Distributed energy new scene-oriented power distribution network operation risk assessment method

Abstract

The invention relates to the technical field of power distribution network operation assessment, in particular to a distributed energy new scene-oriented power distribution network operation risk assessment method, which comprises the following steps of: establishing a wind-solar output joint probability distribution function based on historical photovoltaic and wind power data; carrying out sampling and inverse transformation on the joint probability distribution function to obtain wind and light output data generated by considering correlation in each time period; clustering the sampled wind and light output data generated by considering the correlation in each time period to generate wind power and photovoltaic typical day scenes; and performing improved Latin hypercube sampling on each generated wind power and photovoltaic typical day scene, and calculating the node voltage out-of-limit risk of the power distribution network through load flow. Compared with the prior art, the method fully considers the correlation characteristics between distributed energy sources, considers the correlation of random variables in a sampling mode, and is more in line with the current operation situation of the novel power distribution network.

Description

Technical Field

[0001] The present invention relates to the field of distribution network operation, and in particular to a distribution network operation risk assessment method for a new distributed energy scenario. Background Art

[0002] In the distribution network, both distributed energy and loads are random. The output of wind turbines and photovoltaics in distributed energy has great volatility. Currently, few studies have taken the correlation between wind and solar into account in the distribution network operation assessment. There is a weak correlation between distributed energy such as distributed photovoltaics and distributed wind power. Considering the access of distributed energy alone will ignore the impact of the correlation between distributed energy. The access of electric vehicles to the distribution network will have a significant impact on the voltage of the distribution network nodes. Distributed energy can serve as the voltage support of the distribution network connected to electric vehicles, which will significantly improve the voltage level of the distribution network. The access of electric vehicles to different nodes will have different impacts on the operation risk of the distribution network, which will make the results of the distribution network operation risk inaccurate. The traditional consideration of the distribution network operation risk assessment does not consider the correlation between distributed photovoltaics and wind power, and rarely considers the impact of the access of electric vehicles to different nodes on the operation risk of the distribution network. Summary of the invention

[0003] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a distribution network operation risk assessment method taking into account the new scenario of distributed energy.

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] A distribution network operation risk assessment method for a new distributed energy scenario, comprising the following steps:

[0006] Based on historical photovoltaic and wind power data, a joint probability distribution function of wind and solar power output is established;

[0007] The joint probability distribution function is sampled and inversely transformed to obtain the wind and solar power output data generated in each period taking into account the correlation;

[0008] Cluster the wind and solar output data generated by considering the correlation in each time period obtained by sampling to generate typical daily scenes of wind power and photovoltaic power;

[0009] Improved Latin hypercube sampling is performed on the generated typical daily scenarios of wind power and photovoltaic power generation, and the operation risk of the distribution network is calculated through power flow.

[0010] As a preferred technical solution, the specific steps of establishing the joint probability distribution function include:

[0011] The kernel density estimation method is used to generate the probability density function of wind and solar output in each period of 24 hours for historical photovoltaic and wind power data;

[0012] A photovoltaic and wind power joint probability distribution function based on the Frank-Copula function is established for each time period.

[0013] As a preferred technical solution, the kernel density estimation method adopts a Gaussian kernel function:

[0014]

[0015] In the formula, d is the total number of days of historical data input, n is any hour in 24 hours, and x n represents the expected estimated probability density point, X n is the point in the sample data.

[0016] As an optimal technical solution, the joint probability distribution function is sampled and the inverse transformation is performed to obtain the wind and solar power output in each period, specifically:

[0017] The joint function and joint probability distribution function of each period are sampled using cubic spline interpolation method to solve the wind and solar output sampled in each period corresponding to the cumulative probability;

[0018] According to the sampling results and the inverse transformation of the joint probability distribution function, the output of distributed photovoltaic and wind power is calculated, and the wind and photovoltaic outputs generated by considering the correlation in each time period are obtained.

[0019] As a preferred technical solution, the spline interpolation method is expressed as:

[0020] S i (x) = a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3

[0021] In the formula, S i (x) indicates the sample diversity, x i Represents a node.

[0022] As a preferred technical solution, the distributed photovoltaic output model is as follows:

[0023] P=AG·η

[0024] Where P represents the instantaneous power output of the photovoltaic system, A represents the effective area of ​​the photovoltaic module, G represents the solar radiation intensity, and η represents the conversion efficiency of the photovoltaic module.

[0025] As a preferred technical solution, the distributed wind power output model is as follows:

[0026]

[0027] Where P w is the active power output of wind power; k and c represent the shape and scale parameters of Weibull distribution respectively; μ w and σ w Represent the mean and standard deviation of wind power active output respectively.

[0028] As a preferred technical solution, the wind and solar power output data is clustered using the K-means clustering algorithm, and the specific steps are:

[0029] Randomly select k data points as the initial cluster centers; assign data points, and for each data point in the data set, assign it to the nearest cluster center; update the cluster centers, calculate the average value of all points in each cluster, and update the average value as the new cluster center; repeat the above steps until the cluster center no longer changes or the preset number of iterations is reached.

[0030] As a preferred technical solution, the improved Latin hypercube sampling has the following specific steps:

[0031] According to the type of input variables and the linear correlation coefficient matrix C X , and obtain the correlation coefficient matrix C of the standard normal variable after Nataf transformation Z , and the correlation coefficient matrix C W Perform Cholesky decomposition to obtain the lower triangular matrix B;

[0032] Sampling M standard normal variables obtains the first sample matrix W N×M , and obtain the correlation coefficient matrix C between the columns of the first sample matrix W , for the correlation coefficient matrix C W Perform Cholesky decomposition to obtain the lower triangular matrix Q;

[0033] According to Z=W(Q T ) -1 B T The correlation matrix is ​​C Z The second sample matrix Z is then used to obtain the order matrix L according to the order of its elements in each column. S ;

[0034] The first sample matrix W N×M According to the order matrix L S After sorting, Latin hypercube sampling is performed to obtain the final sampling matrix S.

[0035] As an optimal technical solution, the final sampling matrix S is calculated through power flow to obtain the node voltage over-limit risk, as follows:

[0036] The node voltage over-limit risk considers the voltage over-limit probability. The voltage over-limit probability is the probability of exceeding the voltage deviation range allowed under the current voltage level of the node:

[0037]

[0038] Where V i Represents the voltage amplitude of node i; V i max and V i min Indicates the upper and lower limits of the node voltage; F(V) represents the voltage accumulation function;

[0039] Node voltage over-limit risk severity index Sev(V i ) is expressed as follows:

[0040]

[0041]

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention provides a new distribution network operation evaluation method that takes into account the new distributed energy scenario. It considers the correlation characteristics between distributed energy sources, establishes a joint probability distribution function based on the Frank-Copula function, and then obtains the wind and solar power output data generated by considering the correlation in each time period through sampling and inverse transformation; and proposes an improved Latin hypercube sampling, which takes into account the correlation of random variables in the sampling method, which is more in line with the current operation status of the new distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of the risk assessment of distribution network operation for the new distributed energy scenario of the present invention;

[0045] Figure 2 It is the topology diagram of IEEE34-node distribution network;

[0046] Figure 3 It is the photovoltaic time series distribution diagram based on the Frank-Copula correlation function;

[0047] Figure 4 It is the wind power time series distribution diagram based on the Frank-Copula correlation function;

[0048] Figure 5 To improve the Latin hypercube sampling method flow chart;

[0049] Figure 6It is the voltage over-limit risk diagram of each node in the distribution network. DETAILED DESCRIPTION

[0050] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0051] Example 1

[0052] The present invention provides a distribution network operation risk assessment method for a new distributed energy scenario, such as Figure 1 As shown, the specific implementation method is as follows:

[0053] S1: The kernel density estimation method is used to generate the probability density function of wind and solar output in each period of 24 hours for historical photovoltaic and wind power data;

[0054] S2: Establish a joint probability distribution function based on the Frank-Copula function for each time period;

[0055] S3: Sampling the joint Copula function and the joint probability distribution function of each time period, and finally obtaining the wind power and photovoltaic power output of each time period according to the sampling results and the inverse transformation of the joint probability distribution function, that is, the wind and photovoltaic output generated by considering the correlation;

[0056] S4: Cluster the sampled data to generate 20 typical day scenarios for wind power and photovoltaic power;

[0057] S5: Improved Latin hypercube sampling is performed on each typical daily scenario of wind power and photovoltaic power generation, and the operation risk of the distribution network is calculated through power flow.

[0058] In step S1, the kernel density estimation method is used for the historical photovoltaic and wind power data to analyze the input photovoltaic and wind power data for each time period.

[0059] The kernel density estimation method selects the Gaussian kernel function:

[0060]

[0061] In the formula, d is the total number of days of historical data input, n is any hour in 24 hours, and x n represents the expected estimated probability density point, X n is the point in the sample data.

[0062] In step S2, based on the Frank-Copula function combined probability distribution function, taking into account the weak correlation between photovoltaic and wind power, and according to research literature, the correlation function between photovoltaic and wind power is close to the Frank-Copula empirical function, a photovoltaic and wind power combined probability distribution function based on the Frank-Copula function is established.

[0063] Frank-Copula distribution function:

[0064]

[0065] Frank-Copula density function:

[0066]

[0067] Wind-solar joint probability model for each period

[0068] F(x, y) = C(F X (x), F Y (y))

[0069] In step S3, the joint distribution function is sampled, and the wind and solar outputs of each time period are obtained by inverse transformation, and the wind and solar outputs sampled in each time period corresponding to the cumulative probability are solved by cubic spline interpolation method.

[0070] Spline interpolation method:

[0071] S i (x) = a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3

[0072] In the formula, S i (x) indicates the sample diversity, x i Represents a node.

[0073] Secondly, the output of distributed photovoltaic and wind power is calculated. The modeling method of distributed photovoltaic output is as follows:

[0074] P=A·G·η

[0075] Where, Ω represents the instantaneous power output of the PV system, A represents the effective area of ​​the PV module, G represents the solar radiation intensity, and η represents the conversion efficiency of the PV module, which is usually between 15% and 20%.

[0076] The distributed wind power output modeling method is as follows:

[0077]

[0078] Where P w is the active power output of wind power; k and c represent the shape and scale parameters of Weibull distribution respectively; μ w and σ w Represent the mean and standard deviation of wind power active output respectively.

[0079] In step S4, a clustering algorithm generates a typical day scene, and a K-means clustering algorithm is used to perform clustering on the N groups of sampling results.

[0080] The specific steps of the K-means clustering algorithm are: randomly select k data points as the initial cluster centers (centroids); assign data points, for each data point in the data set, assign it to the nearest cluster center. Usually the Euclidean distance is used to measure the distance; update the cluster center, calculate the average value of all points in each cluster, and update the average value as the new cluster center; iterate, repeat the above steps until the cluster center no longer changes or the preset number of iterations is reached.

[0081] In step S5, improved Latin hypercube sampling is performed on each wind power and photovoltaic scenario, and the distribution network operation risk is calculated through power flow.

[0082] The present invention considers improving the Latin Hypercube Sampling (LHS) sampling process mainly based on the inverse function conversion method. Assuming that the sampling scale is N, the number of random variables is M, which are X1, X2, ..., X M , random variable X K The cumulative probability distribution function of is:

[0083] Y K =F K (X K ), K = 1, 2, ..., M

[0084] Divide the value range of the distribution function [0, 1] into N non-overlapping equally spaced subintervals [n / N, (n+1) / N], n = 0, 1, ..., (N-1) so that the length of each interval is 1 / N; then select a T in each subinterval n , that is, T n =(n+1-ξ) / N. The selection methods of ξ are mainly the following: 0.5, Hammersley point set, interval boundary, random number. Among them, ξ=0.5 is the point matrix sampling. n Then use the inverse transform to get the sample value:

[0085]

[0086] in, Yes F k The inverse transformation of (·).

[0087] However, when LHS processes multiple input random variables, the simulation accuracy is not only affected by the sampling value, but also by the correlation of the sampling values ​​of different input random variables. Different correlation control methods use different correlation coefficients, and the specific methods include: CLMCS method, single-key optimization algorithm, Cholesky decomposition, genetic algorithm, simulated annealing algorithm, etc. After obtaining a sorting matrix with a certain correlation, the sampling matrix is ​​arranged according to the sorting matrix to obtain a sampling matrix with the same correlation.

[0088] The present invention adopts a LHCS sorting method that processes the correlation of random variables more accurately, and the specific steps are:

[0089] 5.1) Calculate the cumulative probability distribution function of photovoltaic and wind power historical data and input system parameters;

[0090] 5.2) Input system parameters, distribution function of random variables, and linear correlation coefficient matrix C of variables X , where the variable linear correlation coefficient matrix C X It consists of the Pearson correlation coefficient between two variables;

[0091] 5.3) According to the type of input variables and the linear correlation coefficient matrix C X Get the correlation coefficient matrix C of the standard normal variable after Nataf transformation Z , and the correlation coefficient matrix C Z Perform Cholesky decomposition to obtain the lower triangular matrix B;

[0092] 5.4) Sample M standard normal variables to obtain the sample matrix W N×M , solve to get the correlation coefficient matrix C between its columns W , and the correlation coefficient matrix C Z Perform Cholesky decomposition to obtain the lower triangular matrix Q;

[0093] 5.5) According to Z = W (Q T ) -1 B T The correlation matrix is ​​C Z The sample matrix Z is then used to obtain the order matrix L according to the order of its elements in each column. S ;

[0094] 5.6) The sample matrix obtained by sampling in the previous text is based on the order matrix L SAfter sorting, Latin hypercube sampling is performed to obtain the final sampling matrix S.

[0095] The final sampling matrix S is calculated through power flow to obtain the node voltage over-limit risk value. In this method, the node voltage over-limit risk takes into account the voltage over-limit probability. The voltage over-limit probability is the probability of exceeding the voltage deviation range allowed under the current voltage level of the node:

[0096]

[0097] Where V i Represents the voltage amplitude of node i, V i max and V i min represents the upper and lower limits of the node voltage, 1.05 and 0.95, and F(V) represents the voltage accumulation function.

[0098] Using Sev(V i ) represents the node voltage over-limit risk severity index, as shown in the following formula:

[0099]

[0100] Traditionally, the correlation between distributed photovoltaic and wind power is rarely considered when considering the risk assessment of distribution network operation. The new distribution network operation assessment method taking into account the new scenario of distributed energy provided by the present invention considers the correlation characteristics between distributed energy sources and the correlation of random variables in the sampling method, which is more in line with the current operation status of the new distribution network.

[0101] Example 2

[0102] As one of the specific implementation examples of the present invention, Figure 2 As shown in the figure, this example uses the IEEE 34-node power distribution system and takes the historical photovoltaic and wind power data of a certain area in eastern China as the basis. The total area of ​​photovoltaic panels is 16000m 2 , the photoelectric conversion efficiency is 14%, the wind farm cut-in wind speed is 3m / s, the cut-out wind speed is 15m / s, the rated wind speed is 13.5m / s, and the rated power of each wind farm is 1MW.

[0103] Based on the historical wind and solar power output data, the kernel density estimation method is used to generate the wind and solar power output probability density function for each hour of 24h, and the Frank-Copula function is selected to describe the wind and solar power output correlation. The joint probability distribution function of each hour is sampled, and finally the wind and solar power correlation output scene and scene probability are obtained by inverse transformation. The typical daily scene is generated according to the K-means clustering method. Figure 3 , 4 shown.

[0104] like Figure 5,The present invention adopts a LHCS sampling method that considers correlation, takes the extracted samples as input parameters for ,power flow calculation, and finally outputs the node voltage over-limit risk.

[0105] like Figure 6 As shown in the figure, from the calculation results of voltage over-limit risk, it can be seen that the maximum voltage over-limit risk occurs at node 34, which is also the solar and wind power access node, and the voltage over-limit risk value reaches 0.035. The closer to node 34, the greater the voltage over-limit risk value, while the closer to the end node, the lower the voltage amplitude, and nodes 1-6 basically do not produce voltage over-limit.

[0106] In summary, the present invention proposes a new distribution network operation risk assessment method taking into account new distributed energy scenarios. First, the distributed energy correlation scenario is modeled. Secondly, based on correlation control, the present invention proposes an LHCS sampling method that considers correlation, and performs flow calculations to summarize the node voltage over-limit risks. The new distribution network operation risk assessment method taking into account new distributed energy scenarios provided by the present invention proposes a distributed energy sampling method that is more suitable for considering correlation, which can provide a reference for current research on sampling of correlation scenarios.

[0107] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A distribution network operation risk assessment method for a new distributed energy scenario, characterized in that the steps include: Based on historical photovoltaic and wind power data, a joint probability distribution function of wind and solar power output is established; The joint probability distribution function is sampled and inversely transformed to obtain the wind and solar power output data generated in each period taking into account the correlation; Cluster the wind and solar output data obtained from the sampling in each period to generate typical daily scenes of wind power and photovoltaic power; Improved Latin hypercube sampling is performed on the generated typical daily scenarios of wind power and photovoltaic power generation, and the voltage over-limit risk of distribution network nodes is calculated through power flow.

2. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 1, characterized in that: The specific steps of establishing the joint probability distribution function include: The kernel density estimation method is used to generate the probability density function of wind and solar output in each period of 24 hours for historical photovoltaic and wind power data; A photovoltaic and wind power joint probability distribution function based on the Frank-Copula function is established for each time period.

3. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 2, characterized in that: The kernel density estimation method described uses a Gaussian kernel function: In the formula, d is the total number of days of historical data input, n is any hour in 24 hours, and x n represents the expected estimated probability density point, X n is the point in the sample data.

4. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 2, characterized in that: The joint probability distribution function is sampled and inversely transformed to obtain the wind and solar power output in each period, specifically: The joint function and joint probability distribution function of each period are sampled using cubic spline interpolation method to solve the wind and solar output sampled in each period corresponding to the cumulative probability; According to the sampling results and the inverse transformation of the joint probability distribution function, the output of distributed photovoltaic and wind power is calculated, and the wind and photovoltaic outputs generated by considering the correlation in each time period are obtained.

5. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 4, characterized in that: The spline interpolation method is expressed as: S i (x)=a i +b i (x-x i )+c i (x-x i ) 2 +d i (x-x i ) 3 In the formula, S i (x) indicates the sample diversity, x i Represents a node.

6. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 4, characterized in that: The distributed photovoltaic output modeling is as follows: P=A·G·η Where, Ω represents the instantaneous power output of the photovoltaic system, A represents the effective area of ​​the photovoltaic module, G represents the solar radiation intensity, and η represents the conversion efficiency of the photovoltaic module.

7. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 4, characterized in that: The distributed wind power output modeling is as follows: Where P w is the active power output of wind power; k and c represent the shape and scale parameters of Weibull distribution respectively; μ w and σ w Represent the mean and standard deviation of wind power active output respectively.

8. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 1, characterized in that: The wind and solar power output data are clustered using the K-means clustering algorithm, and the specific steps are as follows: Randomly select k data points as the initial cluster centers; assign data points, and for each data point in the data set, assign it to the nearest cluster center; update the cluster centers, calculate the average value of all points in each cluster, and update the average value as the new cluster center; repeat the above steps until the cluster center no longer changes or the preset number of iterations is reached.

9. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 1, characterized in that: The improved Latin hypercube sampling has the following specific steps: According to the type of input variables and the linear correlation coefficient matrix C X , and obtain the correlation coefficient matrix C of the standard normal variable after Nataf transformation z , and the correlation coefficient matrix C W Perform Cholesky decomposition to obtain the lower triangular matrix B; Sampling M standard normal variables obtains the first sample matrix W N×M , and obtain the correlation coefficient matrix C between the columns of the first sample matrix W , for the correlation coefficient matrix C W Perform Cholesky decomposition to obtain the lower triangular matrix Q; According to Z=W(Q T ) -1 B T The correlation matrix is ​​C z The second sample matrix Z is then used to obtain the order matrix L according to the order of its elements in each column. S ; The first sample matrix W N×M According to the order matrix L S After sorting, Latin hypercube sampling is performed to obtain the final sampling matrix S.

10. A distribution network operation risk assessment method for a new distributed energy scenario according to claim 9, characterized in that: For the final sampling matrix S, the voltage over-limit risk of the distribution network nodes is obtained through power flow calculation, as follows: The node voltage over-limit risk considers the voltage over-limit probability, which is the probability of exceeding the voltage deviation range allowed under the current voltage level of the node: Where V i Represents the voltage amplitude of node i; V imax and V imi n represents the upper and lower limits of the node voltage; F(V) represents the voltage accumulation function; Node voltage over-limit risk severity index Seυ(V i ) is expressed as follows:

Images