Prefecture-level photovoltaic power generation power prediction method, system, equipment and medium

Abstract

The invention discloses a prefecture-level photovoltaic power generation power prediction method, system and device and a medium, and relates to the field of new energy. The method comprises the following steps: firstly, acquiring historical power data of photovoltaic power stations with different voltage grades from a D5000 system, a power distribution automation system and a use and acquisition system, wherein the voltage grades of 380V to 35kV are covered; acquiring grid numerical weather forecast historical data from an electric power meteorological data system; then constructing an initial input variable data set; preprocessing the multi-source data in the initial input variable data set, and constructing a spatial-temporal characteristic quantity based on a dynamic graph based on the preprocessed data set; carrying out the dimension reduction of the spatial-temporal characteristic quantity based on Tucker decomposition, taking the characteristic quantity after dimension reduction as the input of an LSTM network, taking the sum of the power of all photovoltaic power stations as the output, training the LSTM network, and obtaining a photovoltaic power generation power prediction model after the training is completed; the photovoltaic power generation power prediction model is adopted to predict the photovoltaic power generation power of the prefecture-level target area, so that the photovoltaic power generation power prediction precision can be remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of new energy technologies, and particularly to a method, system, device, and medium for predicting photovoltaic power at the prefecture-level city. Background Art

[0002] With the rapid development of social economy, a series of problems such as global resource shortage, climate warming, and environmental pollution have become increasingly prominent. Photovoltaic power generation is transitioning from a supplementary energy source to an alternative energy source. The photovoltaic power has the characteristics of randomness, periodicity, and volatility. When it is connected to the grid at various voltage levels, it brings many challenges to the safe and economic operation of power grids at all levels. To improve the photovoltaic power consumption capacity of power grids at all levels, reasonably arrange the dispatching plan, and manage photovoltaic power stations, the prediction and operation monitoring of photovoltaic power generation have received great attention in the power industry.

[0003] Currently, most of the photovoltaic power predictions at the prefecture-level city are for centralized photovoltaic power stations connected to the grid at 10 kV and above, without considering distributed photovoltaic power stations such as 380 V. This results in an actually smaller predicted value of photovoltaic power, and with the increase in the installed capacity of distributed photovoltaics, the prediction deviation becomes larger. In addition, the resolution of meteorological data used in the prediction process is insufficient, resulting in the need to improve the prediction accuracy of solar radiation and photovoltaic output. Summary of the Invention

[0004] The purpose of the present application is to provide a method, system, device, and medium for predicting photovoltaic power at the prefecture-level city to improve the prediction accuracy of photovoltaic power generation.

[0005] To achieve the above purpose, the present application provides the following solutions.

[0006] In the first aspect, the present application provides a method for predicting photovoltaic power at the prefecture-level city, including:

[0007] Obtaining historical power data of photovoltaic power stations at different voltage levels from the D5000 system, the distribution automation system, and the power consumption information acquisition system; the different voltage levels cover the voltage levels from 380 V to 35 kV;

[0008] Obtaining historical data of grid-based numerical weather forecasts from the power meteorological data system;

[0009] Constructing an initial input variable data set based on the historical power data of photovoltaic power stations and the historical data of grid-based numerical weather forecasts;

[0010] Preprocessing the multi-source data in the initial input variable data set to obtain a preprocessed data set;

[0011] Constructing spatio-temporal feature quantities based on the dynamic graph based on the preprocessed data set;

[0012] Reduce the dimensionality of spatio-temporal feature quantities based on Tucker decomposition to obtain the feature quantities after dimensionality reduction;

[0013] Use the feature quantities after dimensionality reduction as the input of the LSTM network, and use the sum of the powers of all photovoltaic power stations as the output of the LSTM network to train the LSTM network. After the training is completed, a photovoltaic power prediction model is obtained;

[0014] Use the photovoltaic power prediction model to predict the photovoltaic power of the target area at the prefecture-level city.

[0015] Optionally, the obtaining of the historical power data of photovoltaic power stations with different voltage levels from the D5000 system, the distribution automation system, and the user collection system specifically includes:

[0016] Obtain the historical power data of centralized photovoltaic power stations above 10 kV from the D5000 system;

[0017] Obtain the historical power data of 10 kV distributed photovoltaic power stations from the distribution automation system;

[0018] Obtain the historical power data of 380 V distributed photovoltaic power stations from the user collection system.

[0019] Optionally, the obtaining of the historical data of grid-based numerical weather forecasts from the power meteorological data system specifically includes:

[0020] Obtain b different types of historical data of grid-based numerical weather forecasts from the power meteorological data system, including historical data of irradiance values, temperature, humidity, wind speed, wind direction, air pressure, and precipitation at different longitude and latitude positions.

[0021] Optionally, the construction of the initial input variable data set based on the historical power data of photovoltaic power stations and the historical data of grid-based numerical weather forecasts specifically includes:

[0022] Construct an initial input variable data set D, which contains d feature quantities and T time series; d = M × a = M × (b + 1); M is the number of photovoltaic power stations; a is all the variable types obtained, including b different types of historical data of grid-based numerical weather forecasts and 1 type of historical power data of photovoltaic power stations;

[0023] Construct an output variable data set Y, which contains 1 variable and T time series; where 1 variable is the sum of the powers of all photovoltaic power stations.

[0024] Optionally, the preprocessing of the multi-source data in the initial input variable data set to obtain the preprocessed data set specifically includes:

[0025] Perform one or more preprocessings on the multi-source data in the initial input variable dataset D, including data cleaning, abnormal data detection, data conversion, data dimensionality reduction, data classification and encoding, data difference and aggregation, association analysis, and data standardization, to obtain the preprocessed dataset D p ; D p contains d p feature quantities and T time series; where d p = M * a', and a' is the remaining variable type after preprocessing

[0026] Optionally, constructing spatio-temporal feature quantities based on the dynamic graph from the preprocessed dataset specifically includes:

[0027] Denote the value of the k-th variable type of the i-th distributed photovoltaic power station at time t in the preprocessed dataset D p as x k,i,t , and use the formula to calculate the spatio-temporal feature x l ; where t' is the specified time window length; is the mean value of the k-th variable type of the i-th photovoltaic power station within the time window length of t'; x k,j,t is the value of the k-th variable type of the j-th photovoltaic power station at time t; is the mean value of the k-th variable type of the j-th distributed photovoltaic power station within the time window length of t'; k ∈ (1, 2,..., a'); i, j ∈ (1, 2,..., M);

[0028] Take the set of all spatio-temporal features x l as the spatio-temporal feature quantity X loc ; The dimension of the spatio-temporal feature quantity X loc is M × M × a' dimensions

[0029] Optionally, perform dimensionality reduction on the spatio-temporal feature quantity based on Tucker decomposition to obtain the dimensionality-reduced feature quantity, specifically including:

[0030] Based on the Tucker decomposition formula X loc = G × 1 A × 2 B × 3 C, split the spatio-temporal feature quantity X loc into a core tensor matrix G and three factor matrices A, B, and C; where × n , n = 1, 2, 3 is the Tucker decomposition symbol; the dimension of the factor matrix A is e × M, the dimension of the factor matrix B is e × M, the dimension of the factor matrix C is e × a', and the dimension of the core tensor matrix G is e 3 ; e is a constant term

[0031] Take the set of all elements in matrices G, A, B, and C as the feature quantity X' after dimensionality reduction loc .

[0032] In a second aspect, the present application also provides a prefecture-level photovoltaic power prediction system, including:

[0033] A power data acquisition module, configured to acquire historical power data of photovoltaic power stations at different voltage levels from the D5000 system, the distribution automation system, and the power consumption information acquisition system; the different voltage levels cover the voltage levels from 380V to 35kV;

[0034] A meteorological data acquisition module, configured to acquire historical data of grid-based numerical weather forecasts from the power meteorological data system;

[0035] A dataset construction module, configured to construct an initial input variable dataset based on the historical power data of photovoltaic power stations and the historical data of grid-based numerical weather forecasts;

[0036] A data preprocessing module, configured to preprocess the multi-source data in the initial input variable dataset to obtain a preprocessed dataset;

[0037] A spatio-temporal feature construction module, configured to construct spatio-temporal feature quantities based on a dynamic graph based on the preprocessed dataset;

[0038] A Tucker decomposition module, configured to perform dimensionality reduction on the spatio-temporal feature quantities based on Tucker decomposition to obtain feature quantities after dimensionality reduction;

[0039] A prediction model training module, configured to use the feature quantities after dimensionality reduction as the input of the LSTM network, and use the sum of the powers of all photovoltaic power stations as the output of the LSTM network to train the LSTM network, and obtain a photovoltaic power prediction model after the training is completed;

[0040] A power prediction module, configured to use the photovoltaic power prediction model to predict the photovoltaic power of the prefecture-level target area.

[0041] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the prefecture-level photovoltaic power prediction method.

[0042] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the prefecture-level photovoltaic power prediction method is implemented.

[0043] According to the specific embodiments provided by the present application, the following technical effects are disclosed in the present application.

[0044] A method, system, device, and medium for predicting photovoltaic power at the prefecture-level provided by this application adopt key technologies such as multi-source data preprocessing, numerical weather forecast output correction, and multi-data condition prediction. It can utilize historical data of grid-based numerical weather forecasts for all photovoltaic power generation resources at the prefecture-level to achieve multi-time scale prediction of photovoltaic power, effectively improving the accuracy of photovoltaic power prediction and enhancing the operation and control level of the regional power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is a flowchart of a method for predicting photovoltaic power at the prefecture-level in this application;

[0047] Figure 2 It is a principle block diagram of a method for predicting photovoltaic power at the prefecture-level in this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only some embodiments of this application, rather than all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of this application.

[0049] To improve the accuracy of photovoltaic power prediction at the prefecture-level and the application level of the system, and meet the requirements of stable operation and precise control of the regional power grid, this application proposes a method, system, device, and medium for predicting photovoltaic power at the prefecture-level. Based on key technologies such as multi-source data preprocessing, numerical weather forecast output correction, and prediction methods under multi-data conditions, the overall architecture and functional design of photovoltaic power prediction at the prefecture-level are proposed. The application of this application can effectively improve the accuracy of photovoltaic power prediction and enhance the operation and control level of the regional power grid.

[0050] To make the above objects, features, and advantages of this application more obvious and understandable, the following will further elaborate on this application in conjunction with the drawings and specific embodiments.

[0051] In an exemplary embodiment, this application provides a method for predicting photovoltaic power at the prefecture-level. Its flowchart is as Figure 1 shown, including the following steps 1 to 8; its principle block diagram is as Figure 2 shown.

[0052] Step 1: Obtain historical power data of photovoltaic power stations with different voltage levels from the D5000 system, the distribution automation system, and the power consumption information acquisition system; the different voltage levels cover the voltage levels from 380V to 35kV.

[0053] The photovoltaic power prediction method for the prefecture-level range is different from the new energy modules of the in-station monitoring system of photovoltaic power stations and the main network dispatching. There are differences in data sources, business functions, etc., which are particularly prominent in the following two aspects: 1) The access voltage levels cover 380V to 35kV, there are many business attribution departments, and the operation monitoring of new energy involves data from multiple business systems, which are placed and managed dispersedly; 2) The accuracy and resolution of numerical weather forecasts for the provincial-level range are limited, resulting in the need to improve the prediction accuracy of solar radiation and photovoltaic output.

[0054] Most of the existing photovoltaic power prediction systems are mainly for centralized photovoltaic power stations connected to the grid at 10kV and above, without considering distributed photovoltaic power stations such as 380V, resulting in the actual predicted value of photovoltaic power being on the small side, and with the increase of the installed capacity of distributed photovoltaics, the prediction deviation is greater. To solve the problem of incomplete photovoltaic power data designed by the existing photovoltaic power prediction system, this application obtains historical power data of photovoltaic power stations with different voltage levels from the D5000 system, the distribution automation system, and the power consumption information acquisition system; the different voltage levels cover the voltage levels from 380V to 35kV.

[0055] Among them, the D5000 system is an important technical support system for power grid operation control and dispatching production management. The D5000 system has functions such as data collection and exchange, and steady-state monitoring of power grid operation, and mainly provides power data information of centralized photovoltaic power stations at 10kV and above in this application.

[0056] The distribution automation system (DAS) is an automation system that enables distribution enterprises to monitor, coordinate, and operate distribution equipment in real time from a distance. The distribution automation system collects real-time operation data of distribution equipment, conducts status monitoring, fault diagnosis, and early warning, and realizes the automated management and control of the distribution network. In this application, it mainly provides power data information of 10kV distributed photovoltaic power stations.

[0057] The power consumption information acquisition system, that is, the power user electricity information acquisition system, is mainly used for collecting, processing, and storing the electricity consumption information of power users. The power consumption information acquisition system collects the electricity consumption data of users in real time or regularly through the acquisition equipment installed at the user end, such as voltage, current, power, electricity consumption, etc., and transmits these data to the main station system through the communication network for processing and analysis. In this application, it mainly provides power data information of 380V distributed photovoltaic power stations.

[0058] This application obtains the historical power data of centralized photovoltaic power stations above 10 kV from the D5000 system, the historical power data of 10 kV distributed photovoltaic power stations from the distribution automation system, and the historical power data of 380 V distributed photovoltaic power stations from the power consumption and collection system, jointly constituting the historical power data of photovoltaic power stations with different voltage levels.

[0059] Step 2: Obtain the historical data of grid-based numerical weather forecasts from the power meteorological data system.

[0060] The numerical weather forecasts adopted by existing photovoltaic power prediction systems only use single data at the prefecture-level city, without considering the meteorological differences at different longitudes and latitudes within the same prefecture-level city, resulting in the need to improve the prediction accuracy of solar radiation and photovoltaic output. To solve the problem of insufficient resolution of the meteorological data used by existing photovoltaic power prediction systems, this application obtains the historical data of grid-based numerical weather forecasts from the power meteorological data system, which can effectively improve the resolution of meteorological data and thus enhance the prediction accuracy of photovoltaic power.

[0061] Among them, grid-based numerical weather forecast refers to dividing the forecast area into multiple small grids and conducting refined numerical weather forecasts for each grid. This method can capture smaller details of weather changes and improve the accuracy and precision of the forecast. Specifically, imagine the forecast area as a grid composed of multiple squares with side lengths of several kilometers (such as 5 kilometers), and achieve precise spatial forecasts for these small square areas. Grid-based numerical weather forecasts are superior to ordinary meteorological data in terms of spatial resolution, time accuracy, forecast elements, technical basis, and service applications.

[0062] This application obtains b different types of historical data of grid-based numerical weather forecasts from the power meteorological data system, including historical data of irradiance values, temperature, humidity, wind speed, wind direction, air pressure, and precipitation at different longitude and latitude positions (different geographical locations).

[0063] Step 3: Construct an initial input variable data set based on the historical power data of photovoltaic power stations and the historical data of grid-based numerical weather forecasts.

[0064] Based on the obtained multi-source historical data (including the historical power data of photovoltaic power stations and the historical data of grid-based numerical weather forecasts), construct an initial input variable data set D, which contains d characteristic quantities and T time series. The T time series are divided according to the acquisition time of the obtained historical data. Among them,

[0065] d = M × a = M × (b + 1) (1)

[0066] M is the number of photovoltaic power stations within the studied prefecture-level region. a is all the variable types obtained, including b different types of gridded numerical weather forecast historical data (abbreviated as meteorological data) and 1 type of photovoltaic power station historical power data (abbreviated as power data).

[0067] For subsequent model training, it is also necessary to construct an output variable dataset Y, which contains 1 variable (i.e., the sum of the powers of all photovoltaic power stations y) and T time series.

[0068] Step 4: Preprocess the multi-source data in the initial input variable dataset to obtain a preprocessed dataset.

[0069] The preprocessing of the multi-source data in the initial input variable dataset D may include: one or more preprocessing methods such as data cleaning, abnormal data detection, data transformation, data dimensionality reduction, data classification and encoding, data difference and aggregation, association analysis, and data standardization to obtain the preprocessed dataset D p ; D p contains d p feature quantities and T time series. Among them,

[0070] d p = M * a' (2)

[0071] a' is the remaining variable type after preprocessing.

[0072] In an exemplary embodiment, the preprocessing process of the multi-source data in D includes: 4.1) Detect missing (abnormal) data based on the K-means algorithm; 4.2) Interpolate and fill the missing (abnormal) data based on the KNN algorithm; 4.3) Perform min-max standardization processing on the photovoltaic power station power data and meteorological data; 4.4) Extract the main variable types through the PCA algorithm to obtain a new dataset D p , containing d p feature quantities and T time series, where d p = M * a', a' is the variable type after PCA dimensionality reduction, and M is the number of distributed photovoltaic power stations.

[0073] Step 5: Construct spatio-temporal feature quantities based on the dynamic graph based on the preprocessed dataset.

[0074] For a' variable types, construct spatio-temporal feature quantities X loc . Denote the preprocessed dataset D p in which the value of the k-th variable type of the i-th distributed photovoltaic power station at time t is x k,i,t , and the value of the k-th variable type of the j-th photovoltaic power station at time t is x k,j,t , then there is the formula:

[0075]

[0076] where t' is the specified time window length; is the mean value of the k-th variable type of the i-th photovoltaic power station within the time window length of t'; is the mean value of the k-th variable type of the j-th distributed photovoltaic power station within the time window length of t'; k ∈ (1, 2,..., a'); i, j ∈ (1, 2,..., M); x l is the spatio-temporal feature based on the dynamic graph, l ∈ (1, 2,..., M×M×a').

[0077] Take the set of all spatio-temporal features x l as the spatio-temporal feature quantity X loc , then the dimension of the spatio-temporal feature quantity X loc is M×M×a' - dimensional, that is, a three - dimensional matrix.

[0078] Step 6: Perform dimensionality reduction on the spatio-temporal feature quantity based on Tucker decomposition to obtain the dimensionality - reduced feature quantity.

[0079] Based on Tucker decomposition, split the spatio-temporal feature quantity X loc into a core tensor matrix G and three factor matrices A, B, and C. The Tucker decomposition can be expressed as:

[0080] X loc = G× 1 A× 2 B× 3 C (4)

[0081] where × n , n = 1, 2, 3 is the Tucker decomposition symbol. The dimension of factor matrix A is e×M, the dimension of factor matrix B is e×M, the dimension of factor matrix C is e×a', and the sum of the dimensions of the three factor matrices is ea' + 2eM; e is a constant term. The dimension of the core tensor matrix G is e 3 .

[0082] Take the set of all elements in matrices G, A, B, and C as the dimensionality - reduced feature quantity X' loc , then the dimension of X' loc is e 3 + ea' + 2eM.

[0083] Step 7: Use the dimensionality - reduced feature quantity as the input of the LSTM network, and use the sum of the powers of all photovoltaic power stations as the output of the LSTM network to train the LSTM network. After training, obtain the photovoltaic power prediction model.

[0084] Use the dimensionality - reduced feature quantity X'loc Input into the LSTM network, and the output of the LSTM network is the corresponding power data in the output variable dataset Y (i.e., the sum of the powers of all photovoltaic power stations y). Initialize the parameters of the LSTM network and use the processed data X' loc and the output variable dataset Y to train the LSTM network.

[0085] The LSTM network adds a forget gate, an input gate, a cell state, and an output gate on the basis of the RNN network. The forget gate can be expressed as:

[0086] f t = σ(W f * [h t-1 , x t + b f ) (5)

[0087] where f t is the output of the forget gate at the current time step; σ represents the Sigmoid activation function; W f represents the weight matrix of the forget gate; * is the convolution operation; h t-1 represents the hidden state at the previous time step; x t represents the input at the current time step, which is the sample data at time t corresponding to X' loc ; [h t-1 , x t represents the concatenated input; b f is the bias term of the forget gate.

[0088] The input gate can be expressed as:

[0089] i t = σ(W i * [h t-1 , x t + b i ) (6)

[0090]

[0091] where i t represents the input gate at the current time step; W i represents the weight matrix of the input gate, and b i is the bias term of the input gate. represents the candidate memory cell, and W c and b c are the weight matrix and bias term of the candidate memory cell respectively.

[0092] The LSTM updates the old memory cells through the forget gate and the input gate to obtain new memory cells, which can be expressed as:

[0093]

[0094] Among them, c t-1 represents the old memory cell, and c t represents the updated memory cell.

[0095] The output gate at the current time step can be expressed as:

[0096] o t = σ(W o * [h t-1 , x t + b o ) (9)

[0097] Among them, W o and b o represent the weight matrix and bias term of the output gate respectively, both of which are constants.

[0098] The final output result can be expressed as:

[0099] y t = o t * tanh(c t ) (10)

[0100] y t is the value of the predicted power data y at time t.

[0101] After the LSTM network training is completed, the photovoltaic power prediction model of this application can be obtained.

[0102] Step 8: Use the photovoltaic power prediction model to predict the photovoltaic power of the prefecture-level target area.

[0103] When predicting the photovoltaic power of the prefecture-level target area, first obtain the current power data of the photovoltaic power stations within the prefecture-level target area and the grid-based numerical weather forecast data, then construct the spatio-temporal feature quantities based on the dynamic graph according to the foregoing method, and perform dimensionality reduction based on Tucker decomposition to obtain the X' loc data at the current moment, and then input the X' loc data at the current moment into the trained photovoltaic power prediction model to obtain the photovoltaic power prediction value.

[0104] Furthermore, by adjusting the time window t', power prediction on multiple time scales (such as ultra-short term, short term) can also be achieved.

[0105] This application adopts key technologies such as multi-source data preprocessing, numerical weather prediction output correction, and multi-data condition prediction, and establishes a method for predicting photovoltaic power at the municipal level. It can utilize grid-based meteorological data for all photovoltaic power generation resources at the municipal level to achieve multi-time scale prediction of photovoltaic power, effectively improving the accuracy of photovoltaic power prediction and enhancing the operation and control level of the regional power grid.

[0106] In an exemplary embodiment, this application also provides a system for predicting photovoltaic power at the municipal level, including:

[0107] A power data acquisition module, configured to acquire historical power data of photovoltaic power stations at different voltage levels from the D5000 system, the distribution automation system, and the power consumption acquisition system; the different voltage levels cover the voltage levels from 380V to 35kV;

[0108] A meteorological data acquisition module, configured to acquire historical grid-based numerical weather prediction data from the power meteorological data system;

[0109] A dataset construction module, configured to construct an initial input variable dataset based on the historical power data of photovoltaic power stations and the historical grid-based numerical weather prediction data;

[0110] A data preprocessing module, configured to preprocess the multi-source data in the initial input variable dataset to obtain a preprocessed dataset;

[0111] A spatio-temporal feature construction module, configured to construct spatio-temporal feature quantities based on a dynamic graph based on the preprocessed dataset;

[0112] A Tucker decomposition module, configured to perform dimensionality reduction on the spatio-temporal feature quantities based on Tucker decomposition to obtain dimensionality-reduced feature quantities;

[0113] A prediction model training module, configured to use the dimensionality-reduced feature quantities as the input of the LSTM network and the sum of the powers of all photovoltaic power stations as the output of the LSTM network to train the LSTM network, and obtain a photovoltaic power prediction model after the training is completed;

[0114] A power prediction module, configured to use the photovoltaic power prediction model to predict the photovoltaic power of the target area at the municipal level.

[0115] Aiming at the problem that the inaccurate prediction results of photovoltaic power at the municipal level are caused by incomplete photovoltaic power data, this system acquires the historical power data of photovoltaic power stations at each voltage level from the D5000 system, the power consumption acquisition system, and the distribution automation system, which can cover all photovoltaic power generation resources at the municipal level, and solves the problem that the deviation between the prediction results of photovoltaic power at the municipal level and the actual situation is too large.

[0116] Aiming at the problem that the prediction result of photovoltaic power using single longitude and latitude meteorological data is low due to insufficient resolution of meteorological data, this system obtains grid-based numerical weather forecasts from the power meteorological data system, which can effectively improve the resolution of meteorological data and thus enhance the prediction accuracy of photovoltaic power.

[0117] In an exemplary embodiment, the present application further provides a computer device, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface, and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. 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, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used for the processor to exchange information with external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the above-mentioned prefecture-level photovoltaic power prediction method.

[0118] In an exemplary embodiment, the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the above-mentioned prefecture-level photovoltaic power prediction method.

[0119] In an exemplary embodiment, the present application further provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the above-mentioned prefecture-level photovoltaic power prediction method.

[0120] Those of ordinary skill in the art will understand that all or part of the processes in the above-described embodiment methods can be completed by hardware related to computer program instructions. 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 embodiment of the above method. Among them, any reference to a memory or other medium provided in the embodiments of the present application can include at least one of non-volatile and volatile memories. 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), magnetoresistive 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. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0121] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data 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, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0122] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, 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, it should be considered as the scope described in this specification.

[0123] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for predicting photovoltaic power generation at the prefecture-level, characterized in that: include: Obtain historical power data of photovoltaic power plants of different voltage levels from the D5000 system, distribution automation system and power supply system; different voltage levels cover voltage levels from 380V to 35kV; obtain gridded numerical weather forecast historical data from the power meteorological data system; construct an initial input variable data set based on the historical power data of photovoltaic power plants and gridded numerical weather forecast historical data; preprocess the multi-source data in the initial input variable data set to obtain the preprocessed data set; Construct spatiotemporal feature quantities based on dynamic graphs based on preprocessed data sets; Based on Tucker decomposition, the dimension of the spatiotemporal feature quantity is reduced to obtain the feature quantity after dimension reduction; The feature quantity after dimensionality reduction is used as the input of the LSTM network, and the sum of the power of all photovoltaic power stations is used as the output of the LSTM network. The LSTM network is trained, and a photovoltaic power generation prediction model is obtained after the training is completed. The photovoltaic power generation prediction model is used to predict the photovoltaic power generation in the target area at the municipal level.

2. The method for predicting photovoltaic power generation at the prefecture-level according to claim 1, characterized in that: The acquisition of historical power data of photovoltaic power stations of different voltage levels from the D5000 system, the distribution automation system and the power collection system specifically includes: Obtain historical power data of centralized photovoltaic power plants of 10kV and above from the D5000 system; Obtain historical power data of 10kV distributed photovoltaic power stations from the distribution automation system; Obtain historical power data of 380V distributed photovoltaic power stations from the power supply system.

3. The method for predicting photovoltaic power generation at the prefecture-level according to claim 2, characterized in that: The obtaining of gridded numerical weather forecast historical data from the electric power meteorological data system specifically includes: Obtain b different types of gridded numerical weather forecast historical data from the power meteorological data system, including historical data of irradiance, temperature, humidity, wind speed, wind direction, air pressure and precipitation at different longitude and latitude locations.

4. The method for predicting photovoltaic power generation at the prefecture-level according to claim 3 is characterized in that: The initial input variable data set is constructed based on the historical power data of the photovoltaic power station and the historical data of the gridded numerical weather forecast, specifically including: Construct the initial input variable data set D, which contains d feature quantities and T time series; d = M × a = M × (b + 1); M is the number of photovoltaic power stations; a is all the variable types obtained, including b different types of gridded numerical weather forecast historical data and 1 type of photovoltaic power station historical power data; Construct an output variable dataset Y, which includes 1 variable and T time series; one variable is the sum of the power of all PV power stations.

5. The method for predicting photovoltaic power generation at the prefecture-level according to claim 4, characterized in that: The preprocessing of the multi-source data in the initial input variable data set to obtain the preprocessed data set specifically includes: Perform one or more preprocessing of data cleaning, abnormal data detection, data conversion, data dimension reduction, data classification and coding, data difference and aggregation, association analysis, and data standardization on the multi-source data in the initial input variable data set D to obtain the preprocessed data set D p ;D p Contains d p feature quantities and T time series; where d p =M*a', a' is the variable type remaining after preprocessing.

6. The method for predicting photovoltaic power generation at the prefecture-level according to claim 5, characterized in that: The step of constructing a spatiotemporal feature quantity based on a dynamic graph based on the preprocessed data set specifically includes: Note that the preprocessed data set D p In the example, the value of the kth variable type of the i-th distributed photovoltaic power station at time t is x k,i,t , using the formula Calculate spatiotemporal features x based on dynamic graphs l ; Where t' is the specified time window length; is the mean value of the kth variable type of the i-th PV plant within the t' window length; x k,j,t is the kth variable of the jth PV plant The value of type at time t; is the mean value of the kth variable type of the jth distributed photovoltaic power station within the t window length; k∈(1,2,...,a'); i,j∈(1,2,...,M); All spatiotemporal features x l The set of spatiotemporal features X loc ; Spatiotemporal feature quantity X loc The dimension is M×M×a'.

7. The method for predicting photovoltaic power generation at the prefecture-level according to claim 6, characterized in that: The dimensionality reduction of the spatiotemporal feature quantity based on Tucker decomposition to obtain the feature quantity after dimensionality reduction specifically includes: Based on Tucker decomposition formula X loc =G×1A×2B×3C, the spatiotemporal feature quantity X loc Split into a core tensor matrix G and three factor matrices A, B, C; where × n , n = 1, 2, 3 is the Tucker decomposition symbol; the dimension of factor matrix A is e × M, the dimension of factor matrix B is e × M, the dimension of factor matrix C is e × a', and the dimension of core tensor matrix G is e 3 ; e is a constant term; The set of all elements in the matrix G, A, B, and C is used as the feature quantity X after dimensionality reduction l ' oc .

8. A prefecture-level photovoltaic power generation prediction system, characterized in that: include: The power data acquisition module is used to obtain the historical power data of photovoltaic power plants of different voltage levels from the D5000 system, the distribution automation system and the power collection system; the different voltage levels cover the voltage levels of 380V to 35kV; A meteorological data acquisition module is used to obtain gridded numerical weather forecast historical data from the electric power meteorological data system; A data set construction module is used to construct an initial input variable data set based on the historical power data of photovoltaic power plants and the historical data of gridded numerical weather forecasts; A data preprocessing module is used to preprocess the multi-source data in the initial input variable data set to obtain a preprocessed data set; The spatiotemporal feature construction module is used to construct spatiotemporal feature quantities based on dynamic graphs based on the preprocessed data set; Tucker decomposition module, used to reduce the dimension of spatiotemporal feature quantities based on Tucker decomposition to obtain feature quantities after dimension reduction; The prediction model training module is used to use the feature quantity after dimensionality reduction as the input of the LSTM network, and the sum of the power of all photovoltaic power stations as the output of the LSTM network, to train the LSTM network, and after the training is completed, a photovoltaic power generation prediction model is obtained; The power prediction module is used to predict the photovoltaic power generation power of the target area at the prefecture-level city level using the photovoltaic power generation power prediction model.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting photovoltaic power generation at the municipal level according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting photovoltaic power generation at the municipal level described in any one of claims 1 to 7 is implemented.

Images