine parameters are: vci=3m / s, vr=10m / s and vco=20m / s; the wind speed obeys the Weibull distribution with k=2.17 and c=8.34. The photovoltaic parameters are: Ir=1kW / m2; the parameters of the beta distribution are: α=1.95 and β=2.28. The load of each node refers to the original data of IEEE33 nodes as the mean, and the standard deviation is 10% of the mean. The investment and operation and maintenance costs of WT are 10,000 yuan / kW and 4,000 yuan / kW·h respectively; the investment and operation and maintenance costs of PV are 13,000 yuan / kW and 2,500 yuan / kW·h respectively. The unit power supply cost of the substation is 0.5 yuan / kW·h.

[0168] First, the scenarios of wind, light and load are constructed, and then the initial scenarios are reduced using the K-means clustering algorithm to obtain typical scenarios.

[0169] The LHS method is used to sample wind, light and load, with a sampling scale of 800. Then, the wind speed and solar irradiance are converted into PV and WT output efficiency according to the probability model formula. 800 basic scenarios can be obtained by combining the node load data. Then, through scenario clustering, Figure 3 The number of clusters available is 8, and the typical scenarios are shown in Table 2.

[0170] Table 2 Results of scene reduction

[0171] Scenario Fan efficiency Photovoltaic efficiency Load factor Scenario Probability 1 0.3142 0.8523 0.7452 0.1268 2 0.0215 0.8864 0.4876 0.1099 3 0.9689 0.8167 0.5103 0.1396 4 0.7234 0.5621 0.5977 0.0988 5 0.6579 0.4873 0.8601 0.1395 6 0.0016 0.3486 0.7989 0.1629 7 0.5193 0.1529 0.7672 0.1228 8 0.9779 0.1985 0.7381 0.0997

[0172] Afterwards, the traditional voltage stability index is improved based on the load level, and in the site selection process, the improved voltage stability index is used to determine the scope of distributed power access.

[0173] In order to reduce the calculation scale of distributed power generation planning, the present invention sorts the voltage stability index from high to low and selects L ij Then select the node at the end of the branch as the node to be planned for distributed generation. When the load level is not considered, L ij The order from largest to smallest is: 5, 2, 27, 28, 3, 4, 23, 8, 12; after considering the load level, L ij The order from largest to smallest is: 2, 5, 3, 4, 23, 30, 28, 9.

[0174] It can be seen that after multiplying the load level weight coefficient, the branch ranking changes significantly, and the stability of important loads directly affects people's production and life and even life safety. Therefore, the importance of loads is an important factor that cannot be ignored in distribution network planning. In summary, the distributed power supply is optimized within the terminal node range of the above branches.

[0175] Finally, different algorithms are used to simulate the Pareto frontier. Figure 4 As shown. This embodiment uses NSGA-II, WOA, and MIWOALS to solve the planning model. The initial population number and total number of iterations of the three algorithms are set to 100. Finally, the Pareto frontier comparison is obtained. Figure 4 As shown, it is shown that the Pareto frontier obtained by the MIWOALS algorithm proposed in this technical solution is closer to the front, and the distribution and uniformity of the solution are significantly better than those of NSGA-II and WOA.

Claims

1. A distributed power planning method based on an improved whale algorithm, characterized in that: The following steps are involved: S1. Obtain network topology and source-load information, use LHS Latin hypercube sampling method to construct scenarios for wind, light and load, obtain initial scenarios, and use K-means clustering algorithm to reduce the initial scenarios to obtain typical scenarios; S2. Based on typical scenarios, a distributed power planning model is established with the goal of minimizing the improved voltage stability index and the annual comprehensive cost; S3, determining the nodes to be planned for the distributed power source according to the numerical order of the improved voltage stability index; S4. Use the improved whale algorithm to solve the distributed power planning model and obtain the planning results corresponding to the nodes to be planned; The solution process of the improved whale algorithm in step S4 is specifically as follows: First, the values ​​of each individual objective function are calculated, and then the TOPSIS method based on information entropy is used to select the solution with the highest comprehensive satisfaction; In the first n iterations, the Nelder-Mead simplex method is used to generate better solutions on the Pareto optimal frontier to replace the original solutions to establish a moving strong Pareto frontier and accelerate convergence; After n iterations, PAES is used to randomly select N solutions from the archive for local search every p iterations to find a better solution and add it to the archive to increase the coverage of the solution; Then, in order to prevent falling into the local optimum and increase population diversity, the whale will follow a certain probability O R Search in the direction opposite to the chosen best solution until the opposite situation becomes worse.

2. According to claim 1, a distributed power planning method based on an improved whale algorithm is characterized in that: The step S1 specifically includes the following steps: S11. Probabilistic modeling is performed for wind power, photovoltaic power and load respectively, where the wind power probability model is represented by a two-parameter Weibull distribution, the photovoltaic probability model is represented by a Beta distribution, and the load probability model is represented by a normal distribution; S12. Each random variable is sampled by the LHS method to generate a large number of initial scenarios, and then the initial scenarios are clustered by the improved K-means clustering algorithm to reduce the typical scenarios.

3. A distributed power planning method based on improved whale algorithm according to claim 2, characterized in that: The wind power probability model in step S11 is specifically: The two-parameter Weibull distribution is used to describe the wind speed, and its probability density function is: Among them, v is the actual wind speed, k and c are shape parameters and scale parameters respectively; The actual output of the fan P w The relationship between it and wind speed v is expressed as: Among them, P wr is the rated power of the fan, v ci 、v co and v r They are the fan's cut-in, cut-out and rated wind speed respectively.

4. A distributed power planning method based on an improved whale algorithm according to claim 2, characterized in that: The photovoltaic probability model in step S11 is specifically: Beta distribution is used to describe solar radiation, and its probability density function is: Among them, α and β are the two shape parameters of Beta distribution, I and I r are the actual value and maximum value of solar radiation respectively; The actual output of photovoltaic power t The relationship with solar radiation I is as follows: Among them, P tr is the photovoltaic rated power.

5. A distributed power planning method based on improved whale algorithm according to claim 2, characterized in that: The load probability model in step S11 is specifically: The load size is represented by normal distribution, and its probability density function is: Among them, P G is the active load, μ and σ are its expected value and standard deviation respectively, Q G is the reactive load; is the load power factor angle.

6. A distributed power planning method based on improved whale algorithm according to claim 2, characterized in that: The specific process of clustering the initial scene by using the improved K-means clustering algorithm in step S12 is as follows: First, determine the scene set to be clustered: X={x i ∈X p ,|i=1,2,…,n} Among them, x i is the i-th initial scene, n is the total number of initial scenes in the scene set; Then set the number of clusters to K and select the variance S(x c )The first K scenes with the smallest number are taken as the initial cluster centers, where c = 1, 2, ..., n, and the variance S(x c ) is calculated as follows: Where d is the Euclidean distance, is the mean of all scenes to be clustered in the scene set; Then calculate the Euclidean distance between the remaining scenes and each cluster center, and classify them into the cluster with the closest cluster center, and re-obtain the cluster center of each cluster; Finally, the scenes except the cluster center are deleted, and the scene probability is added to the scene as the cluster center, so as to obtain K typical scenes and corresponding scene probabilities.

7. A distributed power planning method based on improved whale algorithm according to claim 1, characterized in that: The distributed power planning model established in step S2 includes an objective function and constraints, wherein the objective function is specifically: f=min[f1,f2] Among them, f1 is the annual comprehensive cost, f2 is the improved voltage stability index; The constraints include power balance constraints, branch capacity constraints, node voltage constraints and DG penetration rate constraints.

8. The distributed power planning method based on the improved whale algorithm according to claim 7 is characterized in that: The annual comprehensive cost includes DG investment cost, operation and maintenance cost and network loss cost: Among them, P(s) is the probability of scene s, C I is the investment cost of DG, C OM is the DG operation and maintenance cost, C L Active power loss cost of distribution network; The improved voltage stability index is specifically: L=max(L ij ),i,j=1,2,…,N L ij =ω j H ij Among them, L is the largest improved voltage stability index in the system, L ij is the improved voltage stability index of the branch between node i and node j, H ij is the original voltage stability index of the branch between node i and node j, ω j is the load level coefficient, which is used to indicate the importance of load point j. ij The larger the value, the worse the system stability. ij The smaller it is, the more stable the system is.

9. A distributed power generation planning method based on improved whale algorithm according to claim 1, characterized in that: The step S3 specifically comprises the following steps: S31, arranging the improved voltage stability index corresponding to each branch in descending order of value; S32, selecting the top Z improved voltage stability index values, and using the end nodes of the corresponding branches as the nodes to be planned for the distributed power generation.

Images