Joint prediction method, device and equipment for generating capacity of hydropower station cluster and storage medium

Abstract

The invention relates to the technical field of power generation capacity prediction, and discloses a hydropower station cluster power generation capacity combined prediction method, device and equipment and a storage medium. The hydropower station cluster power generation capacity combined prediction method comprises the steps that historical power generation capacity data of each hydropower station in a hydropower station cluster and multiple pieces of water regimen data related to the power generation capacity are collected, collecting geographical distance data between hydropower stations; constructing a generating capacity space influence prediction item based on the generating capacity matrix time sequence and the geographic distance data, constructing a generating capacity time influence prediction item based on a historical moment generating capacity matrix in the generating capacity matrix time sequence, and constructing a water regimen time influence prediction item based on a historical moment water regimen matrix in the water regimen matrix time sequence; constructing a matrix value time sequence space autoregression model; training the matrix value time sequence space autoregression model to obtain a generating capacity combined prediction model of the hydropower station cluster; and predicting the generating capacity of the hydropower station cluster. The generating capacity prediction accuracy can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power generation prediction, and in particular to a method, device, equipment and storage medium for joint prediction of power generation of a hydropower station cluster. Background Art

[0002] A hydropower station consists of a hydraulic system, a mechanical system, and an electric energy generating device. It is a key water conservancy project that realizes the conversion of water energy into electric energy. The sustainability of electric energy production requires the uninterrupted use of water energy in a hydropower station. Through the construction of a hydropower station reservoir system, the distribution of water resources in time and space is artificially adjusted and changed to achieve sustainable use of water resources. In order to effectively convert the water energy in the reservoir into electric energy, it is necessary to predict the power generation in a hydropower station so that the staff can make corresponding allocations of electricity and backup plans to ensure the continuity of power supply.

[0003] Existing methods for predicting power generation of hydropower stations are mainly developed from the aspects of time series analysis and deep learning. In traditional time series analysis, an autoregressive model is only constructed for a single hydropower station, and single variable or vector data modeling is performed based on historical power generation data. Deep learning models such as LSTM and extreme learning machines are also mostly based on the perspective of a single hydropower station, and comprehensive information such as the inflow flow, water level, and head of the hydropower station is used for modeling. Current methods are unable to model the spatial effects between hydropower station clusters, and do not utilize the natural correlation characteristics between hydropower station clusters, resulting in low accuracy in power generation prediction. Summary of the invention

[0004] In view of this, the present invention provides a method, device, equipment and storage medium for joint prediction of power generation of a hydropower station cluster to improve the accuracy of power generation prediction.

[0005] In a first aspect, the present invention provides a method for jointly predicting power generation of a cluster of hydropower stations, and the method comprises: collecting historical power generation data of each hydropower station in the hydropower station cluster and a plurality of water condition data related to power generation; constructing a power generation matrix time series of the hydropower station cluster based on the historical power generation data of each hydropower station; constructing a water condition matrix time series of the hydropower station cluster based on the water condition data of each hydropower station; constructing a data set based on the power generation matrix time series and the water condition matrix time series; constructing a power generation spatial impact prediction item based on the power generation matrix time series and geographic distance data, constructing a power generation time impact prediction item based on the power generation matrix at historical moments in the power generation matrix time series, and constructing a water condition time impact prediction item based on the water condition matrix at historical moments in the water condition matrix time series; constructing a matrix value time series spatial autoregression model in combination with the power generation spatial impact prediction item, the power generation time impact prediction item and the water condition time impact prediction item; training the matrix value time series spatial autoregression model based on the data set to obtain a joint power generation prediction model of the hydropower station cluster; and predicting the power generation of the hydropower station cluster based on the joint power generation prediction model.

[0006] In this implementation, the present application starts from the perspective of joint prediction of multiple hydropower stations in a hydropower station cluster, based on the similar power generation trend characteristics of hydropower stations under similar hydrological and meteorological conditions, fully considers the mutual influence between different hydropower stations, and comprehensively considers the influence of historical power generation matrix time series and other water condition matrix time series on power generation to construct a model. Specifically, a matrix-valued time series spatial autoregressive model is constructed based on the spatial effect influence of the natural geographical relationship between multiple hydropower stations, the pure time dynamic effect and the water condition variable effect, which not only considers the mutual influence relationship in the time dimension, but also considers the correlation in the spatial dimension, which can improve the accuracy of power generation prediction of hydropower station clusters.

[0007] In an optional embodiment, a power generation matrix time series of a cluster of hydropower stations is constructed based on the historical power generation data of each hydropower station, including: for daily historical power generation data, defining the matrix rows as each hydropower station, defining the matrix columns as the power generation of each hydropower station at different times, and obtaining the power generation matrix corresponding to each day; constructing a power generation matrix time series based on the power generation matrix of consecutive dates; constructing a water condition matrix time series of a cluster of hydropower stations based on the water condition data of each hydropower station, including: for daily water condition data, calculating the mean of each water condition variable of each hydropower station; defining each matrix row as each hydropower station, defining each matrix column as the mean of each water condition variable of each hydropower station, and obtaining the water condition matrix corresponding to each day; constructing a water condition matrix time series based on the water condition matrix of consecutive dates.

[0008] In this implementation, the date is used as the matrix establishment standard to construct the power generation matrix of the hydropower station and multiple water condition matrices of the hydropower station. The spatial relationship of the data can be considered, and the time series can be constructed in chronological order. The temporal relationship of the data can be considered, which provides a good data set environment for subsequent model training to further improve the model training effect.

[0009] In an optional embodiment, a prediction item of spatial impact of power generation is constructed based on the power generation matrix time series and the geographic distance data, including: constructing a hydropower station cluster spatial weight matrix based on the geographic distance data, the larger the geographic distance data, the smaller the corresponding element value in the hydropower station cluster spatial weight matrix; constructing a concurrent prediction item of spatial impact of power generation based on the current moment power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster; constructing a lagged prediction item of spatial impact of power generation based on the historical moment power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster.

[0010] In this implementation, the concurrent prediction items of the spatial influence on power generation and the lagged prediction items of the spatial influence on power generation are introduced into the matrix-valued time series spatial autoregressive model, which takes into account the pure spatial effect of power generation due to the natural geographical relationship between multiple hydropower stations and the spatial effect of time lag of power generation. It is more suitable for the joint power generation prediction of the current hydropower station cluster and improves the accuracy of subsequent model application.

[0011] In an optional embodiment, a concurrent prediction item of spatial impact of power generation is constructed based on the current moment power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster, including: fusing the current moment power generation matrix and the spatial weight matrix of the hydropower station cluster to obtain a first fusion matrix; adjusting the first fusion matrix using a first diagonal matrix to obtain a concurrent prediction item of spatial impact of power generation; constructing a lagged prediction item of spatial impact of power generation based on the historical moment power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster, including: fusing the historical moment power generation matrix of the first lag order and the spatial weight matrix of the hydropower station cluster The matrices are fused to obtain a second fused matrix; the second fused matrix is ​​adjusted by using the second diagonal matrix to obtain the spatial impact lag prediction item of power generation; the power generation time impact prediction item is constructed based on the historical moment power generation matrix in the power generation matrix time series, including: adjusting the historical moment power generation matrix of the second lag order by using the first regression coefficient matrix to obtain the power generation time impact prediction item; the water condition time impact prediction item is constructed based on the historical moment water condition matrix in the water condition matrix time series, including: adjusting the historical moment water condition matrix of the third lag order by using the second regression coefficient matrix to obtain the water condition time impact prediction item.

[0012] In an optional embodiment, a matrix-valued time series spatial autoregressive model is trained based on a data set to obtain a joint prediction model for power generation of a cluster of hydropower stations, including: dividing the data set into a training set and a test set in time series according to a division ratio; using the training set, estimating parameters of the matrix-valued time series spatial autoregressive model to obtain a trained matrix-valued time series spatial autoregressive model; testing the trained matrix-valued time series spatial autoregressive model based on the test set to determine the lag order to obtain a joint prediction model for power generation of a cluster of hydropower stations.

[0013] In this implementation, the iterative estimation method of the present application can effectively model endogeneity problems.

[0014] In an optional embodiment, parameters of a matrix-valued time series spatial autoregressive model are estimated to obtain a trained matrix-valued time series spatial autoregressive model, including: obtaining a plurality of preset lag order combinations; using a training set, parameters of the matrix-valued time series spatial autoregressive models corresponding to the plurality of preset lag order combinations are estimated to obtain a plurality of trained matrix-valued time series spatial autoregressive models; the trained matrix-valued time series spatial autoregressive model is tested based on a test set to determine the lag order to obtain a joint prediction model for power generation of a hydropower station cluster, including: calculating model errors of the plurality of trained matrix-valued time series spatial autoregressive models based on the test set; calculating a Bayesian information criterion value based on the model error, and selecting the preset lag order combination with the minimum Bayesian information criterion value to obtain the corresponding joint prediction model for power generation.

[0015] In this implementation, different order combinations are verified and selected to determine the optimal combination, avoiding the problem that the model may be under-fitting when the order is too small, and the problem that the model may be over-fitting when the order is too large. This method can make the model show a good fitting effect on the existing data and reduce prediction errors.

[0016] In a second aspect, the present invention provides a device for jointly predicting the power generation of a cluster of hydropower stations, and the device comprises: a collection module for collecting historical power generation data of each hydropower station in the hydropower station cluster and a plurality of water condition data related to power generation; collecting geographical distance data between hydropower stations; a first construction module for constructing a power generation matrix time series of the hydropower station cluster based on the historical power generation data of each hydropower station; constructing a water condition matrix time series of the hydropower station cluster based on the water condition data of each hydropower station; constructing a data set based on the power generation matrix time series and the water condition matrix time series; a second construction module for constructing a data set based on the power generation matrix time series and the water condition matrix time series. The spatial impact prediction item of power generation is constructed based on the geographical distance data, the time impact prediction item of power generation is constructed based on the historical moment power generation matrix in the power generation matrix time series, and the time impact prediction item of water condition is constructed based on the historical moment water condition matrix in the water condition matrix time series; a matrix value time series spatial autoregressive model is constructed by combining the spatial impact prediction item of power generation, the time impact prediction item of power generation and the time impact prediction item of water condition; a training module is used to train the matrix value time series spatial autoregressive model based on the data set to obtain a joint prediction model of power generation of a hydropower station cluster; a prediction module is used to predict the power generation of a hydropower station cluster based on the joint prediction model of power generation.

[0017] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for joint prediction of power generation of a cluster of hydropower stations according to the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for jointly predicting power generation of a cluster of hydropower stations according to the first aspect or any corresponding embodiment thereof.

[0019] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the method for jointly predicting power generation of a cluster of hydropower stations according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 It is a flow chart of a method for joint prediction of power generation of a hydropower station cluster according to an embodiment of the present invention; Figure 2 is a flow chart of another method for joint prediction of power generation of a hydropower station cluster according to an embodiment of the present invention; Figure 3 is a structural block diagram of a device for joint prediction of power generation of a hydropower station cluster according to an embodiment of the present invention; Figure 4 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0023] With the enhancement of information collection capabilities, more and more complex structured data are regularly stored and recorded, which has brought great impact and innovation to the ideas and methods of statistics, and also given new meaning to the important field of time series analysis. Traditional time series analysis is mostly based on single variable or vector observations. However, due to the addition of complex structured data, traditional time series analysis methods cannot be directly applied. In complex time series, a common form is matrix valued time series, that is, the observation value at each time point is a matrix. Time series data with such a structure exist in many real-world scenarios, for example, power generation data of different hydropower stations with a daily sampling interval of 15 minutes.

[0024] There are certain correlation characteristics between the hydropower stations in the hydropower station cluster, which will have an important impact on the prediction of the power generation of the hydropower station, and the spatial effects produced by this correlation characteristic are different. Therefore, it is of certain significance to analyze the structural characteristics of the natural matrix value time series data of the power generation data of the hydropower station cluster and comprehensively examine the spatial effects between different hydropower stations. Therefore, this application proposes a joint prediction method for the power generation of a hydropower station cluster, which considers the spatial effect and time effect between the hydropower stations to predict the power generation, so as to improve the accuracy of the power generation prediction.

[0025] According to an embodiment of the present invention, an embodiment of a method for joint prediction of power generation of a cluster of hydropower stations is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] In this embodiment, a method for joint prediction of power generation of a hydropower station cluster is provided. Figure 1 is a flow chart of a method for joint prediction of power generation of a hydropower station cluster according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment does not use Figure 1 The process sequence shown is limited. Figure 1 As shown, the process includes the following steps: Step S101 , collecting historical power generation data and a plurality of water condition data related to power generation of each hydropower station in the hydropower station cluster, and collecting geographical distance data between the hydropower stations.

[0027] The hydropower station cluster includes multiple hydropower stations, and each hydropower station affects each other. In one implementation, the hydropower station cluster is a small cascade hydropower station with hydraulic connection.

[0028] Among them, the water conditions data related to power generation include the inflow, outflow, water level, head, etc. of each hydropower station.

[0029] Specifically, the present application collects the power generation data of each hydropower station in the hydropower station cluster according to the preset sampling interval, obtains the historical power generation data, and records the timestamp of each power generation data; collects the historical water conditions data such as the inflow, outflow, water level, head, etc. of each hydropower station that are related to the power generation forecast of the hydropower station, and records the timestamp of each water conditions data.

[0030] Exemplarily, the preset sampling interval is 15 minutes, and the power generation data of each hydropower station in the hydropower station cluster is collected every 15 minutes.

[0031] At the same time, the present application collects geographical distance data between any two hydropower stations in the hydropower station cluster, wherein the geographical distance between the hydropower stations represents the degree of mutual influence between the two hydropower stations, and the closer the geographical distance, the greater the degree of influence.

[0032] Step S102, constructing a power generation matrix time series of a hydropower station cluster based on the historical power generation data of each hydropower station; constructing a water condition matrix time series of a hydropower station cluster based on the water condition data of each hydropower station; and constructing a data set based on the power generation matrix time series and the water condition matrix time series.

[0033] The historical power generation data of each hydropower station and multiple water condition data related to power generation are divided according to unit time intervals. The historical power generation data in each time unit is used to construct a power generation matrix, which is arranged in chronological order to obtain a power generation matrix time series; similarly, the water condition data in each time unit is used to construct a water condition matrix, which is arranged in chronological order to obtain a water condition matrix time series. The power generation matrix time series and the water condition matrix time series constitute the data set.

[0034] In one implementation, the unit time interval is one day. Specifically, the historical power generation data observed at each hydropower station every day is used to construct a power generation matrix, and the power generation matrices on different dates form a power generation matrix time series. , a water condition matrix is ​​constructed according to the water condition data observed at each hydropower station every day, and the water condition matrices of different dates form a water condition matrix time series ,in, Represents time.

[0035] Step S103, constructing a prediction item for the spatial impact of power generation based on the power generation matrix time series and the geographic distance data, constructing a prediction item for the time impact of power generation based on the power generation matrix at historical moments in the power generation matrix time series, and constructing a prediction item for the time impact of water conditions based on the water condition matrix at historical moments in the water condition matrix time series; and constructing a matrix value time series spatial autoregressive model by combining the power generation spatial impact prediction item, the power generation time impact prediction item and the water condition time impact prediction item.

[0036] This application introduces the spatial regression term of the power generation matrix time series, the spatial regression term of the power generation matrix time series lag period and the regression term of the water regime matrix time series lag period to construct a matrix valued time series spatial autoregressive model (Matrix Spatial-Temporal Lag Autoregressive Model, MSTLAR).

[0037] Specifically, the power generation matrix time series is combined with the geographical distance data between power stations, and the influence of the power generation of the hydropower station on the power generation forecast in the geographical space dimension is considered to obtain the power generation space impact prediction item; the power generation matrix at the historical moment in the power generation matrix sequence is determined according to the number of lags, and the influence of the power generation of the hydropower station on the power generation forecast in the time dimension is considered to obtain the power generation time impact prediction item; similarly, the water condition power generation matrix in the water condition matrix sequence is determined according to the number of lags, and the influence of the water condition data related to the power generation of the hydropower station on the power generation forecast in the time dimension is considered to obtain the water condition time impact prediction item. Combining multiple prediction regression items, the matrix value time series spatial autoregressive model of the present application is obtained.

[0038] Among them, in this implementation, the lag order of the power generation matrix at the historical moment and the lag order of the historical water condition matrix are not limited.

[0039] Step S104, training the matrix value time series spatial autoregressive model based on the data set to obtain a joint prediction model for power generation of the hydropower station cluster.

[0040] In the data set, the historical power generation matrix is ​​determined in the power generation matrix time series according to the lag order, the historical water regime matrix is ​​determined in the water regime matrix time series, and a group of training samples is determined by combining the current power generation matrix, the historical power generation matrix and the historical water regime matrix.

[0041] For example, when the lag order of the historical power generation matrix is ​​1 and the lag order of the historical water regime matrix is ​​2, The power generation matrix at each moment, The historical power generation matrix at time The historical water condition matrix at each moment is used as a set of training samples.

[0042] The matrix-valued time series spatial autoregressive model constructed in the above steps is trained using multiple groups of training samples to obtain a joint prediction model for power generation of a hydropower station cluster.

[0043] Step S105: predicting the power generation of the hydropower station cluster based on the power generation joint prediction model.

[0044] The historical power generation data and multiple water condition data corresponding to the lag order are obtained, and the power generation matrix and water condition matrix are constructed. The data are input into the power generation joint prediction model of the hydropower station cluster, and the power generation at the future moment is obtained by prediction output.

[0045] The method for joint prediction of power generation of a cluster of hydropower stations provided in this embodiment is triggered from the perspective of joint prediction of multiple hydropower stations in a cluster of hydropower stations. It is based on the similar power generation trend characteristics of hydropower stations under similar hydrological and meteorological conditions, fully considers the mutual influence between different hydropower stations, and comprehensively considers the influence of historical power generation matrix time series and other water condition matrix time series on power generation to construct a model. Specifically, a matrix value time series spatial autoregressive model is constructed based on the spatial effect influence of the natural geographical relationship between multiple hydropower stations, the pure time dynamic effect and the water condition data influence effect, which not only considers the mutual influence relationship in the time dimension, but also considers the correlation relationship in the spatial dimension, which can improve the accuracy of power generation prediction of a cluster of hydropower stations.

[0046] In this embodiment, a method for joint prediction of power generation of a hydropower station cluster is provided. Figure 2is a flow chart of another method for joint prediction of power generation of a hydropower station cluster according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment does not use Figure 2 The process sequence shown is limited. Figure 2 As shown, the process includes the following steps: Step S201 , collecting historical power generation data and a plurality of water condition data related to power generation of each hydropower station in the hydropower station cluster, and collecting geographical distance data between the hydropower stations.

[0047] Specifically, a sampling time interval is set, such as 15 minutes, and the historical power generation data and its timestamp of real-time monitoring of each hydropower station in the hydropower station cluster are collected based on the data collection and monitoring system of each hydropower station. The inflow, outflow, water level, head and other water situation data related to the power generation prediction of the hydropower station and its timestamp are collected based on the reservoir hydrological monitoring system. The multiple data collected in this application are all time series data.

[0048] Furthermore, the collected variables related to the prediction of the power generation of the hydropower station are screened for relevance, and the water regime variables with higher relevance are selected to obtain the corresponding water regime data.

[0049] In one implementation, the Pearson correlation coefficients between multiple water regime variable data related to the power generation prediction of the hydropower station and the corresponding historical power generation data are calculated, and a preset number of water regime variables with high correlation are selected as inputs for subsequent model training, and the corresponding water regime data are extracted. The five water regime variables with the highest correlation are selected, and the corresponding water regime data are extracted.

[0050] Furthermore, the collected data is preprocessed.

[0051] Specifically, the abnormal parts in the collected data are identified through clustering algorithms, 3-σ principle and other methods, and then the missing values ​​are filled and the abnormal values ​​are replaced by moving average method or exponential smoothing method; the trend, seasonality and cyclical characteristics of time series data are identified by combining statistical methods such as unit root test, and the non-stationary time series are converted into stationary time series by using methods such as difference, deterministic detrending and structural change. This method can ensure the stationarity of the input data of the time series model; and each time series is normalized.

[0052] At the same time, this application collects geographical distance data between any two hydropower stations in the hydropower station cluster.

[0053] Step S202, constructing a power generation matrix time series of a hydropower station cluster based on the historical power generation data of each hydropower station; constructing a water condition matrix time series of a hydropower station cluster based on the water condition data of each hydropower station; and constructing a data set based on the power generation matrix time series and the water condition matrix time series.

[0054] Specifically, the above step S202 includes: Step S2021, for daily historical power generation data, define the matrix rows as each hydropower station, and define the matrix columns as the power generation of each hydropower station at different times, to obtain the power generation matrix corresponding to each day.

[0055] For example, when the sampling time interval is 15 minutes, for the historical power generation data of each hydropower station, each row is defined as a different hydropower station, and each column is the power generation collected every 15 minutes every day, that is, each row represents the power generation of a hydropower station at 96 moments a day, and a power generation matrix corresponding to each day is obtained. .

[0056] Step S2022: construct a power generation matrix time series based on the power generation matrix of consecutive dates.

[0057] The power generation matrix of different dates is constructed in chronological order to obtain the power generation matrix time series .

[0058] Step S2023, for daily water regime data, calculate the mean of each water regime variable of each hydropower station.

[0059] Exemplarily, when the sampling time interval is 15 minutes and there are 5 water regime variables, the mean of the water regime data collected at 96 times a day for the 5 different water regime variables of each hydropower station is calculated.

[0060] Step S2024, define each row of the matrix as each hydropower station, define each column of the matrix as the mean value of each water regime variable of each hydropower station, and obtain the water regime matrix corresponding to each day.

[0061] For example, for the water regime data of each hydropower station, the rows are defined as different hydropower stations, and the columns are the means of 5 different water regime variables, and a water regime matrix corresponding to each day is obtained: .

[0062] Step S2025, constructing a water condition matrix time series based on the water condition matrix of consecutive dates.

[0063] The water regime matrices of different dates are constructed in chronological order to obtain the water regime matrix time series. .

[0064] Step S2026, constructing a data set based on the power generation matrix time series and the water regime matrix time series.

[0065] In this implementation, the date is used as the matrix establishment standard to construct a matrix of power generation of multiple hydropower stations, and a water condition matrix of multiple hydropower stations. The spatial relationship of the data can be considered, and the time series can be constructed in chronological order. The time relationship of the data can be considered, which provides a good data set environment for subsequent model training to further improve the model training effect.

[0066] Step S203, constructing a prediction item for the spatial impact of power generation based on the power generation matrix time series and the geographic distance data, constructing a prediction item for the time impact of power generation based on the power generation matrix at historical moments in the power generation matrix time series, and constructing a prediction item for the time impact of water conditions based on the water condition matrix at historical moments in the water condition matrix time series; and constructing a matrix value time series spatial autoregressive model by combining the power generation spatial impact prediction item, the power generation time impact prediction item and the water condition time impact prediction item.

[0067] The hydropower station cluster spatial weight matrix can reflect the distance of correlation or association between the relevant data of two hydropower stations. Generally, in spatial econometrics, it is usually assumed that the hydropower station cluster spatial weight matrix is ​​known. There are many ways to set the spatial weight matrix, such as the common economic distance spatial weight matrix, geographic distance spatial weight matrix, etc.

[0068] In this implementation, a hydropower station cluster spatial weight matrix is ​​constructed based on geographic distance data. The larger the geographic distance data, the smaller the corresponding element value in the hydropower station cluster spatial weight matrix.

[0069] Specifically, the spatial weight matrix of the hydropower station cluster is Each element in ,in, Expressed as The hydropower station and The geographical distance between the hydropower stations, and .

[0070] In one implementation, the spatial impact prediction item of power generation includes a concurrent spatial impact prediction item of power generation and a delayed spatial impact prediction item of power generation.

[0071] Specifically, the current moment power generation matrix of the power generation matrix time series is combined with the spatial weight matrix of the hydropower station clusters between the power stations, and the influence of the current hydropower station power generation on the power generation forecast in the geographic space dimension is considered to obtain the concurrent prediction item of the spatial influence of power generation; the historical moment power generation matrix of the power generation matrix time series is combined with the spatial weight matrix of the hydropower station clusters between the power stations, and the influence of the lagged hydropower station power generation with a certain number of lags in the geographic space dimension on the power generation forecast is considered to obtain the lagged prediction item of the spatial influence of power generation.

[0072] Furthermore, in this implementation, a matrix-valued time series spatial autoregressive model is constructed by combining the concurrent prediction item of spatial impact on power generation, the delayed prediction item of spatial impact on power generation, the temporal impact prediction item of power generation and the temporal impact prediction item of water regime.

[0073] Specifically, the matrix-valued time series spatial autoregressive model is constructed The specific form is: .

[0074] in, is the power generation matrix time series, and the left side of the model equation is the power generation matrix at the current moment. for dimensional matrix, representing day Hydropower station Observed values ​​of power generation at different times.

[0075] The first term on the right side of the model equation is the concurrent prediction term of the spatial impact of power generation, which represents the pure spatial effect of power generation.

[0076] Specifically, the current power generation matrix and the hydropower station cluster spatial weight matrix are fused to obtain a first fusion matrix; the first fusion matrix is ​​adjusted using the first diagonal matrix to obtain the prediction item of the power generation spatial impact during the same period. yes The dimensional spatial weight matrix of the hydropower station cluster has zero elements on the main diagonal, which can describe the dependencies between different hydropower stations. and Respectively and The first diagonal matrix of dimension. Among them, for the matrix Middle Elements The first term contributes No. The contemporaneous spatial effect of the column, It is The spatial influence coefficient of each position, It is The adjustment coefficient of an indicator for spatial effects.

[0077] The second term on the right side of the model equation is the prediction term of the time impact of power generation, which represents the pure dynamic effect of power generation, that is, the time lag effect.

[0078] Specifically, the first regression coefficient matrix is ​​used to adjust the historical power generation matrix of the second lag order to obtain the power generation time impact prediction item. and They are and dimensional first regression coefficient matrix, and They are and No. OK, is the second lag order. Among them, No. The elements can be represented as , which indicates that the second term on the right side of the model is important for predicting The contribution of the matrix The linear combination of represents the lagged prediction relationship in the time dimension.

[0079] The third term on the right side of the model equation is the lagged prediction term of the spatial impact of power generation, which represents the spatial effect of the time lag of power generation.

[0080] Specifically, the historical power generation matrix of the first lag order and the hydropower station cluster spatial weight matrix are fused to obtain a second fusion matrix; the second fusion matrix is ​​adjusted using the second diagonal matrix to obtain the lag prediction item of the power generation spatial impact. and Respectively and The second diagonal matrix of dimension , is the first lag order. Among them, for the matrix Middle Elements The third term contributes No. The lagged spatial effect of the column, It is The spatial influence coefficient of each position, It is The adjustment coefficient of an indicator for spatial effects.

[0081] The fourth term on the right side of the model equation is the prediction term for the time impact of water regime, which represents the time impact effect of water regime variables.

[0082] Specifically, the second regression coefficient matrix is ​​used to adjust the historical water regime matrix of the third lag order to obtain the water regime time impact prediction item. is the water regime matrix time series, for dimensional matrix, representing day Hydropower station Observed values ​​of water regime variables. Indicates the length of the collected data time series. and They are and dimensional second regression coefficient matrix. is the third lag order.

[0083] The fifth term on the right side of the model equation is the error term, is a dimensional white noise matrix and satisfies ,in , .

[0084] In one implementation, when the lag order is 1, The specific form of the model is: (1) Step S204, training the matrix-valued time series spatial autoregressive model based on the data set to obtain a joint prediction model for power generation of the hydropower station cluster.

[0085] In some optional implementations, the above step S204 includes: Step S2041, dividing the data set into a training set and a test set in time series according to the division ratio.

[0086] Specifically, the data set is divided into a training set and a test set according to a preset ratio, such as a preset ratio of 7:3.

[0087] Step S2042, using the training set, performing parameter estimation on the matrix-valued time series spatial autoregressive model to obtain a trained matrix-valued time series spatial autoregressive model.

[0088] In one implementation, the first-order lag The model is used as an example to estimate parameters. exists on both sides of the model formula, so and This will lead to endogeneity problems. In this case, we can directly use the least squares method to and Regression will produce inconsistent estimates. Therefore, this paper proposes an estimation method based on the Yule-Walker equation to solve the endogeneity problem. The Yule-Walker equation is an equation that describes the relationship between the parameters of an autoregressive sequence and its covariance function. Therefore, the Yule-Walker equation does not contain the information of the error term and can effectively eliminate the endogeneity problem in the original regression model.

[0089] The MSTLAR model proposed in this application consists of three parts: pure spatial effect, pure dynamic effect (time lag effect) and time lag spatial effect. The estimation methods of these three parts are given respectively, and then an iterative calculation method is constructed. Specifically, some parameters are given, the other parameters are solved, and then the solution is iteratively obtained. Specifically, the estimation process is mainly divided into the following steps: Step 1: Estimate the time-dependent impact of power generation forecast items.

[0090] Specifically, given the matrix , , , , and The estimated result is , , , , and .

[0091] definition , then formula (1) is converted to , the model does not contain endogeneity problems. At this time, the least squares method can be used to obtain and The consistent estimate of , its parameter optimization problem can be expressed as: (2) In order to optimize the objective function given in equation (2), the matrix Frobenius norm is first converted into the form of matrix trace, and then according to the differentiation rule of the trace to the matrix, the matrix and Taking the partial derivative and setting it to 0, we can get: (3) Pair Matrix and The calculation can be estimated iteratively as follows: Given the matrix Estimates , based on formula (3), we can get the matrix Updated estimate of : , (4) Similarly, given the matrix Estimates , based on formula (3), we can get the matrix Updated estimate of : (5) Step 2: Estimate the water regime time impact prediction items.

[0092] Specifically, similar to step 1, given the matrix , , , , and The estimated result is , , , , and After that, the model does not have endogeneity problems, and the least squares method can be used to obtain parameter estimation results: , ; .

[0093] in, .

[0094] Step 3: Estimate the lagged prediction term of the spatial impact of power generation.

[0095] Specifically, given the matrix , , and The estimated result is , , and , solve the parameter matrix and The objective function is defined as: (6) in, Similarly, there is no endogenous problem in parameter solution at this time, and the iterative estimation method is used to solve and The least squares estimate of . Given Estimates ,definition The minimization problem given in equation (6) can be rewritten as . Regularly remove all The relevant formula, at this time about The objective optimization function is: (7) in, and They are and No. OK. In order to solve The least squares estimate of Taking the derivative and setting its partial derivative to 0, we get: (8) In the estimation After that, normalization is required , so that its Frobenius norm is 1.

[0096] Similarly, given the parameter matrix Estimates , we can get The updated estimate is: (9) in, , and They are and No. List.

[0097] Step 4: Estimate the spatial impact of power generation on the concurrent forecast items.

[0098] Specifically, given the parameter matrix , , and The estimate is , , and , based on the Yule–Walker equation to solve the parameter matrix and For any ,definition Based on model (3.1), the following Yule–Walker equation can be obtained: (10) in, yes dimensional unit matrix. Regularly take out The relevant lines, , .

[0099] in, Represents the spatial weight matrix No. OK, Represents the unit matrix No. OK, for No. OK. , using the sample covariance matrix instead and , , .

[0100] Then, the least squares method is used to solve the problem by minimizing the following objective function : .

[0101] definition , by changing the above objective function to Find the partial derivative and set it to 0, and we get The generalized Yule–Walker estimate of is: (11) In the estimation After that, normalization is required , so that its Frobenius norm is 1.

[0102] Similarly, regularly take out the The relevant lines: , .

[0103] Can get The generalized Yule–Walker estimate of is: (12) in, .

[0104] In summary, the overall iterative solution process can be summarized as follows: ① Given the matrix , , , , and The initial value of , , , , and , using step 1 to give the matrix and The initial value of and ;②For the In the iteration, the current estimation result is updated using steps 1 to 3. ; ③ Repeat step ② until convergence, giving the final parameter estimate . This paper sets the number of iterations to 100. This paper sets the iteration stop condition as follows: the number of iterations reaches 100, or the sum of the absolute differences of all parameter estimates of two consecutive iterations is less than 0.001. The parameter estimation process of model (1) can be summarized as an iterative algorithm based on the Yule-Walker equation, as follows: The algorithm input is: random initial value .

[0105] The output of the algorithm is: parameter estimates .

[0106] The algorithm steps are: 1. , .

[0107] .

[0108] 2. , .

[0109] .

[0110] 3. , .

[0111] .

[0112] 4. , .

[0113] .

[0114] 5. Repeat 1 to 3 until the iterative estimation of each coefficient matrix converges, and output the estimated value: .

[0115] In this implementation, the spatial impact contemporaneous prediction item of power generation, the temporal impact prediction item of power generation, the lagged prediction item of spatial impact of power generation and the temporal impact prediction item of water regime are introduced into the matrix-valued time series spatial autoregressive model. Not only the influence of historical data and water regime data, but also the spatial effect of different hydropower stations are considered. It is more suitable for the joint power generation prediction of the current hydropower station cluster and improves the accuracy of subsequent model application.

[0116] In some optional implementations, the above step S2042 includes: Step a1, obtaining multiple preset lag order combinations.

[0117] In general, when performing a matrix-valued time series When fitting a model, it is not possible to know in advance , and The optimal order of . , and If the value of is selected to be small, the model may be underfitting; , and If the value of is large, the model may be overfitted. In order to make the model have a better fitting effect on the existing data and have a lower prediction error, it is necessary to , and Therefore, the present application presets a plurality of lag order combinations.

[0118] Step a2, using the training set, performing parameter estimation on the matrix-valued time series spatial autoregressive models corresponding to multiple preset lag order combinations to obtain multiple trained matrix-valued time series spatial autoregressive models.

[0119] According to the corresponding preset lag order combination, the training set is determined, and the parameters of the matrix-valued time series spatial autoregressive model corresponding to each preset lag order combination are estimated by using the above-mentioned Yule-Walker equation and the least squares method, so as to obtain multiple trained matrix-valued time series spatial autoregressive models. The specific parameter estimation method refers to the above content and will not be repeated here.

[0120] Step S2043, testing the trained matrix-valued time series spatial autoregressive model based on the test set, and determining the lag order to obtain a joint prediction model for power generation of the hydropower station cluster.

[0121] In some optional implementations, the above step S2043 includes: Step b1, calculating the model errors of multiple trained matrix-valued time series spatial autoregressive models based on the test set.

[0122] Among them, for Model, construct the error function as the covariance estimate of the residual : .

[0123] in, , is the test value of the trained matrix-valued time series spatial autoregressive model, is the actual value of the test set.

[0124] Step b2, calculating the Bayesian information criterion value based on the error function.

[0125] Specifically, the Bayesian Information Criterion (BIC) is: .

[0126] in, express The value of the determinant of .

[0127] Step b3, selecting a preset lag order combination with the smallest Bayesian information criterion value, and determining the lag order to obtain a corresponding power generation joint prediction model.

[0128] By minimizing the Bayesian information criterion value, the optimal lag order combination is obtained , and obtain the corresponding power generation joint prediction model.

[0129] In this implementation, different order combinations are verified and selected to determine the optimal combination, avoiding the problem that the model may be under-fitting when the order is too small, and the problem that the model may be over-fitting when the order is too large. This method can make the model show a good fitting effect on the existing data and reduce prediction errors.

[0130] Step S205: predicting the power generation of the hydropower station cluster based on the power generation joint prediction model.

[0131] The historical power generation data and multiple water condition data corresponding to the lag order are obtained, and the power generation matrix and water condition matrix are constructed. The data are input into the power generation joint prediction model of the hydropower station cluster, and the power generation at the future moment is obtained by prediction output.

[0132] The method for joint prediction of power generation of a cluster of hydropower stations provided in this embodiment can predict the power generation of a cluster of hydropower stations while maintaining the matrix value data structure, and comprehensively considers the hydrological influencing factors of the cluster of hydropower stations and the correlation characteristics between hydropower stations. Not only the mutual influence relationship in the time dimension is considered, but also the correlation relationship in the spatial dimension between the clusters of hydropower stations is considered. In terms of parameter estimation, iterative estimation combined with the Yule-Walker equation and the least squares method can solve the problem of model endogeneity. In addition, a large number of parameters to be estimated are generated when modeling. This application solves the problem that too many parameters may cause parameter expansion and overfitting under the same sample data.

[0133] In this embodiment, a device for joint prediction of power generation of a hydropower station cluster is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0134] This embodiment provides a joint prediction device for the power generation of a hydropower station cluster, such as Figure 3 As shown, including: The collection module 301 is used to collect the historical power generation data and multiple water condition data related to the power generation of each hydropower station in the hydropower station cluster, and collect the geographical distance data between the hydropower stations.

[0135] The first construction module 302 is used to construct a power generation matrix time series of a hydropower station cluster based on the historical power generation data of each hydropower station; to construct a water condition matrix time series of a hydropower station cluster based on the water condition data of each hydropower station; and to construct a data set based on the power generation matrix time series and the water condition matrix time series.

[0136] The second construction module 303 is used to construct a prediction item for the spatial impact of power generation based on the power generation matrix time series and the geographic distance data, to construct a prediction item for the time impact of power generation based on the power generation matrix at historical moments in the power generation matrix time series, and to construct a prediction item for the time impact of water conditions based on the water conditions matrix at historical moments in the water conditions matrix time series; and to construct a matrix value time series spatial autoregressive model by combining the spatial impact prediction item for power generation, the time impact prediction item for power generation and the time impact prediction item for water conditions.

[0137] The training module 304 is used to train the matrix value time series spatial autoregressive model based on the data set to obtain a joint prediction model for power generation of the hydropower station cluster.

[0138] The prediction module 305 is used to predict the power generation of the hydropower station cluster based on the power generation joint prediction model.

[0139] In some optional implementations, the first building module 302 includes: The first matrix construction unit is used to define the matrix rows as each hydropower station and the matrix columns as the power generation of each hydropower station at different times for the daily historical power generation data, so as to obtain the power generation matrix corresponding to each day.

[0140] The first sequence construction unit is used to construct a power generation matrix time series based on the power generation matrix of consecutive dates.

[0141] The calculation unit is used to calculate the mean value of each water condition variable of each hydropower station based on the daily water condition data.

[0142] The second matrix construction unit is used to define each row of the matrix as each hydropower station, and define each column of the matrix as the mean value of each water regime variable of each hydropower station, so as to obtain the water regime matrix corresponding to each day.

[0143] The second sequence construction unit is used to construct a water regime matrix time series based on the water regime matrix of consecutive dates.

[0144] In some optional implementations, the second building module 303 includes: The first model building unit is used to build a hydropower station cluster spatial weight matrix based on the geographic distance data. The larger the geographic distance data, the smaller the corresponding element value in the hydropower station cluster spatial weight matrix. It is used to build a synchronous prediction item of the spatial impact of power generation based on the current moment power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster. It is used to build a lagged prediction item of the spatial impact of power generation based on the historical moment power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster.

[0145] In some optional implementations, the second building module 303 includes: The first model building unit is used to fuse the current moment power generation matrix and the hydropower station cluster spatial weight matrix to obtain a first fusion matrix; the first fusion matrix is ​​adjusted using a first diagonal matrix to obtain the power generation spatial impact contemporaneous prediction item. The historical moment power generation matrix of the first lag order and the hydropower station cluster spatial weight matrix are fused to obtain a second fusion matrix; the second fusion matrix is ​​adjusted using a second diagonal matrix to obtain the power generation spatial impact lag prediction item.

[0146] The second model building unit is used to adjust the historical moment power generation matrix of the second lag order by using the first regression coefficient matrix to obtain the power generation time impact prediction item.

[0147] The third model building unit is used to adjust the water condition matrix of the historical moment of the third lag order by using the second regression coefficient matrix to obtain the water condition time impact prediction item.

[0148] In some optional implementations, the training module 304 includes: The partitioning unit is used to divide the data set into a training set and a test set in time series according to the partitioning ratio.

[0149] The training unit is used to estimate the parameters of the matrix-valued time series spatial autoregressive model by using the training set, and determine the lag order to obtain the trained matrix-valued time series spatial autoregressive model.

[0150] The testing unit is used to test the trained matrix value time series spatial autoregressive model based on the test set to obtain a joint prediction model for power generation of a hydropower station cluster.

[0151] In some optional implementations, the training unit includes: The acquisition subunit is used to acquire a plurality of preset lag order combinations, wherein the preset lag order combinations include a preset first lag order, a preset second lag order, and a preset third lag order.

[0152] The training subunit is used to use the training set to estimate the parameters of the matrix-valued time series spatial autoregressive model corresponding to multiple preset lag order combinations, and determine the lag order to obtain multiple trained matrix-valued time series spatial autoregressive models.

[0153] In some optional embodiments, the test unit includes: The first calculation subunit is used to calculate the model errors of multiple trained matrix-valued time series spatial autoregressive models based on the test set.

[0154] The second calculation subunit is used to calculate the Bayesian information criterion value based on the error function, and select a preset lag order combination when the Bayesian information criterion value is the smallest to obtain a corresponding power generation joint prediction model.

[0155] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0156] The joint prediction device for the power generation of a cluster of hydropower stations in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0157] The embodiment of the present invention also provides a computer device having the above Figure 4 The joint prediction device for power generation of a hydropower station cluster is shown.

[0158] See also Figure 4 , Figure 4 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 4As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0159] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0160] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0161] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0162] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0163] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.

[0164] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device may be a touch screen.

[0165] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0166] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0167] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for joint prediction of power generation of a hydropower station cluster, characterized in that: The method comprises: Collect historical power generation data and multiple water condition data related to power generation forecast for each hydropower station in the hydropower station cluster; collect geographical distance data between hydropower stations; Based on the historical power generation data of each hydropower station, a power generation matrix time series of the hydropower station cluster is constructed; based on the water regime data of each hydropower station, a water regime matrix time series of the hydropower station cluster is constructed; based on the power generation matrix time series and the water regime matrix time series, a data set is constructed; Constructing a prediction item of spatial impact of power generation based on the power generation matrix time series and the geographic distance data, constructing a prediction item of temporal impact of power generation based on the power generation matrix at historical moments in the power generation matrix time series, and constructing a prediction item of temporal impact of water regime based on the water regime matrix at historical moments in the water regime matrix time series; constructing a spatial autoregressive model of matrix value time series by combining the prediction item of spatial impact of power generation, the prediction item of temporal impact of power generation and the prediction item of temporal impact of water regime; Training the matrix-valued time series spatial autoregressive model based on the data set to obtain a joint prediction model for power generation of the hydropower station cluster; The power generation of the hydropower station cluster is predicted based on the power generation joint prediction model.

2. The method for joint prediction of power generation of a hydropower station cluster according to claim 1 is characterized in that: The step of constructing a power generation matrix time series of the hydropower station cluster based on the historical power generation data of each hydropower station includes: For the daily historical power generation data, define the matrix rows as each hydropower station, and define the matrix columns as the power generation of each hydropower station at different times, to obtain the power generation matrix corresponding to each day; Based on the power generation matrix of consecutive dates, construct the power generation matrix time series; The step of constructing a water regime matrix time series of the hydropower station cluster based on the water regime data of each hydropower station includes: Calculate the mean value of each water condition variable of each hydropower station for the daily water condition data; Define each row of the matrix as each hydropower station, define each column of the matrix as the mean value of each water regime variable of each hydropower station, and obtain the water regime matrix corresponding to each day; Based on the water condition matrix of consecutive dates, the water condition matrix time series is constructed.

3. The method for joint prediction of power generation of a hydropower station cluster according to claim 1 is characterized in that: The constructing of a prediction item of spatial impact of power generation based on the power generation matrix time series and the geographic distance data comprises: Constructing a hydropower station cluster spatial weight matrix based on the geographic distance data, wherein the larger the geographic distance data, the smaller the corresponding element value in the hydropower station cluster spatial weight matrix; Constructing a concurrent prediction item of spatial impact of power generation based on the current power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster; Based on the historical moment power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster, a power generation spatial impact lag prediction item is constructed.

4. The method for joint prediction of power generation of a hydropower station cluster according to claim 3 is characterized in that: The constructing of the concurrent prediction item of the spatial impact of power generation based on the current power generation matrix of the power generation matrix time series and the spatial weight matrix of the hydropower station cluster comprises: The current moment power generation matrix and the hydropower station cluster spatial weight matrix are merged to obtain a first fusion matrix; The first fusion matrix is ​​adjusted by using a first diagonal matrix to obtain a prediction item of the spatial impact of power generation in the same period; The constructing of the power generation spatial impact lag prediction item based on the historical power generation matrix of the power generation matrix time series and the hydropower station cluster spatial weight matrix comprises: The power generation matrix at the historical moment of the first lag order and the spatial weight matrix of the hydropower station cluster are fused to obtain a second fusion matrix; The second fusion matrix is ​​adjusted by using a second diagonal matrix to obtain the power generation spatial impact lag prediction item; The constructing of a power generation time impact prediction item based on the historical moment power generation matrix in the power generation matrix time series includes: Using the first regression coefficient matrix, the historical moment power generation matrix of the second lag order is adjusted to obtain the power generation time impact prediction item; The constructing of a water regime time impact prediction item based on the water regime matrix at a historical moment in the water regime matrix time series includes: The water condition matrix at the historical moment of the third lag order is adjusted using the second regression coefficient matrix to obtain the water condition time impact prediction item.

5. The method for joint prediction of power generation of a hydropower station cluster according to claim 4 is characterized in that: The step of training the matrix-valued time series spatial autoregressive model based on the data set to obtain a joint prediction model for power generation of the hydropower station cluster includes: Dividing the data set into a training set and a test set in time series according to a division ratio; Using the training set, the matrix-valued time series spatial autoregressive model is used to perform parameter estimation to obtain a trained matrix-valued time series spatial autoregressive model; The trained matrix-valued time series spatial autoregressive model is tested based on the test set, and the lag order is determined to obtain a joint prediction model for power generation of the hydropower station cluster.

6. The method for joint prediction of power generation of a hydropower station cluster according to claim 5 is characterized in that: The step of performing parameter estimation on the matrix-valued time series spatial autoregressive model to obtain a trained matrix-valued time series spatial autoregressive model includes: Acquire a plurality of preset lag order combinations, wherein the preset lag order combinations include a preset first lag order, a preset second lag order, and a preset third lag order; Using the training set, performing parameter estimation on the matrix-valued time series spatial autoregressive models corresponding to the plurality of preset lag order combinations to obtain a plurality of trained matrix-valued time series spatial autoregressive models; The step of testing the trained matrix-valued time series spatial autoregressive model based on the test set and determining the lag order to obtain a joint prediction model for power generation of the hydropower station cluster includes: Calculating model errors of the plurality of trained matrix-valued time series spatial autoregressive models based on the test set; The Bayesian information criterion value is calculated based on the model error, and the preset lag order combination when the Bayesian information criterion value is the smallest is selected to obtain the corresponding power generation joint prediction model.

7. A joint prediction device for power generation of a hydropower station cluster, characterized in that: The device comprises: The collection module is used to collect the historical power generation data of each hydropower station in the hydropower station cluster and multiple water conditions data related to power generation; collect the geographical distance data between the hydropower stations; A first construction module is used to construct a power generation matrix time series of the hydropower station cluster based on the historical power generation data of each hydropower station; to construct a water condition matrix time series of the hydropower station cluster based on the water condition data of each hydropower station; and to construct a data set based on the power generation matrix time series and the water condition matrix time series; The second construction module is used to construct a prediction item of spatial impact of power generation based on the power generation matrix time series and the geographic distance data, to construct a prediction item of time impact of power generation based on the power generation matrix at historical moments in the power generation matrix time series, and to construct a prediction item of time impact of water conditions based on the water conditions matrix at historical moments in the water conditions matrix time series; and to construct a spatial autoregressive model of matrix value time series by combining the prediction item of spatial impact of power generation, the prediction item of time impact of power generation and the prediction item of time impact of water conditions; A training module, used for training the matrix-valued time series spatial autoregressive model based on the data set to obtain a joint prediction model for power generation of the hydropower station cluster; A prediction module is used to predict the power generation of the hydropower station cluster based on the power generation joint prediction model.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for joint prediction of power generation of a hydropower station cluster according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for jointly predicting the power generation of a hydropower station cluster according to any one of claims 1 to 6.

10. A computer program product, characterized in that It comprises computer instructions, and the computer instructions are used to enable a computer to execute the method for jointly predicting the power generation of a hydropower station cluster according to any one of claims 1 to 6.

Images