Power grid key section power probability distribution prediction method based on deep learning

Abstract

The invention discloses a power grid key section power probability distribution prediction method based on deep learning, and the method comprises the steps: representing new energy output and load power probability distribution as effective input features required by a deep learning model, and representing power grid key section power probability distribution as deep learning model output features; secondly, determining non-parametric probability distribution of key section power of the power grid by utilizing the new energy output and load power probability distribution; establishing a sample set containing input and output feature vectors; then, training a deep learning probability distribution prediction model by using the sample set; and finally, inputting the new energy output and load power probability distribution feature vectors corresponding to the power probability distribution of the key section of the power grid to be predicted into the trained deep learning probability distribution prediction model, and outputting the model to obtain the power probability distribution of the key section of the power grid to be predicted. According to the method, the prediction efficiency is improved while the power probability distribution prediction accuracy of the key section of the power grid is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of safe operation and control of power grids, and in particular to a method for predicting power probability distribution of key sections of power grids based on deep learning. Background Art

[0002] Understanding the transmission power limit of each transmission section and ensuring that it can meet the load without exceeding the limit is of great help to the stability and economy of the power system operation. In traditional power grids, the impact of uncertainty factors is relatively small, and deterministic power flow calculations are sufficient to estimate the overall operating status of the system. In modern power systems, due to the substantial increase in the proportion of new energy, the increase in uncertain factors has caused drastic changes in the power flow mode of the power system. The uncertainty of the power flow in each transmission section has greatly increased, and traditional methods are difficult to evaluate whether the system can operate safely and stably. How to obtain the probability distribution of power in each key section while considering the uncertainty of new energy and load fluctuations has become an urgent problem to be solved.

[0003] The basis for solving the power probability distribution of key sections is the uncertainty analysis method represented by the probability tide. In the existing technology, there are three probabilistic power flow algorithms: simulation method, analytical method and approximate method, but there are still some challenges in the actual application of high-proportion renewable energy power systems. The main reasons are: the approximate method and the analytical method have clear concepts and fast calculation speeds, but they often have large errors in systems with high renewable energy penetration rates; the simulation method has high calculation accuracy, but the high computational burden brought by large-scale sampling and deterministic power flow calculations is a pain point that makes it difficult to apply online.

[0004] Therefore, in the relevant technology, there is an urgent need for a method that can improve the accuracy and speed of power probability distribution prediction at key sections of the power grid. Summary of the invention

[0005] Based on this, it is necessary to provide a power probability distribution prediction method for key sections of the power grid based on deep learning, which can improve the accuracy and speed of power probability distribution prediction for key sections of the power grid in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for predicting power probability distribution of key sections of a power grid based on deep learning. The method comprises the following steps:

[0007] S1. Represent the probability distribution of renewable energy output and load power as the effective input features required by the deep learning model, and represent the power probability distribution of key sections of the power grid as the output features of the deep learning model;

[0008] S2. Using the probability distribution of the new energy output and load power, determine the non-parametric probability distribution of power in key sections of the power grid, and establish a sample set containing input and output feature vectors;

[0009] S3, use the sample set to train the deep learning probability distribution prediction model;

[0010] S4. Input the new energy output and load power probability distribution feature vectors corresponding to the power probability distribution of the key section of the power grid to be predicted into the trained deep learning probability distribution prediction model, and the model output obtains the power probability distribution of the key section of the power grid to be predicted.

[0011] Optionally, in one embodiment of the present application, during prediction, if the operating mode of the synchronous generator of the power grid is different from the operating mode used when training the deep learning probability distribution prediction model, the power probability distribution of the key section of the power grid obtained in S4 is corrected.

[0012] Optionally, in one embodiment of the present application, in S1, obtaining the effective input features and output features includes the following method:

[0013] Take M points in the range (0, 1] for the cumulative distribution function of each of the new energy output, load power, and key section power of the power grid, and obtain the power quantile according to the domain value corresponding to the M points, and the dimension is (K r +K l +L)×M feature vector;

[0014] Among them, K r , K l ,L are the number of uncertain new energy,loads and key sections in the power grid; K r ×M,K l ×M are the dimensions of the input feature vectors consisting of the probability distribution of renewable energy output and load power respectively; L×M are the dimensions of the output feature vectors consisting of the power probability distribution of key sections of the power grid.

[0015] Optionally, in one embodiment of the present application, in S2, the method for determining the non-parametric probability distribution of power at a key section of a power grid includes:

[0016] The output of new energy and the probability distribution of load power are used to obtain the key section power through Monte Carlo sampling, and then the non-parametric probability distribution of the key section power of the power grid is obtained through non-parametric estimation.

[0017] Optionally, in one embodiment of the present application, based on the sensitivity perturbation value of the synchronous motor output to the active power of the key section of the power grid, the power probability distribution of the key section of the power grid to be predicted obtained by the model output in S4 is corrected.

[0018] Optionally, in one embodiment of the present application, the sensitivity perturbation value is expressed as:

[0019]

[0020] Among them, P L,l is the active power of the lth target critical section, l=1,2,...,L,P′ L,l It represents the power of the lth key section after the active power of the synchronous generator changes, and ε is the change in the active power of the synchronous generator.

[0021] In a second aspect, the present application also provides a power probability distribution prediction device for key sections of a power grid based on deep learning. The device comprises:

[0022] Data acquisition and processing module, used to obtain the probability distribution of new energy output and load power;

[0023] The input and output feature vector determination module is used to represent the probability distribution of renewable energy output and load power as the effective input features required by the deep learning model, and to represent the power probability distribution of key sections of the power grid as the output features of the deep learning model;

[0024] A power grid key section power non-parametric probability distribution establishment module is used to determine the power grid key section power non-parametric probability distribution by using the new energy output and load power probability distribution, and establish a sample set containing input and output feature vectors;

[0025] A deep learning probability distribution prediction model training module is used to train a deep learning probability distribution prediction model using a sample set containing input and output feature vectors;

[0026] The power probability distribution determination module for key sections of the power grid is used to input the new energy output and load power probability distribution feature vectors corresponding to the power probability distribution of key sections of the power grid to be predicted into the trained deep learning probability distribution prediction model, and the model output obtains the power probability distribution of key sections of the power grid to be predicted.

[0027] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the methods described in the above embodiments.

[0028] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in each of the above embodiments are implemented.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. The present invention determines the non-parametric probability distribution of the power of key sections of the power grid through Monte Carlo simulation for each probability distribution of new energy output and load power, constructs a deep learning feature vector in the form of probability distribution, and establishes a sample set containing input and output feature quantities based on this;

[0031] 2. The present invention can realize nonlinear mapping from the probability distribution of renewable energy output and load power to the probability distribution of power in key sections of the power grid. The deep learning method adopted takes into account a large number of probability distributions of renewable energy output and load power, and has higher calculation accuracy than the probability flow convolution method. The sensitivity calculated based on the perturbation method solves the problem that the operating mode of the synchronous generator of the power grid during prediction is different from the operating mode adopted during deep learning model training, thereby improving the accuracy of the prediction of the power probability distribution in key sections of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is an application environment diagram of a power grid key section power probability distribution prediction method based on deep learning in one embodiment;

[0033] Figure 2 A schematic diagram of a flow chart of a method for predicting power probability distribution of key sections of a power grid based on deep learning in one embodiment;

[0034] Figure 3 A schematic diagram of a sample generation method flow for input and output in the form of probability distribution according to an embodiment;

[0035] Figure 4 A schematic diagram of Monte Carlo simulation sampling in one embodiment;

[0036] Figure 5 A schematic diagram of constructing a probability distribution feature vector based on a cumulative distribution function in one embodiment;

[0037] Figure 6 A schematic diagram of a prediction result of a deep learning probability distribution prediction model in one embodiment;

[0038] Figure 7 A schematic diagram of correcting data results using a sensitivity method in one embodiment;

[0039] Figure 8 is a schematic diagram of an IEEE-118 node system containing a high proportion of new energy in one embodiment;

[0040] Fig. 9 A schematic diagram of a comparison between a deep learning probability distribution prediction model and a DC power flow convolution method in one embodiment;

[0041] Fig.10It is a structural block diagram of a power grid key section power probability distribution prediction device based on deep learning in one embodiment;

[0042] Fig.11 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0044] The present application provides a method for predicting power probability distribution of key sections of a power grid based on deep learning, which can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0045] In one embodiment, Figure 2 As shown in the figure, a power probability distribution prediction method for key sections of power grid based on deep learning is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0046] S201: Obtain the probability distribution of new energy output and load power.

[0047] In the embodiment of the present application, firstly, a large amount of new energy output and load power probability distribution data are collected.

[0048] S203: Represent the probability distribution of renewable energy output and load power as the effective input features required by the deep learning model, and represent the power probability distribution of key sections of the power grid as the output features of the deep learning model.

[0049] In the present application embodiment, Figure 3 As shown, it is a schematic diagram of the sample generation method process with probability distribution input and output. new and Q new are the active power and reactive power of renewable energy output, P load and Qload are the active power and reactive power of the load respectively. After the power flow calculation, the node voltage V bus , phase angle δ bus and the active power P of the key sections of the power grid branch and reactive power Q branch Take M points in the range (0,1] for the cumulative distribution function of each renewable energy output, load power and key section power of the power grid, and obtain the power quantile according to the domain value corresponding to these M points, and the construction dimension is (K r +K l +L)×L feature vector, where K r , K l ,L are the number of uncertain new energy,loads and key sections in the power grid; K r ×M,K l ×M are the dimensions of the input feature vectors composed of the probability distribution of new energy output and load, and L×M is the dimension of the output feature vector composed of the non-parametric probability distribution of power in key sections of the power grid. Then, a feature vector with the dimension of S×(K r +K l )×M input feature sample set, and the output feature sample set with the dimension of S×L×M. Figure 5 As shown, it is a schematic diagram of constructing a probability distribution feature vector based on the cumulative distribution function.

[0050] S205: Using the new energy output and load power probability distribution, determine the non-parametric probability distribution of power at key sections of the power grid, and establish a sample set containing input and output feature vectors.

[0051] In the embodiment of the present application, for each new energy output and load power probability distribution, the non-parametric probability distribution of the power of the key section of the power grid is simulated and established by the Monte Carlo method. Figure 4 The figure shows a schematic diagram of Monte Carlo simulation sampling.

[0052] In one embodiment of the present application, it is assumed that the number of new energy sources and loads with design and uncertainty is K, and its cumulative distribution function is Y k =F k (X k ), k = 1, 2, ..., K, divide the range of each cumulative distribution function [0, 1] into N equally spaced intervals, sample each interval, and select the midpoint y of each interval kn As Y k The sampling value of X is calculated based on the inverse function of the cumulative distribution function. k The nth sample value of is as follows:

[0053]

[0054] Among them, xkn For the sampled renewable energy output and load power, N samplings of K input random variables constitute an initial sampling matrix with a dimension of K×N, which is randomly arranged to form a sampling matrix.

[0055] According to the renewable energy output and load power in each column of the sampling matrix, the power of L key sections of the power grid is obtained through power flow calculation. After N power flow calculations, a key section power matrix with a dimension of L×N is formed. Non-parametric estimation is performed to obtain the power probability distribution of L key sections of the power grid.

[0056] S207: Use the sample set to train a deep learning probability distribution prediction model.

[0057] S209: Input the new energy output and load power probability distribution feature vectors corresponding to the power probability distribution of the key section of the power grid to be predicted into the trained deep learning probability distribution prediction model, and the model output obtains the power probability distribution of the key section of the target power grid to be predicted.

[0058] In an embodiment of the present application, the sample set of input and output feature vectors is divided into a training sample set and a test sample set, the deep learning probability distribution prediction model is trained based on the training sample set, and the deep learning probability distribution prediction model is tested based on the test sample set.

[0059] The training sample set is input into the deep learning probability distribution prediction model, the deep learning probability distribution prediction model is continuously optimized using the loss function and the activation function, the training of the deep learning probability distribution prediction model is completed, and the testing of the deep learning probability distribution model is completed based on the test sample set.

[0060] The mean square error function is selected as the loss function, and the specific formula is as follows.

[0061]

[0062] Where d is y k and The dimension of y k is the true value, that is, the output feature vector, It is the output of the deep learning probability distribution prediction model, that is, the initial output prediction result.

[0063] The activation function of the model is the ReLU function. ReLU is a piecewise function. When the horizontal axis is in the negative half axis, the output is 0, and when the horizontal axis is in the positive half axis, the output is equal to the horizontal axis value. When ReLU is used as the activation function, when the neuron is activated in the positive half axis, the gradient disappearance phenomenon will not occur. The specific formula is as follows.

[0064]

[0065] Here, x represents the horizontal axis.

[0066] In one embodiment of the present application, the method further includes:

[0067] Based on the sensitivity perturbation value of synchronous motor output to the active power of key sections of power grid, the power probability distribution of key sections of power grid to be predicted obtained by model output is corrected.

[0068] In one embodiment of the present application, after the model training is completed, the operation mode of the synchronous generator of the power grid may be different from the operation mode used during the training. Therefore, based on the sensitivity perturbation value of the synchronous motor output to the active power of the key section of the power grid, the probability distribution of the power of the key section of the power grid to be predicted obtained by the model output is corrected. Based on the operation mode of the power grid when the model is applied, the power flow calculation is performed to obtain the power of L key sections.

[0069] Apply a change in the output of the i-th synchronous generator (active power changes):

[0070] P′ G,i =P G,i +ε

[0071] Among them, P G,i is the active output of the i-th generator, i=1,2,...,I, I is the total number of synchronous generators, ε is the change in the active power of the synchronous generator. Perform power flow calculation to obtain the power of L key sections after the active power of the synchronous generator changes.

[0072] In one embodiment of the present application, the sensitivity perturbation value of the synchronous motor output to the active power of the key section of the power grid can be expressed as:

[0073]

[0074] Among them, P L,l is the active power of the lth target critical section, l=1,2,...,L,P′ L,l It represents the power of the lth key section after the active power of the synchronous generator changes, and ε is the change in the active power of the synchronous generator.

[0075] Apply the change amount to each synchronous generator output, calculate the change of power of each key section of the power grid before and after the change, and obtain the sensitivity matrix λ with a dimension of L×I:

[0076]

[0077] The power probability distribution of key sections of the power grid obtained by the deep learning probability distribution prediction model is corrected according to the sensitivity matrix. The specific correction formula is shown below.

[0078] P′ branch =P branch +λΔP

[0079] Among them, P branch is the power probability distribution of the key sections of the power grid obtained by the deep learning probability distribution prediction model, and ΔP is the difference between the output of the synchronous generator during application and training. Figure 6 , 7 As shown in the figure, they are schematic diagrams of the prediction results of the deep learning probability distribution prediction model and the sensitivity correction data results calculated based on the perturbation method. Figure 6 , Figure 7 It can be seen that based on the power prediction interval of the key section of the power grid with a confidence level of 95%, the actual power fall rate of the key section of the power grid is higher after sensitivity correction.

[0080]

[0081] In this embodiment, the power probability distribution of the key section of the power grid is corrected by using the sensitivity calculated by the perturbation method, thereby solving the problem that the operation mode of the synchronous generator of the power grid is different from the operation mode adopted during the training of the deep learning probability distribution prediction model.

[0082] In one embodiment of the present application, Figure 8 As shown in the figure, it is a schematic diagram of the IEEE-118 node system with a high proportion of renewable energy, in which the traditional units of nodes 59 and 61 are replaced by photovoltaic power stations with equal capacity, and the traditional units of nodes 25, 26, 49, 65, 66, 100, 103, and 111 are replaced by wind turbines with equal capacity, and the renewable energy penetration rate is 49.44%. The improved IEEE118 node system with renewable energy access, complete topology, generates a total of 10,000 samples, of which 80% are used as training sample sets and 20% are used as test sample sets. The AC-based Monte Carlo (ACMC) simulation method based on AC power flow is used as a benchmark. The analysis method based on DC power flow and convolution (DC-Conv), the critical section power probability distribution calculation method based on multi-layer perceptron (MLP), the critical section power probability distribution calculation method based on convolutional neural network (CNN), the critical section power probability distribution calculation method based on long short-term memory network (LSTM), and the critical section power probability distribution calculation method based on graph convolution network (GCN) are compared. The comparison results are shown in the figure. Fig. 9 As shown in the following table and Fig. 9 It can be seen that the method based on deep learning probability distribution prediction model has higher calculation accuracy than the probabilistic power flow convolution method in high-proportion renewable energy power systems.

[0083]

[0084] Based on the same inventive concept, the embodiment of the present application also provides a power probability distribution prediction device for a key section of a power grid based on deep learning, which is used to implement the above-mentioned power probability distribution prediction method for a key section of a power grid based on deep learning. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the power probability distribution prediction device for a key section of a power grid based on deep learning provided below can be referred to the limitations of the power probability distribution prediction method for a key section of a power grid based on deep learning above, and will not be repeated here.

[0085] In one embodiment, Fig.10 As shown, a power grid key section power probability distribution prediction device 1000 based on deep learning is provided, comprising: a data acquisition processing module 1001, an input and output feature vector determination module 1003, a power grid key section power non-parametric probability distribution establishment module 1005, a deep learning probability distribution prediction model training module 1007 and a power grid key section power probability distribution determination module 1009, wherein:

[0086] The data acquisition processing module 1001 is used to obtain the probability distribution of new energy output and load power.

[0087] The input and output feature vector determination module 1003 is used to represent the probability distribution of new energy output and load power as the effective input features required by the deep learning model, and to represent the non-parametric probability distribution of power at key sections of the power grid as the output features of the deep learning model.

[0088] The power grid key section power non-parametric probability distribution establishment module 1005 is used to determine the power grid key section power non-parametric probability distribution by using the new energy output and load power probability distribution, and establish a sample set including input and output feature vectors.

[0089] The deep learning probability distribution prediction model training module 1007 is used to train the deep learning probability distribution prediction model using a sample set containing input and output feature vectors.

[0090] The power probability distribution determination module 1009 for key sections of the power grid is used to input the new energy output and load power probability distribution feature vectors corresponding to the power probability distribution of key sections of the power grid to be predicted into the trained deep learning probability distribution prediction model, and the model output obtains the power probability distribution of key sections of the power grid to be predicted.

[0091] In one embodiment of the present application, the data acquisition processing module is further used to:

[0092] Take M points in the range (0, 1] for the cumulative distribution function of each of the new energy output, load power, and key section power of the power grid, and obtain the power quantile according to the domain value corresponding to the M points, and the dimension is (K r +K l +L)×M feature vector;

[0093] Among them, K r ×M,K l ×L are the dimensions of the input feature vectors consisting of the probability distribution of renewable energy output and load power respectively; L×M are the dimensions of the output feature vectors consisting of the probability distribution of power in key sections of the power grid.

[0094] In one embodiment of the present application, the power grid key section power non-parametric probability distribution establishment module is also used for:

[0095] The output of new energy and the probability distribution of load power are used to obtain the power of key sections of the power grid through sampling by Monte Carlo method, and then the non-parametric probability distribution of the power of key sections of the power grid is obtained through non-parametric estimation.

[0096] In one embodiment of the present application, the deep learning probability distribution prediction model training module is also used to:

[0097] Dividing the sample set containing the input and output feature vectors into a training sample set and a test sample set;

[0098] The training sample set is input into the deep learning probability distribution prediction model to obtain an initial output prediction result.

[0099] The power grid key section power probability distribution prediction device based on deep learning also includes a power grid key section power probability distribution correction module. In one embodiment of the present application, the power grid key section power probability distribution correction module is used to:

[0100] During prediction, if the operation mode of the synchronous generator in the power grid is different from the operation mode used when training the deep learning probability distribution prediction model, based on the sensitivity perturbation value of the synchronous motor output to the active power of the key section of the power grid, the power probability distribution of the key section to be predicted obtained by the model output in S4 is corrected.

[0101] In one embodiment of the present application, the sensitivity can be expressed as:

[0102]

[0103] Among them, P L,l is the active power of the lth target critical section, l=1,2,...,L,P′ L,lIt represents the power of the lth key section after the active power of the synchronous generator changes, and ε is the change in the active power of the synchronous generator.

[0104] Each module in the above-mentioned power grid key section power probability distribution prediction device based on deep learning can be implemented in whole or in part by software, hardware and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0105] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.11 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a power probability distribution prediction method for key sections of a power grid based on deep learning is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse.

[0106] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0107] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0108] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0109] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0110] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0111] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0112] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0113] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for predicting power probability distribution of key sections of a power grid based on deep learning, characterized in that: The method comprises: S1. Represent the probability distribution of renewable energy output and load power as the effective input features required by the deep learning model, and represent the power probability distribution of key sections of the power grid as the output features of the deep learning model; S2. Using the probability distribution of the new energy output and load power, determine the non-parametric probability distribution of power in key sections of the power grid, and establish a sample set containing input and output feature vectors; S3, use the sample set to train the deep learning probability distribution prediction model; S4. Input the new energy output and load power probability distribution feature vectors corresponding to the power probability distribution of the key section of the power grid to be predicted into the trained deep learning probability distribution prediction model, and the model output obtains the power probability distribution of the key section of the power grid to be predicted.

2. According to a method for predicting power probability distribution of key sections of a power grid based on deep learning according to claim 1, it is characterized in that: During prediction, if the operation mode of the synchronous generator of the power grid is different from the operation mode adopted when training the deep learning probability distribution prediction model, the power probability distribution of the key section of the power grid obtained by S4 is corrected.

3. According to a method for predicting power probability distribution of key sections of a power grid based on deep learning according to claim 1, it is characterized in that: In S1, the effective input features and output features are obtained. Includes the following methods: Take M points in the range (0, 1] for the cumulative distribution function of each of the new energy output, load power, and key section power of the power grid, and obtain the power quantile according to the domain value corresponding to the M points, and the dimension is (K r +K l +L)×M feature vector; Among them, K r , K l ,L are the number of uncertain new energy,loads and key sections in the power grid; K r ×M,K l ×M are the dimensions of the input feature vectors consisting of the probability distribution of renewable energy output and load power respectively; L×M are the dimensions of the output feature vectors consisting of the power probability distribution of key sections of the power grid.

4. According to a method for predicting power probability distribution of key sections of a power grid based on deep learning according to claim 1, it is characterized in that: In S2, the method for determining the non-parametric probability distribution of power at a key section of a power grid includes: The output of new energy and the probability distribution of load power are used to obtain the key section power through Monte Carlo sampling, and then the non-parametric probability distribution of the key section power of the power grid is obtained through non-parametric estimation.

5. According to a method for predicting power probability distribution of key sections of a power grid based on deep learning according to claim 2, it is characterized in that: Based on the sensitivity perturbation value of the synchronous motor output to the active power of the key section of the power grid, the power probability distribution of the key section of the power grid to be predicted obtained by the model output in S4 is corrected.

6. The method for predicting power probability distribution of key sections of a power grid based on deep learning according to claim 5 is characterized in that: The sensitivity perturbation value is expressed as: Among them, P L,l is the active power of the lth target critical section, l=1,2,...,L,P′ L,l is the power of the lth key section after the active power of the synchronous generator changes, and ε is the change in the active power of the synchronous generator.

7. A power grid key section power probability distribution prediction device based on deep learning, characterized in that: The device comprises: Data acquisition and processing module, used to obtain the probability distribution of new energy output and load power; The input and output feature vector determination module is used to represent the probability distribution of renewable energy output and load power as the effective input features required by the deep learning model, and to represent the power probability distribution of key sections of the power grid as the output features of the deep learning model; A power grid key section power non-parametric probability distribution establishment module is used to determine the power grid key section power non-parametric probability distribution by using the new energy output and load power probability distribution, and establish a sample set containing input and output feature vectors; A deep learning probability distribution prediction model training module is used to train a deep learning probability distribution prediction model using a sample set containing input and output feature vectors; The power probability distribution determination module for key sections of the power grid is used to input the new energy output and load power probability distribution feature vectors corresponding to the power probability distribution of key sections of the power grid to be predicted into the trained deep learning probability distribution prediction model, and the model output obtains the power probability distribution of key sections of the power grid to be predicted.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images