Distributed photovoltaic area distribution network scheduling risk assessment method considering photovoltaic power border crossing

Abstract

The invention relates to a distributed photovoltaic area distribution network scheduling risk assessment method considering photovoltaic power border crossing, and belongs to the field of distribution network scheduling risk assessment. According to the method, firstly, richness of photovoltaic power prediction models is considered, prediction power sequences of two different photovoltaic power prediction models are selected, and the DTW distance between the two prediction power sequences is calculated through a dynamic time warping method; secondly, defining a photovoltaic prediction power uncertainty boundary, and giving a power border-crossing label calculation method; and calculating the photovoltaic power border crossing probability of a single node through a data-driven logistic regression model. And finally, based on a Copula function theory, analyzing the distribution network scheduling risk of the distributed photovoltaic area.

Description

Technical Field

[0001] The present invention belongs to the field of distribution network scheduling risk assessment, and specifically relates to a distribution network scheduling risk assessment method for a distributed photovoltaic area taking into account photovoltaic power out-of-bounds. Background Art

[0002] With the increasing global demand for clean energy, photovoltaic power generation has become an important way to achieve a low-carbon economy and reduce greenhouse gas emissions. Distributed photovoltaic systems are rapidly becoming popular because they can be flexibly deployed in buildings, factories and residential areas. However, the output of photovoltaic systems is affected by multiple variables such as weather, temperature, and light intensity. The intermittent and uncertain output of photovoltaic systems poses challenges to the stable operation of the power grid. To meet this challenge, researchers have proposed a variety of methods to deal with the uncertainty of distributed photovoltaic output. Among them, robust optimization methods can provide a solution that performs well in the worst case by considering these uncertainties in the design and planning stages, so they are widely used. Robust optimization methods usually assume that photovoltaic power is always within a certain range and give a scheduling plan based on this assumption. However, if the actual photovoltaic power exceeds the upper and lower limits, it will affect the actual scheduling process, bring economic losses, and even cause line power overload problems, affecting the safe and stable operation of the system. Therefore, it is crucial to reasonably give the range of changes in photovoltaic predicted power and evaluate the risk of photovoltaic power crossing the boundary.

[0003] The prior art has the following defects:

[0004] 1. Existing research usually focuses on improving the accuracy of distributed photovoltaic output forecasting. However, in actual operation, photovoltaic output is easily affected by weather conditions, and its forecast will inevitably have deviations. Existing research usually ignores the evaluation of predicted power deviation.

[0005] 2. Although previous optimization scheduling methods take into account the uncertainty of distributed photovoltaic output, they usually directly give the possible fluctuation range (uncertainty set) of photovoltaic output without considering the situation where the photovoltaic power fluctuation exceeds this range and the corresponding cross-boundary risk. Summary of the invention

[0006] The purpose of the present invention is to provide a method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power taking into account photovoltaic power out-of-bounds. First, considering the richness of photovoltaic power prediction models, two different prediction models' predicted power sequences are selected, and the DTW distance of the two sequences is calculated by the dynamic time warping method. Secondly, the uncertainty boundary of photovoltaic predicted power is defined, and a method for calculating power out-of-bounds labels is given. The probability of photovoltaic power out-of-bounds of a single node is calculated by a data-driven logistic regression model. Finally, based on the Copula function theory, the risk of regional distribution network scheduling containing distributed photovoltaic power is analyzed.

[0007] To achieve the above purpose, the technical solution of the present invention is: a method for risk assessment of regional distribution network scheduling containing distributed photovoltaics considering photovoltaic power out-of-bounds, comprising:

[0008] The predicted power sequences of two different photovoltaic power prediction models are selected, and the DTW distance between the two predicted power sequences is calculated by the dynamic time warping method.

[0009] Define the uncertainty boundary of photovoltaic power prediction, provide a method to calculate the power out-of-bounds label, and calculate the probability of photovoltaic power out-of-bounds of a single node through a data-driven logistic regression model;

[0010] Assess the risks of regional distribution network scheduling containing distributed photovoltaics.

[0011] In one embodiment of the present invention, the specific implementation method of selecting predicted power sequences of two different photovoltaic power prediction models and calculating the DTW distance of the two predicted power sequences by the dynamic time warping method is as follows:

[0012] Assume that two different photovoltaic power prediction models have been established on any node containing distributed photovoltaics, and the output prediction power series of the two photovoltaic power prediction models are time series of the same length:

[0013] P1={P 1,1 ,P 1,2 ,...,P 1,T}

[0014] P2={P 2,1 ,P 2,2 ,...,P 2,T}

[0015] Where P1 is the main prediction sequence, P2 is the auxiliary prediction sequence, and T is the length of the time series; it is assumed that the prediction accuracy of the main prediction sequence is higher than that of the auxiliary prediction sequence;

[0016] The DTW distance between P1 and P2 is calculated by the dynamic time warping method:

[0017] First calculate the Euclidean distance between any two time points:

[0018] d(i,j)=dist(P 1,i ,P 2,j )=|P 1,i -P 2,j |

[0019] Where d(i,j) represents the i-th element P of sequence P1 1,i and the jth element P of sequence P2 2,j The distance between

[0020] Construct a T×T distance matrix M:

[0021]

[0022] Calculate the cumulative distance C through the distance matrix M:

[0023] C(i,j)=d(i,j)+min{C(i-1,j),C(i,j-1),C(i-1,j-1)}

[0024] Where C(i,j) represents the minimum cumulative distance from the starting point to position i,j, min{C(i-1,j),C(i,j-1),C(i-1,j-1)} represents the minimum value from the top, left or diagonal direction; the boundary conditions of the cumulative distance are:

[0025] C(0,0)=d(0,0)

[0026]

[0027] The final DTW distance between sequence P1 and sequence P2 is expressed as:

[0028] d DTW (P1,P2)=C(T,T).

[0029] In one embodiment of the present invention, the specific implementation method of defining the uncertainty boundary of photovoltaic power prediction is as follows:

[0030] Assume that the time series of the predicted power sequence of a single node photovoltaic power is;

[0031] P pred = {P pred,1 ,P pred,2 ,...,P pred,T}

[0032] Where T is the prediction time domain;

[0033] The actual output of photovoltaic power will always deviate from the predicted value. The uncertainty set method is used to limit the uncertain variables to a certain range. The upper and lower limits of the fluctuation of photovoltaic power in the uncertainty set, that is, the error band, are defined as α%, then:

[0034]

[0035] Among them, P upper represents the upper bound of the uncertain set, P lower Represents the lower bound of the uncertain set.

[0036] In one embodiment of the present invention, the specific implementation of the power out-of-bounds label calculation method is as follows:

[0037] Assuming that there is historical photovoltaic operation data on any node containing distributed photovoltaics, the photovoltaic power prediction model obtains the historical photovoltaic power main prediction sequence P1 and the historical photovoltaic power actual sequence P0, thereby defining the calculation method of the power out-of-bounds label:

[0038]

[0039] Where P 0,i is the ith element of the actual output sequence, P 1,i To predict the i-th element of the output sequence, sign(x) is the sign function:

[0040]

[0041] When y=1, it means that the actual power sequence is out of bounds, and when y=0, it means that it is not out of bounds. Obviously, the value of the out-of-bounds label is related to α;

[0042] When the labels of the training set samples are unevenly distributed, the photovoltaic power prediction model tends to predict the majority class, which makes the photovoltaic power prediction model inaccurate for the minority class. The Latin hypercube sampling method is used to construct the training samples, so that the input variable α is random and the output label y is as uniform as possible. The details are as follows:

[0043] First calculate the error band boundary values:

[0044]

[0045] Determine the sampling interval as Divide the interval into N subintervals of equal length; the length of each subinterval is:

[0046]

[0047] Therefore, the i-th subinterval is:

[0048]

[0049] Where, i = 1, 2, ..., N;

[0050] A point is randomly selected in each subinterval and randomly sorted to form a set of values ​​of the input variable α of the training set.

[0051] In one embodiment of the present invention, the specific implementation method of calculating the probability of photovoltaic power crossing the boundary of a single node by using a data-driven logistic regression model is as follows:

[0052] D DTW , α is used as the input of the model training set, and the sample label y is used as the output of the model training set;

[0053] The inner layer of the logistic regression model uses a model with polynomial features to capture the nonlinear relationship between input variables and enhance the model's fitting ability. The two-input variable polynomial function used by the model with polynomial features is as follows:

[0054]

[0055] Where w1, w2, ..., w5 are the weights of the model, b is the bias term, x1 and x2 are the model input features, i.e., d DTW , α;

[0056] The outer layer of the logistic regression model uses the sigmoid function to obtain the output probability value:

[0057]

[0058] The loss function of the logistic regression model uses the cross entropy function:

[0059]

[0060] Where i = 1, 2, …, N, N is the total number of samples in the model.

[0061] The logistic regression model is trained by historical data, and the DTW distance d of the power sequence is predicted in actual scheduling. DTW The actual power out-of-bounds probability P(y=1) of a single-node photovoltaic is obtained by using α.

[0062] In one embodiment of the present invention, since the inner layer is a polynomial function, its weight dimension is more than that of a linear function, and it is more likely to cause overfitting problems; the L2 regularization method is used here to reduce the weights of unimportant features and enhance the generalization ability of the model, so the loss function with regularization terms is:

[0063]

[0064] Where λ is the regularization term coefficient, which is used to balance the complexity and fitting accuracy of the model.

[0065] In one embodiment of the present invention, the risk of regional distribution network scheduling containing distributed photovoltaics is evaluated based on Copula function theory.

[0066] In one embodiment of the present invention, based on the Copula function theory, the specific implementation method of evaluating the risk of the distribution network scheduling in the distributed photovoltaic area is as follows:

[0067] The out-of-bounds situation of the photovoltaic power of a single node is unknown during scheduling, so the out-of-bounds situation Y is a random variable in the scheduling stage. Let x be a binary variable to determine whether the power is out of bounds, then the survival function of the random variable Y is:

[0068] S(x)=P(Y≥x)x∈{0,1}

[0069] The two-dimensional Gumbel-Copula function is expressed as follows:

[0070]

[0071] Where u1 and u2 are random variables; θ∈(0,1) is the correlation coefficient between random variables u1 and u2; if the object to be described is a multivariate joint distribution function, then the N-dimensional Gumbel-Copula function is:

[0072]

[0073] Sklar's theorem shows that for any d-dimensional joint survival function There exists a Copula function C such that the joint survival function is expressed as the marginal survival function S i (x i ) and Copula functions:

[0074]

[0075] When the photovoltaic power of any node exceeds the limit, it will affect the scheduling. When the number of distributed photovoltaic access nodes in the regional distribution network is N, the evaluation index function of the scheduling scheme in the region is obtained:

[0076] R(x)=P(Y1<x,Y2<x,...,Y N <x)

[0077] Where Y1, Y2, ..., Y N is the power out-of-bounds situation of each distributed photovoltaic node. When x is 1, it means that the power of each node is within the bounds, that is, the dispatch is reliable. Considering the definition of the survival function, the above formula is converted into the following form by the addition formula:

[0078]

[0079] Writing the above equation in compact form:

[0080]

[0081] According to Sklar theory, there exists an N-dimensional Copula function such that:

[0082]

[0083] According to the definition of Gumbel-Copula function, we can get:

[0084]

[0085] Therefore, the overall evaluation index of the regional scheduling plan is obtained as follows:

[0086]

[0087] The larger the R(x) value is, the lower the probability of power out-of-bounds risk in the corresponding scheduling scheme is.

[0088] The present invention also provides a distributed photovoltaic regional distribution network scheduling risk assessment system that takes into account photovoltaic power out-of-bounds, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0089] The present invention also provides a computer-readable storage medium, on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0090] Compared with the prior art, the present invention has the following beneficial effects: The method of the present invention, firstly, takes into account the richness of photovoltaic power prediction models, selects predicted power sequences of two different prediction models, and calculates the DTW distance of the two sequences by the dynamic time warping method. Secondly, the uncertainty boundary of photovoltaic predicted power is defined, and a method for calculating the power out-of-bounds label is given. The probability of photovoltaic power out-of-bounds of a single node is calculated by a data-driven logistic regression model. Finally, based on the Copula function theory, the risk of distribution network scheduling in areas containing distributed photovoltaics is analyzed; the method of the present invention can realize the risk assessment of distribution network scheduling in areas containing distributed photovoltaics. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 The figure is a flow chart of the method of the present invention.

[0092] Figure 2 It is a typical topological structure of a regional distribution network including distributed photovoltaics.

[0093] Figure 3 Schematic diagram for analyzing photovoltaic power out-of-bounds situation. DETAILED DESCRIPTION

[0094] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0095] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0096] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0097] The present invention provides a method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds, comprising:

[0098] The predicted power sequences of two different photovoltaic power prediction models are selected, and the DTW distance between the two predicted power sequences is calculated by the dynamic time warping method.

[0099] Define the uncertainty boundary of photovoltaic power prediction, provide a method to calculate the power out-of-bounds label, and calculate the probability of photovoltaic power out-of-bounds of a single node through a data-driven logistic regression model;

[0100] Assess the risks of regional distribution network scheduling containing distributed photovoltaics.

[0101] The following is the specific implementation process of the present invention.

[0102] like Figure 1 As shown, this embodiment provides a method for assessing the risk of regional distribution network scheduling containing distributed photovoltaic power taking into account photovoltaic power out-of-bounds. First, considering the richness of photovoltaic power prediction models, the predicted power sequences of two different prediction models are selected, and the DTW distance of the two sequences is calculated by the dynamic time warping method. Secondly, the uncertainty boundary of photovoltaic predicted power is defined, and a method for calculating the power out-of-bounds label is given. The probability of photovoltaic power out-of-bounds of a single node is calculated by a data-driven logistic regression model. Finally, based on the Copula function theory, the risk of regional distribution network scheduling containing distributed photovoltaic power is analyzed.

[0103] 1. Regional distribution network structure and dispatch risk analysis including distributed photovoltaics

[0104] Typical topological structures of regional distribution networks with distributed photovoltaics are as follows: Figure 2 shown.

[0105] Photovoltaic output is intermittent and uncertain, and is easily affected by the weather, which makes it impossible to make a completely accurate prediction of photovoltaic output at this stage. At present, the method of dealing with the uncertainty of photovoltaic output usually limits the fluctuation range of photovoltaic output to a certain range. For a distribution network with multiple nodes connected to distributed photovoltaics, any node's photovoltaic power out of bounds will have a significant impact on the overall operation of the regional distribution network system.

[0106] 2. Analysis of the fluctuation range and out-of-bounds situation of single-node photovoltaic power prediction

[0107] Assume that the predicted time series of single-node photovoltaic power is

[0108] P pred = {P pred,1 ,P pred,2 ,...,P pred,T} (1)

[0109] Among them, T is the prediction time domain.

[0110] The actual output of photovoltaic power will always have a certain deviation from the predicted value. Therefore, the uncertainty set is usually used to limit the uncertain variables to a certain range. The upper and lower limits (error band) of the fluctuation of photovoltaic power in the uncertainty set are defined as α%, then

[0111]

[0112] Among them, P upper represents the upper bound of the uncertain set, P lower Represents the lower bound of the uncertain set.

[0113] The schematic diagram of the out-of-bounds analysis of photovoltaic power is as follows: Figure 3 shown.

[0114] The red dotted line is the predicted power sequence P pred , the upper and lower bounds of the uncertain set P upper , P lower The shaded area in the figure is composed of the blue solid line, which is the actual power sequence P actual Obviously, the larger the error band α is, the larger the shaded area is, and the actual power sequence P actual The probability of crossing the boundary is lower.

[0115] In the process of optimizing scheduling considering the uncertainty of photovoltaic output, the error band α of the uncertainty set directly affects the economic efficiency of scheduling. The smaller α is, the smaller the fluctuation of photovoltaic output is considered, and the higher the economic efficiency of scheduling is. However, the higher the probability of actual photovoltaic power crossing the boundary is, the greater the scheduling risk is. The present invention intends to use data-driven methods to predict the probability of actual power crossing the boundary through analysis and processing of existing data, so as to provide guidance for the reasonable value of error band α.

[0116] 3. Dynamic Time Warping Method to Calculate DTW Distance of Power Series

[0117] The present invention does not involve the modeling process of the photovoltaic power prediction model. It is assumed that two different power prediction models have been established on any node containing distributed photovoltaics, and the outputs of the two prediction models are time series of the same length:

[0118] P1={P1,1 ,P 1,2 ,...,P 1,T} (4)

[0119] P2={P 2,1 ,P 2,2 ,...,P 2,T} (5)

[0120] Where P1 is the main prediction sequence, P2 is the auxiliary prediction sequence, and T is the length of the time series. It is assumed that under normal circumstances, the prediction accuracy of the main prediction sequence is higher than that of the auxiliary prediction sequence.

[0121] Dynamic Time Warping (DTW) can align and compare time series data and analyze the similarity between two time series. The smaller the DTW distance, the more similar the two series are. The process of calculating the DTW distance between P1 and P2 is as follows.

[0122] First, calculate the Euclidean distance between any two time points

[0123] d(i,j)=dist(P 1,i ,P 2,j )=|P 1,i -P 2,j | (6)

[0124] Where d(i,j) represents the distance between the i-th element of sequence P1 and the j-th element of sequence P2.

[0125] From this, we can construct a T×T order distance matrix M

[0126]

[0127] The cumulative distance C can be calculated through the distance matrix M

[0128] C(i,j)=d(i,j)+min{C(i-1,j),C(i,j-1),C(i-1,j-1)} (8)

[0129] Where C(i,j) represents the minimum cumulative distance from the starting point to position i,j, and min{C(i-1,j),C(i,j-1),C(i-1,j-1)} represents the minimum value from the top, left, or diagonal direction. The boundary conditions of the cumulative distance are

[0130] C(0,0)=d(0,0) (9)

[0131]

[0132] The final DTW distance between sequence P1 and sequence P2 can be expressed as

[0133] d DTW (P1,P2)=C(T,T) (12)

[0134] 4. Definition of training sample labels and sampling method of error band α

[0135] Assuming that there is a certain amount of photovoltaic operation history data on any node containing distributed photovoltaics, the photovoltaic power prediction model can obtain the historical photovoltaic power prediction main sequence P1, and the historical photovoltaic power actual sequence P0 can also be obtained. This can define the calculation method of the power out-of-bounds label

[0136]

[0137] Where P 0,i is the ith element of the actual output sequence, P 1,i is the i-th element of the predicted output sequence, sign(x) is the sign function

[0138]

[0139] When y=1, it indicates that the actual power sequence is out of bounds, and when y=0, it indicates that it is not out of bounds. Obviously, the value of the out-of-bounds label is related to the error band α.

[0140] When the labels of training set samples are unevenly distributed, the model tends to predict the majority class, which makes the model inaccurate for the minority class. The present invention uses Latin hypercube sampling method to construct training samples, so that the input variable α has a certain randomness while the output label y is as uniform as possible.

[0141] First calculate the error band boundary value

[0142]

[0143] Determine the sampling interval as Divide the interval into N subintervals of equal length. The length of each subinterval is

[0144]

[0145] Therefore, the i-th subinterval is

[0146]

[0147] Among them, i=1,2,…,N.

[0148] A point is randomly selected in each subinterval and randomly sorted to form a set of values ​​of the input variable α of the training set.

[0149] 5. Basic structure and model training of logistic regression model

[0150] The input and output sample sets have been constructed in the previous text. Since the output labels have only two categories, 0 and 1, it is essentially a binary classification problem. Therefore, the mapping relationship between input and output can be fitted by the logistic regression method. DTW , α is used as the input of the model training set, and the sample label y is used as the output of the model training set.

[0151] The traditional logistic regression model usually uses a linear function in its inner layer, while the model using polynomial features can capture the nonlinear relationship between input variables and enhance the fitting ability of the model. The two-input variable polynomial function used in the present invention is as follows:

[0152]

[0153] Where w1, w2, ..., w5 are the weights of the model, b is the bias term, x1 and x2 are the model input features, i.e., d DTW 、α.

[0154] The outer layer of logistic regression uses the sigmoid function to obtain the output probability value

[0155]

[0156] The loss function of the logistic regression model uses the cross entropy function

[0157]

[0158] Since the inner layer is a polynomial function, its weight dimension is more than that of a linear function, which is more prone to overfitting problems. Here, the L2 regularization method is used to reduce the weights of unimportant features and enhance the generalization ability of the model, so the loss function with regularization term is

[0159]

[0160] Where λ is the regularization term coefficient, which is used to balance the complexity and fitting accuracy of the model.

[0161] The logistic regression model is trained by historical data, so that the DTW distance d of the power sequence can be predicted in actual scheduling. DTW The actual power out-of-bounds probability P(y=1) of a single-node photovoltaic is obtained by summing the error band α.

[0162] 6. Risk analysis of regional distribution network overall power outage based on copula function

[0163] For regional active distribution networks with multiple nodes connected to distributed photovoltaics, any one node's photovoltaic power out of bounds will have a significant impact on the actual scheduling process, so it is necessary to conduct a risk analysis of the regional distribution network as a whole. The present invention uses Copula theory to conduct risk assessment on regional distribution network scheduling considering multi-node distributed photovoltaic power out of bounds.

[0164] The out-of-bounds situation of the photovoltaic power of a single node is unknown during scheduling, so the out-of-bounds situation Y is a random variable in the scheduling stage. Let x be a binary variable to determine whether the power is out of bounds. The survival function of the random variable Y is

[0165] S(x)=P(Y≥x) x∈{0,1} (22)

[0166] The Copula function is a function that connects a multidimensional distribution function with the marginal distribution functions of each dimension. The Copula function can flexibly specify the dependency structure between variables without changing the marginal distribution. The two-dimensional Gumbel-Copula function can be expressed as follows

[0167]

[0168] Where u1 and u2 are random variables; θ∈(0,1) is the correlation coefficient between random variables u1 and u2. If the object to be described is a multivariate joint distribution function, then the N-dimensional Gumbel-Copula function is

[0169]

[0170] Sklar's theorem shows that for any d-dimensional joint survival function There exists a Copula function C such that the joint survival function can be expressed as the marginal survival function S i (x i ) and Copula functions

[0171]

[0172] When the photovoltaic power of any node exceeds the limit, it will affect the scheduling. When the number of distributed photovoltaic access nodes in the regional distribution network is N, the scheduling scheme evaluation index function of the region can be obtained:

[0173] R(x)=P(Y1<x,Y2<x,...,Y N <x) (26)

[0174] Where Y1, Y2, ..., Y Nis the power out-of-bounds situation of each distributed photovoltaic node. When x is 1, it means that the power of each node is within the bounds, that is, the dispatch is reliable. Considering the definition of the survival function, the addition formula converts equation (26) into the following form:

[0175]

[0176] Write equation (27) in a compact form

[0177]

[0178] According to Sklar theory, there exists an N-dimensional Copula function such that

[0179]

[0180] According to the definition of Gumbel-Copula function, we can get

[0181]

[0182] Therefore, the overall dispatching scheme evaluation index of the region can be obtained as follows:

[0183]

[0184] The larger the R(x) value is, the lower the probability of power out-of-bounds risk in the scheduling scheme is.

[0185] The present invention also provides a distributed photovoltaic regional distribution network scheduling risk assessment system that takes into account photovoltaic power out-of-bounds, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0186] The present invention also provides a computer-readable storage medium, on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0187] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0188] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0189] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0190] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0191] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A method for risk assessment of regional distribution network scheduling containing distributed photovoltaics considering photovoltaic power out-of-bounds, characterized in that: include: The predicted power sequences of two different photovoltaic power prediction models are selected, and the DTW distance between the two predicted power sequences is calculated by the dynamic time warping method. Define the uncertainty boundary of photovoltaic power prediction, provide a method to calculate the power out-of-bounds label, and calculate the probability of photovoltaic power out-of-bounds of a single node through a data-driven logistic regression model; Assess the risks of regional distribution network scheduling containing distributed photovoltaics.

2. According to claim 1, a method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds, characterized in that: Select two different photovoltaic power prediction models to predict the power series, and calculate the DTW distance between the two prediction power series by the dynamic time warping method. The specific implementation method is as follows: Assume that two different photovoltaic power prediction models have been established on any node containing distributed photovoltaics, and the output prediction power series of the two photovoltaic power prediction models are time series of the same length: P1={P 1,1 ,P 1,2 ,...,P 1,T } P2={P 2,1 ,P 2,2 ,...,P 2,T } Where P1 is the main prediction sequence, P2 is the auxiliary prediction sequence, and T is the length of the time series; it is assumed that the prediction accuracy of the main prediction sequence is higher than that of the auxiliary prediction sequence; The DTW distance between P1 and P2 is calculated by the dynamic time warping method: First calculate the Euclidean distance between any two time points: d(i,j)=dist(P 1,i ,P 2,j )=|P 1,i -P 2,j | Where d(i,j) represents the i-th element P of sequence P1 1,i and the jth element P of sequence P2 2,j The distance between Construct a T×T distance matrix M: Calculate the cumulative distance C through the distance matrix M: C(i,j)=d(i,j)+min{C(i-1,j),C(i,j-1),C(i-1,j-1)} Where C(i,j) represents the minimum cumulative distance from the starting point to position i,j, min{C(i-1,j),C(i,j-1),C(i-1,j-1)} represents the minimum value from the top, left or diagonal direction; the boundary conditions of the cumulative distance are: C(0,0)=d(0,0) The final DTW distance between sequence P1 and sequence P2 is expressed as: d DTW (P1,P2)=C(T,T)。 3. A method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds according to claim 2, characterized in that: The specific implementation method of defining the uncertainty boundary of photovoltaic power prediction is as follows: Assume that the time series of the predicted power sequence of a single node photovoltaic power is; P pred ={P pred,1 ,P pred,2 ,...,P pred,T } Where T is the prediction time domain; The actual output of photovoltaic power will always deviate from the predicted value. The uncertainty set method is used to limit the uncertain variables to a certain range. The upper and lower limits of the fluctuation of photovoltaic power in the uncertainty set, that is, the error band, are defined as α%, then: Among them, P upper represents the upper bound of the uncertain set, P lower Represents the lower bound of the uncertain set.

4. A method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds according to claim 3, characterized in that: The specific implementation of the power out-of-bounds label calculation method is as follows: Assuming that there is historical photovoltaic operation data on any node containing distributed photovoltaics, the photovoltaic power prediction model obtains the historical photovoltaic power main prediction sequence P1 and the historical photovoltaic power actual sequence P0, thereby defining the calculation method of the power out-of-bounds label: Where P 0,i is the ith element of the actual output sequence, P 1,i To predict the i-th element of the output sequence, sign(x) is the sign function: When y=1, it means that the actual power sequence is out of bounds, and when y=0, it means that it is not out of bounds. Obviously, the value of the out-of-bounds label is related to α; When the labels of the training set samples are unevenly distributed, the photovoltaic power prediction model tends to predict the majority class, which makes the photovoltaic power prediction model inaccurate for the minority class. The Latin hypercube sampling method is used to construct the training samples, so that the input variable α is random and the output label y is as uniform as possible. The details are as follows: First calculate the error band boundary values: Determine the sampling interval as Divide the interval into N subintervals of equal length; the length of each subinterval is: Therefore, the i-th subinterval is: Where, i = 1, 2, ..., N; A point is randomly selected in each subinterval and randomly sorted to form a set of values ​​of the input variable α of the training set.

5. A method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds according to claim 4, characterized in that: The specific implementation method of calculating the probability of photovoltaic power exceeding the boundary of a single node through a data-driven logistic regression model is as follows: D DTW , α is used as the input of the model training set, and the sample label y is used as the output of the model training set; The inner layer of the logistic regression model uses a model with polynomial features to capture the nonlinear relationship between input variables and enhance the model's fitting ability. The two-input variable polynomial function used by the model with polynomial features is as follows: Where w1, w2, ..., w5 are the weights of the model, b is the bias term, x1 and x2 are the model input features, i.e., d DTW , α; The outer layer of the logistic regression model uses the sigmoid function to obtain the output probability value: The loss function of the logistic regression model uses the cross entropy function: Where i = 1, 2, …, N, N is the total number of samples in the model. The logistic regression model is trained by historical data, and the DTW distance d of the power sequence is predicted in actual scheduling. DTW The actual power out-of-bounds probability P(y=1) of a single-node photovoltaic is obtained by using α.

6. A method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds according to claim 5, characterized in that: Since the inner layer is a polynomial function, its weight dimension is more than that of a linear function, which is more prone to overfitting problems. Here, the L2 regularization method is used to reduce the weights of unimportant features and enhance the generalization ability of the model. Therefore, the loss function with regularization terms is: Where λ is the regularization term coefficient, which is used to balance the complexity and fitting accuracy of the model.

7. A method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds according to claim 6, characterized in that: Based on Copula function theory, the risk of regional distribution network scheduling containing distributed photovoltaics is evaluated.

8. A method for risk assessment of regional distribution network scheduling containing distributed photovoltaic power considering photovoltaic power out-of-bounds according to claim 7, characterized in that: Based on the Copula function theory, the specific implementation method of evaluating the risk of regional distribution network dispatching containing distributed photovoltaics is as follows: The out-of-bounds situation of the photovoltaic power of a single node is unknown during scheduling, so the out-of-bounds situation Y is a random variable in the scheduling stage. Let x be a binary variable to determine whether the power is out of bounds, then the survival function of the random variable Y is: S(x)=P(Y≥x)x∈{0,1} The two-dimensional Gumbel-Copula function is expressed as follows: Where u1 and u2 are random variables; θ∈(0,1) is the correlation coefficient between random variables u1 and u2; if the object to be described is a multivariate joint distribution function, then the N-dimensional Gumbel-Copula function is: Sklar's theorem shows that for any d-dimensional joint survival function There exists a Copula function C such that the joint survival function is expressed as the marginal survival function S i (x i ) and Copula functions: When the photovoltaic power of any node exceeds the limit, it will affect the scheduling. When the number of distributed photovoltaic access nodes in the regional distribution network is N, the evaluation index function of the scheduling scheme in the region is obtained: R(x)=P(Y1<x,Y2<x,...,Y N (x) Where Y1, Y2, ..., Y N is the power out-of-bounds situation of each distributed photovoltaic node. When x is 1, it means that the power of each node is within the bounds, that is, the dispatch is reliable. Considering the definition of the survival function, the above formula is converted into the following form by the addition formula: Writing the above equation in compact form: According to Sklar theory, there exists an N-dimensional Copula function such that: According to the definition of Gumbel-Copula function, we can get: Therefore, the overall evaluation index of the regional scheduling plan is obtained as follows: The larger the R(x) value is, the lower the probability of power out-of-bounds risk in the corresponding scheduling scheme is.

9. A risk assessment system for regional distribution network scheduling containing distributed photovoltaics considering photovoltaic power out-of-bounds, characterized in that: The method comprises a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps as claimed in any one of claims 1 to 8 can be implemented.

10. A computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, and when the processor executes the computer program instructions, the method steps according to any one of claims 1 to 8 can be implemented.

Images