Ultra-short-term photovoltaic power generation power prediction method and device based on short-term prediction result fusion

Abstract

The invention relates to the technical field of photovoltaic power generation power prediction, and discloses a short-term prediction result fused ultra-short-term photovoltaic power generation power prediction method and device, and the method comprises the steps: screening a target meteorological impact factor from a plurality of meteorological impact factors corresponding to historical meteorological data; according to the historical power data and first target meteorological data corresponding to the target meteorological influence factor, predicting the photovoltaic power generation power in the first preset time period; according to second target meteorological data corresponding to the target meteorological influence factor, predicting the photovoltaic power generation power in a second preset time period; and fusing the plurality of predicted first and second photovoltaic power generation power predicted values to obtain a plurality of target photovoltaic power generation power predicted values, according to the invention, the short-term photovoltaic power generation power and the ultra-short-term photovoltaic power generation power are predicted respectively, and the prediction results are fused to obtain the final prediction result, so that the accuracy of the final prediction result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation prediction, and in particular to an ultra-short-term photovoltaic power generation prediction method and device integrating short-term prediction results. Background Art

[0002] Photovoltaic power generation is a technology that uses solar energy to directly convert light energy into electrical energy through the semiconductor materials of solar cells. Due to the intermittent, random and fluctuating characteristics of photovoltaic power generation, it brings a series of problems to the safe operation of the power grid. As the proportion of photovoltaic power stations in the power grid structure increases, the photovoltaic power generation power prediction system becomes particularly important.

[0003] Photovoltaic power generation prediction technology is of great significance in the field of solar photovoltaic power generation. Photovoltaic power generation prediction technology needs to predict the photovoltaic power generation output power in an ultra-short time period in an environment with randomly changing meteorological conditions, thereby providing a reference for the stable operation and scheduling of the power grid.

[0004] Related art methods for ultra-short-term photovoltaic power output prediction rely on direct predictions based on historical power and meteorological data. However, this method only reflects ultra-short-term power fluctuations and lacks a comprehensive understanding of power trends. The prediction results are prone to significant fluctuations and cannot accurately reflect actual changes in photovoltaic power. Summary of the Invention

[0005] In view of this, the present invention provides an ultra-short-term photovoltaic power prediction method and device that integrates short-term prediction results to solve the problem of inaccurate photovoltaic power output power predicted in an ultra-short-term time period caused by the method for predicting photovoltaic power output power in the related art.

[0006] In a first aspect, the present invention provides an ultra-short-term photovoltaic power generation prediction method that integrates short-term prediction results, including: obtaining historical power data and historical meteorological data of a target photovoltaic power station, and screening at least one target meteorological influence factor that affects the ultra-short-term photovoltaic power generation prediction from multiple meteorological influence factors corresponding to the historical meteorological data based on the historical power data and the historical meteorological data; predicting the photovoltaic power generation power of a first preset time period based on first target meteorological data corresponding to the historical power data and at least one target meteorological influence factor, and obtaining multiple first photovoltaic power generation power prediction values ​​of the first preset time period; the first target meteorological data is the meteorological data of the first preset time period; predicting the photovoltaic power generation power of a second preset time period based on second target meteorological data corresponding to at least one target meteorological influence factor, and obtaining multiple second photovoltaic power generation power prediction values ​​of the second preset time period; the duration of the second preset time period is greater than the duration of the first preset time period; the second target meteorological data is the meteorological data of the second preset time period; and fusing multiple first photovoltaic power generation prediction values ​​with multiple second photovoltaic power generation prediction values ​​to obtain multiple target photovoltaic power generation prediction values.

[0007] The present invention obtains historical power data and historical meteorological data of a target photovoltaic power station. Based on the historical power data and historical meteorological data, at least one target meteorological influence factor that affects the ultra-short-term photovoltaic power generation power forecast is selected from multiple meteorological influence factors corresponding to the historical meteorological data. By analyzing the historical power data and historical meteorological data, the target meteorological influence factor is selected from the numerous meteorological influence factors, avoiding the use of irrelevant or redundant meteorological influence factors. The target meteorological influence factor that truly affects the photovoltaic power generation power can be obtained more accurately, thereby improving data utilization efficiency and reducing interference from invalid data. The present invention predicts the photovoltaic power generation power for a first preset time period based on first target meteorological data corresponding to the historical power data and at least one target meteorological influence factor, obtaining multiple first photovoltaic power generation power forecast values ​​for the first preset time period. The ultra-short-term photovoltaic power generation power forecast can keenly capture details of recent power changes. The present invention predicts the photovoltaic power generation power for a second preset time period based on second target meteorological data corresponding to at least one target meteorological influence factor, obtaining multiple second photovoltaic power generation power forecast values ​​for the second preset time period. The short-term photovoltaic power generation power forecast can capture macro power change trends. The present invention fuses multiple first photovoltaic power generation power prediction values ​​and multiple second photovoltaic power generation power prediction values ​​to obtain multiple target photovoltaic power generation power prediction values. By fusing multiple first photovoltaic power generation power prediction values ​​and multiple second photovoltaic power generation power prediction values, the present invention can not only keenly capture the details of recent power changes, but also capture the macro power change trend. Compared with related technologies, the present invention can give full play to the advantages of short-term photovoltaic power generation power prediction and ultra-short-term photovoltaic power generation power prediction, and make up for the shortcomings of single prediction. When facing sudden meteorological changes, the trend information of the short-term prediction can provide the overall direction for the ultra-short-term prediction, avoiding large deviations due to data noise or distribution changes. The detailed information of the ultra-short-term prediction can correct and supplement the short-term prediction, making the prediction result more consistent with the actual power changes. The present invention can more comprehensively and accurately reflect the actual changes in photovoltaic power generation, significantly improve the accuracy and stability of ultra-short-term power prediction, and meet the needs of real-time dispatch and operation management of the power system for high-precision prediction.

[0008] In an optional embodiment, based on historical power data and historical meteorological data, at least one target meteorological influence factor affecting the ultra-short-term photovoltaic power generation power prediction is screened from multiple meteorological influence factors corresponding to the historical meteorological data, including: determining the correlation coefficients between the multiple meteorological influence factors corresponding to the historical meteorological data and the historical power data respectively; based on the historical power data and the historical meteorological data, generating a meteorological factor time series graph of the multiple meteorological influence factors corresponding to the historical meteorological data and a power time series graph corresponding to the historical power data; selecting at least one target meteorological influence factor from the multiple meteorological influence factors whose correlation coefficient is greater than a first preset value and whose similarity between the changing trend of the meteorological factor time series graph and the changing trend of the power time series graph is greater than a second preset value.

[0009] The present invention generates a meteorological factor time series diagram and a power time series diagram by determining the correlation coefficient between historical meteorological influencing factors and historical power data. It can accurately measure the degree of correlation between historical meteorological influencing factors and historical power data in a quantitative manner, and can effectively exclude meteorological factors that have low or no correlation with photovoltaic power generation power, ensuring that the selected target meteorological influencing factors have a significant connection with power changes, thereby improving the accuracy of subsequent predictions.

[0010] In an optional embodiment, the photovoltaic power generation power of a first preset time period is predicted based on the historical power data and the first target meteorological data corresponding to at least one target meteorological influencing factor to obtain a plurality of first photovoltaic power generation power prediction values ​​of the first preset time period, including: inputting the historical power data and the first target meteorological data into a trained ultra-short-term photovoltaic power generation prediction model to obtain a plurality of first photovoltaic power generation power prediction values ​​of the first preset time period; the input of the ultra-short-term photovoltaic power generation prediction model is the historical power data and the first target meteorological data, and the output of the ultra-short-term photovoltaic power generation prediction model is a plurality of first photovoltaic power generation power prediction values; the ultra-short-term photovoltaic power generation prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

[0011] In an optional embodiment, the photovoltaic power generation power of a second preset time period is predicted based on the second target meteorological data corresponding to at least one target meteorological influencing factor to obtain a plurality of second photovoltaic power generation power prediction values ​​of the second preset time period, including: inputting the second target meteorological data into a trained short-term photovoltaic power generation power prediction model to obtain a plurality of second photovoltaic power generation power prediction values ​​of the second preset time period; the input of the short-term photovoltaic power generation power prediction model is historical power data and the second target meteorological data, and the output of the short-term photovoltaic power generation power prediction model is a plurality of second photovoltaic power generation power prediction values; the short-term photovoltaic power generation power prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

[0012] In an optional embodiment, the first photovoltaic power generation power prediction value and the second photovoltaic power generation power prediction value are integrated to obtain a target photovoltaic power generation power prediction value, including: screening multiple corresponding photovoltaic power generation power prediction values ​​in the first preset time period among multiple second photovoltaic power generation power prediction values ​​in the second preset time period; obtaining a first coefficient and a second coefficient; the first coefficient is used to reflect the degree of influence of the first photovoltaic power generation power prediction value on the actual power, and the second coefficient is used to reflect the degree of influence of the second photovoltaic power generation power prediction value on the actual power; according to the product of the first coefficient and each first photovoltaic power generation power prediction value, a plurality of first product results are obtained, and according to the product of the second coefficient and each second photovoltaic power generation power prediction value, a plurality of second product results are obtained; according to the sum of each first product result and the corresponding second product result, a plurality of target photovoltaic power generation power prediction values ​​are obtained.

[0013] The present invention introduces a first coefficient and a second coefficient to reflect the degree of influence of different predicted values ​​on actual power. By respectively calculating the product results of the coefficients and the corresponding predicted values ​​and summing them, the target photovoltaic power generation power prediction value is obtained in a weighted comprehensive manner. The weights can be reasonably allocated according to the degree of influence of different predicted values ​​on actual power, giving full play to the advantages of each predicted value, effectively reducing the prediction error, and improving the accuracy and reliability of the final prediction result, providing more valuable reference data for photovoltaic power station operation, power grid scheduling, etc.

[0014] In an optional embodiment, before predicting the photovoltaic power generation power for a first preset time period based on the first target meteorological data corresponding to the historical power data and at least one target meteorological influencing factor, the method also includes: determining whether the time period corresponding to the historical power data and the first preset time period are on the same day; if the time period corresponding to the historical power data and the first preset time period are not on the same day, selecting a target preset number of first target historical sample days from multiple historical sample days based on the first irradiance data of multiple historical sample days and the second irradiance data of the day to be predicted; updating the historical power data using the first target historical power data of the target preset number of first target historical sample days; and the historical power data for predicting the photovoltaic power generation power for the first preset time period is the updated historical power data.

[0015] In a second aspect, the present invention provides an ultra-short-term photovoltaic power generation prediction device that integrates short-term prediction results, including: a meteorological factor screening module for obtaining historical power data and historical meteorological data of a target photovoltaic power station, and screening at least one target meteorological influence factor that affects the ultra-short-term photovoltaic power generation prediction from multiple meteorological influence factors corresponding to the historical meteorological data based on the historical power data and the historical meteorological data; an ultra-short-term photovoltaic power generation prediction module for predicting the photovoltaic power generation power of a first preset time period based on first target meteorological data corresponding to the historical power data and at least one target meteorological influence factor, and obtaining multiple first photovoltaic power generation power prediction values ​​of the first preset time period; the first target meteorological data is the meteorological data of the first preset time period; a short-term photovoltaic power generation prediction module for predicting the photovoltaic power generation power of a second preset time period based on second target meteorological data corresponding to at least one target meteorological influence factor, and obtaining multiple second photovoltaic power generation power prediction values ​​of the second preset time period; the duration of the second preset time period is greater than the duration of the first preset time period; the second target meteorological data is the meteorological data of the second preset time period; a prediction fusion module for fusing multiple first photovoltaic power generation prediction values ​​and multiple second photovoltaic power generation prediction values ​​to obtain multiple target photovoltaic power generation prediction values.

[0016] 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 computer instructions to execute the ultra-short-term photovoltaic power prediction method integrating the short-term prediction results of the first aspect or any corresponding embodiment thereof.

[0017] 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 an ultra-short-term photovoltaic power prediction method integrating the short-term prediction results of the first aspect or any corresponding embodiment thereof.

[0018] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for enabling a computer to execute an ultra-short-term photovoltaic power prediction method integrating short-term prediction results of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in related technologies, the following briefly introduces the drawings required for use in the specific embodiments or related technical descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 4 is a flow chart of an ultra-short-term photovoltaic power generation prediction method by integrating short-term prediction results according to an embodiment of the present invention.

[0021] Figure 2 4 is a flow chart of another ultra-short-term photovoltaic power generation prediction process of integrating short-term prediction results according to an embodiment of the present invention.

[0022] Figure 3 4 is a flow chart of another ultra-short-term photovoltaic power generation prediction method by integrating short-term prediction results according to an embodiment of the present invention.

[0023] Figure 4 4 is a structural block diagram of an ultra-short-term photovoltaic power generation prediction device integrating short-term prediction results according to an embodiment of the present invention.

[0024] Figure 5 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying 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 making creative efforts shall fall within the scope of protection of the present invention.

[0026] Photovoltaic power generation is characterized by intermittency, randomness, and volatility, posing a series of challenges to the safe operation of power grids. As photovoltaic power stations increase their share of the grid's power supply structure, photovoltaic power generation prediction systems have become increasingly important. Photovoltaic power generation prediction technology is crucial in the field of solar photovoltaic power generation. It requires predicting photovoltaic power output over ultra-short-term periods in an environment characterized by randomly changing meteorological conditions, thereby providing a reference for stable grid operation and scheduling. Currently, machine learning and deep learning technologies are increasingly being used for ultra-short-term power prediction, but prediction accuracy still needs to be improved.

[0027] Short-term power forecasts (typically those for the next few hours to days) can provide relatively macroscopic information on power trends. However, existing ultra-short-term power forecasting methods often overlook the value of short-term forecast results and fail to effectively integrate short-term and ultra-short-term forecasts. This results in an incomplete grasp of power trends, making the forecast results prone to significant fluctuations and failing to accurately reflect actual changes in photovoltaic power generation.

[0028] An embodiment of the present invention provides an ultra-short-term photovoltaic power generation prediction method that integrates short-term prediction results. By predicting the short-term and ultra-short-term photovoltaic power generation powers separately, the prediction results are integrated to obtain a final prediction result, so as to achieve the effect of improving the accuracy of the final prediction result.

[0029] According to an embodiment of the present invention, an embodiment of an ultra-short-term photovoltaic power generation prediction method that integrates short-term prediction results 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.

[0030] In this embodiment, a short-term prediction result fusion ultra-short-term photovoltaic power generation prediction method is provided, which can be used for computer equipment. Figure 1 FIG. 1 is a flow chart of an ultra-short-term photovoltaic power generation prediction method based on short-term prediction results fusion according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0031] Step S101 , obtaining historical power data and historical meteorological data of a target photovoltaic power station, and screening at least one target meteorological influencing factor affecting ultra-short-term photovoltaic power generation power prediction from a plurality of meteorological influencing factors corresponding to the historical meteorological data based on the historical power data and the historical meteorological data.

[0032] Among them, the historical power data is the power generation data at different time points within the historical time period. For example, the historical power data can be the power generation data at the past 96 time points. The 96 time points can be obtained by dividing the time of the past day into 15-minute time intervals. The historical power data includes actual power, maximum power that can be generated, etc.; the historical meteorological data is the meteorological data at different time points within the historical time period, including irradiance, humidity, temperature, etc.

[0033] In some optional embodiments, after obtaining the historical power data and historical meteorological data of the target photovoltaic power station, the historical power data and historical meteorological data of the target photovoltaic power station are preprocessed by missing value interpolation, outlier processing, etc., and data normalization is performed. The data subsequently screened are the historical power data and historical meteorological data after preprocessing and normalization.

[0034] Specifically, the outlier processing method includes: using the DPC (Density Peaks Clustering) algorithm to perform preliminary clustering on the data; extracting the representative points of each cluster (such as core points or a combination of core points and boundary points) from the results of the DPC algorithm as the initial clustering center of the FCM (Fuzzy C-Means) algorithm. After determining the initial clustering center, the FCM algorithm is used for subsequent clustering to obtain the clustering results; judging whether the data point is an outlier based on the clustering results. If the data point does not belong to any cluster or the number of data points in the cluster it belongs to is less than a preset threshold, the data point is judged as an outlier and is deleted or replaced.

[0035] In some optional embodiments, based on historical power data and historical meteorological data, at least one target meteorological influence factor that affects the ultra-short-term photovoltaic power generation power prediction is screened from multiple meteorological influence factors corresponding to the historical meteorological data, including: determining the correlation coefficients between the multiple meteorological influence factors corresponding to the historical meteorological data and the historical power data respectively; based on the historical power data and the historical meteorological data, generating a meteorological factor time series graph of the multiple meteorological influence factors corresponding to the historical meteorological data and a power time series graph corresponding to the historical power data; selecting at least one target meteorological influence factor from the multiple meteorological influence factors whose correlation coefficient is greater than a first preset value and whose similarity between the changing trend of the meteorological factor time series graph and the changing trend of the power time series graph is greater than a second preset value.

[0036] The first preset value and the second preset value may be set according to actual conditions. For example, the first preset value may be 0.7, and the second preset value may be 0.6.

[0037] Step S102: predicting the photovoltaic power generation power of a first preset time period based on the historical power data and the first target meteorological data corresponding to at least one target meteorological influencing factor, and obtaining a plurality of first photovoltaic power generation power prediction values ​​of the first preset time period; the first target meteorological data is the meteorological data of the first preset time period.

[0038] Among them, the first preset time period can be set according to actual needs. For example, the first preset time period can be 16 time points after a certain moment in the future. The 16 time points can be started from a certain moment in the future, and one time point is obtained every 15 minutes to obtain 16 time points.

[0039] In the embodiment of the present invention, the multiple first photovoltaic power generation prediction values ​​refer to the first photovoltaic power generation prediction values ​​corresponding to multiple time points within the first preset time period, and one first photovoltaic power generation prediction value corresponding to one time point.

[0040] In some optional embodiments, before predicting the photovoltaic power generation power for a first preset time period based on the first target meteorological data corresponding to the historical power data and at least one target meteorological influencing factor, the method also includes: determining whether the time period corresponding to the historical power data and the first preset time period are on the same day; if the time period corresponding to the historical power data and the first preset time period are not on the same day, selecting a target preset number of first target historical sample days from multiple historical sample days based on the first irradiance data of multiple historical sample days and the second irradiance data of the day to be predicted; updating the historical power data using the first target historical power data of the target preset number of first target historical sample days; and the historical power data for predicting the photovoltaic power generation power for the first preset time period is the updated historical power data.

[0041] Among them, based on the first irradiance data of multiple historical sample days and the second irradiance data of the day to be predicted, a target preset number of first target historical sample days are selected from the multiple historical sample days, including: calculating the Pearson correlation coefficient between the first irradiance data of each historical sample day and the second irradiance data of the day to be predicted, and selecting the target preset number of first target historical sample days whose Pearson correlation coefficient is greater than the preset correlation coefficient from the multiple historical sample days.

[0042] Among them, the target preset number can be 3.

[0043] In some optional embodiments, the historical power data is updated using the first target historical power data of the first target historical sample day of a target preset number, including: according to the time period corresponding to the historical power data and the first preset time period, obtaining multiple first target time points in the time period corresponding to the historical power data that are not on the same day as the first preset time period; taking a weighted average of the first target historical power data corresponding to each first target time point in the first target historical sample day of the target preset number, and replacing the data corresponding to the first target time point in the historical power data using the weighted average price result to update the historical power data.

[0044] For example, when the historical power data of 96 historical time points are used as input variables, these 96 time points may be on two days apart from the 16 time points to be predicted (i.e., they do not belong to the same day). At this time, if there is a large difference between the meteorological conditions of the sample day where the historical power data is located and the day to be predicted, it will have a negative impact on the power forecast. To this end, an embodiment of the present invention uses the first irradiance data of the historical sample day n days before the day to be predicted and the irradiance data of 96 time points on the day to be predicted to perform similar day calculations, and selects the power data of the time points corresponding to the three days whose first irradiance data in the historical sample days before the n days is most similar to the second irradiance data of the 96 time points on the day to be predicted, and performs weighted averaging to replace the power of all time points in the time range that are not on the same day as the day to be predicted in the 96 time points of the input variable, so as to update the historical power data.

[0045] The calculation process of the first target historical sample day is as follows: using the Pearson correlation coefficient, calculate the similarity of the irradiance data between the predicted day and the previous n historical sample days, and finally select the three most similar days as the first target historical sample days.

[0046] Among them, the weight calculation method for the fusion of similar day data is: based on the training set data and the above-mentioned calculation process of the first target historical sample day, the power data at the corresponding moment of the three-day first target historical sample day corresponding to each time point is calculated, and then a multivariate linear regression equation is constructed, and the corresponding standardized regression coefficient is used as the weight.

[0047] In some optional embodiments, the photovoltaic power generation power of a first preset time period is predicted based on the historical power data and the first target meteorological data corresponding to at least one target meteorological influencing factor to obtain a plurality of first photovoltaic power generation power prediction values ​​of the first preset time period, including: inputting the historical power data and the first target meteorological data into a trained ultra-short-term photovoltaic power generation power prediction model to obtain a plurality of first photovoltaic power generation power prediction values ​​of the first preset time period; the input of the ultra-short-term photovoltaic power generation power prediction model is the historical power data and the first target meteorological data, and the output of the ultra-short-term photovoltaic power generation power prediction model is a plurality of first photovoltaic power generation power prediction values; the ultra-short-term photovoltaic power generation power prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

[0048] Among them, the ultra-short-term photovoltaic power generation prediction model is a neural network model constructed based on a bidirectional long short-term memory network (Bidirectional Long Short-Term Memory, BiLSTM) and a multi-layer perceptron (Multi-Layer Perceptron, MLP). Specifically, the ultra-short-term photovoltaic power generation prediction model uses the output of BiLSTM as the input of MLP, and MLP processes these inputs and generates multiple first photovoltaic power generation power prediction values ​​for the final first preset time period.

[0049] In some optional implementations, BiLSTM can simultaneously process the forward and reverse information of the sequence, and can take into account the previous and subsequent information when processing each time step, thereby more comprehensively understanding the contextual relationship in the sequence and improving the accuracy and robustness of the ultra-short-term photovoltaic power generation prediction model. After receiving the output of the bidirectional BiLSTM, the MLP can further perform nonlinear transformation on this information to extract deeper features, thereby obtaining multiple first photovoltaic power generation prediction values.

[0050] Among them, the forward LSTM includes: forward LSTM input gate, forward LSTM forget gate, forward LSTM cell state update, forward LSTM cell state, forward LSTM output gate, forward LSTM hidden state. Specifically, the formula of the forward LSTM input gate is:

[0051] i t =σ(Wi ix *x t +W ih *h t-1 +b i )

[0052] Among them, i t is the input gate value of the forward LSTM at time step t, x t is the vector representation of the t-th time step of the forward LSTM input sequence, σ() is the activation function of the forward LSTM, and W ix Input vector x to the forward LSTM t The weight matrix to the input gate, W ih is the forward LSTM hidden state h t-1 The weight matrix to the input gate, h t-1 is the hidden state of the forward LSTM at time step t-1, b i is the bias vector of the forward LSTM input gate.

[0053] In some optional implementations, the formula for the forward LSTM forget gate is:

[0054] f t=σ(W fx *x t +W fh *h t-1 +b f )

[0055] Among them, f t is the forget gate value of the forward LSTM at time step t, x t is the vector representation of the t-th time step of the forward LSTM input sequence, σ() is the activation function of the forward LSTM, and W fx Input vector x to the forward LSTM t To the weight matrix of the forget gate, W fh is the forward LSTM hidden state h t-1 To the weight matrix of the forget gate, h t-1 is the hidden state of the forward LSTM at time step t-1, b f is the bias vector of the forward LSTM forget gate.

[0056] In some optional implementations, the formula for updating the forward LSTM cell state is:

[0057]

[0058] in, is the candidate cell state of the forward LSTM at time step t, x t is the vector representation of the t-th time step of the forward LSTM input sequence, tanh() is the activation function of the forward LSTM, W cx Input vector x to the forward LSTM t to the weight matrix of the candidate cell state, W ch is the forward LSTM hidden state h t-1 to the weight matrix of the candidate cell state, h t-1 is the hidden state of the forward LSTM at time step t-1, b c is the bias vector of the forward LSTM candidate cell state.

[0059] In some optional implementations, the formula for the forward LSTM cell state is:

[0060]

[0061] Among them, C t is the cell state of the forward LSTM at time step t, f t is the forget gate value of the forward LSTM at time step t, C t-1 is the cell state of the forward LSTM at time step t-1, i t is the input gate value of the forward LSTM at time step t, is the candidate cell state of the forward LSTM at time step t.

[0062] In some optional implementations, the formula for the forward LSTM output gate is:

[0063] o t =σ(W ox *x t +W oh *h t-1 +b o )

[0064] Among them, t is the output gate value of the forward LSTM at time step t, x t is the vector representation of the t-th time step of the forward LSTM input sequence, σ() is the activation function of the forward LSTM, and W ox Input vector x to the forward LSTM t The weight matrix to the output gate, W oh is the forward LSTM hidden state h t-1 The weight matrix to the output gate, h t-1 is the hidden state of the forward LSTM at time step t-1, b o is the bias vector of the forward LSTM output gate.

[0065] In some optional implementations, the formula for the forward LSTM hidden state is:

[0066]

[0067] Among them, h t is the hidden state of the forward LSTM at time step t, o t is the output gate value of the forward LSTM at time step t, C t is the cell state of the forward LSTM at time step t, and tanh() is the activation function of the forward LSTM.

[0068] In some optional embodiments, the backward LSTM includes: a backward LSTM input gate, a backward LSTM forget gate, a backward LSTM cell state update, a backward LSTM cell state, a backward LSTM output gate, and a backward LSTM hidden state. Specifically, the formula of the backward LSTM input gate is:

[0069] i t '=σ(W ix '*x t +W ih '*h t-1 '+b i ')

[0070] Among them, i t' is the input gate value of LSTM at time step t, x t is the vector representation of the t-th time step of the backward LSTM input sequence, σ() is the activation function of the backward LSTM, and W ih 'Input vector x to the backward LSTM t The weight matrix to the input gate, W ih ' is the backward LSTM hidden state h t-1 'The weight matrix to the input gate, h t-1 ' is the hidden state of the backward LSTM at time step t-1, b i ' is the bias vector of the backward LSTM input gate.

[0071] In some optional implementations, the formula for the backward LSTM forget gate is:

[0072] f t '=σ(W fx '*x t +W fh '*h t-1 '+b f ')

[0073] Among them, f t ' is the forget gate value of the backward LSTM at time step t, x t is the vector representation of the t-th time step of the backward LSTM input sequence, σ() is the activation function of the backward LSTM, and W fx 'Input vector x to the backward LSTM t To the weight matrix of the forget gate, W fh ' is the backward LSTM hidden state h t-1 'To the weight matrix of the forget gate, h t-1 ' is the hidden state of the backward LSTM at time step t-1, b f ' is the bias vector of the backward LSTM forget gate.

[0074] In some optional implementations, the formula for updating the backward LSTM cell state is:

[0075]

[0076] in, is the candidate cell state of the backward LSTM at time step t, x t is the vector representation of the t-th time step of the backward LSTM input sequence, tanh() is the activation function of the backward LSTM, and W cx 'Input vector x to the backward LSTM t to the weight matrix of the candidate cell state, W ch ' is the backward LSTM hidden state h t-1'To the weight matrix of the candidate cell state, h t-1 ' is the hidden state of the backward LSTM at time step t-1, b c ' is the bias vector of the backward LSTM candidate cell state.

[0077] In some optional implementations, the formula for the backward LSTM cell state is:

[0078]

[0079] Among them, C t ' is the cell state of the backward LSTM at time step t, f t ' is the forget gate value of the backward LSTM at time step t, C t-1 'The cell state of the backward LSTM at time step t-1, i t ' is the input gate value of the backward LSTM at time step t, is the candidate cell state of the backward LSTM at time step t.

[0080] In some optional implementations, the formula for the backward LSTM output gate is:

[0081] o t '=σ(W ox '*x t +W oh '*h t-1 '+b o ')

[0082] Among them, t ' is the output gate value of the backward LSTM at time step t, x t is the vector representation of the t-th time step of the backward LSTM input sequence, σ() is the activation function of the backward LSTM, and W ox 'Input vector x to the backward LSTM t The weight matrix to the output gate, W oh ' is the backward LSTM hidden state h t-1 'The weight matrix to the output gate, h t-1 ' is the hidden state of the backward LSTM at time step t-1, b o ' is the bias vector of the backward LSTM output gate.

[0083] In some optional implementations, the formula for the backward LSTM hidden state is:

[0084]

[0085] Among them, h t ' is the hidden state of the backward LSTM at time step t, o t' is the output gate value of the backward LSTM at time step t, C t ' is the cell state of the backward LSTM at time step t, and tanh() is the activation function of the backward LSTM.

[0086] In this embodiment of the present invention, an MLP is a feedforward neural network that performs multiple transformations on input data using a weight matrix and activation function to learn nonlinear relationships in the data. The MLP consists of an input layer, a hidden layer, and an output layer, where every neuron in the previous layer is connected to every neuron in the next layer.

[0087] Among them, the output of the qth layer is:

[0088] y q =σ(W q *y q-1 +b q )

[0089] Among them, y q is the output vector of the qth layer of MLP, σ() is the activation function of MLP, W q is the weight of the qth layer of MLP, y q-1 is the output vector of the q-1th layer of the MLP, b q is the bias vector of the qth layer of the MLP.

[0090] In some optional embodiments, the embodiments of the present invention also include a training process of an ultra-short-term photovoltaic power generation prediction model. The training process of the ultra-short-term photovoltaic power generation prediction model includes: inputting historical training power data and meteorological data at the time to be predicted into the ultra-short-term photovoltaic power generation prediction model for training.

[0091] Step S103, based on the second target meteorological data corresponding to at least one target meteorological influencing factor, the photovoltaic power generation power of the second preset time period is predicted to obtain multiple second photovoltaic power generation power prediction values ​​of the second preset time period; the duration of the second preset time period is greater than the duration of the first preset time period; the second target meteorological data is the meteorological data of the second preset time period.

[0092] In some optional embodiments, the photovoltaic power generation power of a second preset time period is predicted based on the second target meteorological data corresponding to at least one target meteorological influencing factor to obtain a plurality of second photovoltaic power generation power prediction values ​​of the second preset time period, including: inputting the second target meteorological data into a trained short-term photovoltaic power generation power prediction model to obtain a plurality of second photovoltaic power generation power prediction values ​​of the second preset time period; the input of the short-term photovoltaic power generation power prediction model is historical power data and the second target meteorological data, and the output of the short-term photovoltaic power generation power prediction model is a plurality of second photovoltaic power generation power prediction values; the short-term photovoltaic power generation power prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

[0093] Among them, the second preset time period can be set according to actual needs. For example, the second preset time period can be 96 time points after a certain moment in the future. The 96 time points can be started from a certain moment in the future, and one time point is obtained every 15 minutes to obtain 96 time points.

[0094] In the embodiment of the present invention, the multiple second photovoltaic power generation prediction values ​​refer to the second photovoltaic power generation prediction values ​​corresponding to multiple time points within the second preset time period, and one second photovoltaic power generation prediction value corresponding to one time point.

[0095] In the embodiment of the present invention, the construction process of the short-term photovoltaic power generation prediction model and the ultra-short-term photovoltaic power generation prediction model are similar and will not be described in detail here.

[0096] In some optional embodiments, the embodiments of the present invention also include a training process of a short-term photovoltaic power generation prediction model. The training process of the short-term photovoltaic power generation prediction model includes: inputting training meteorological data corresponding to at least one target meteorological influencing factor into the short-term photovoltaic power generation prediction model for training.

[0097] Step S104 : fusing the plurality of first photovoltaic power generation prediction values ​​and the plurality of second photovoltaic power generation prediction values ​​to obtain a plurality of target photovoltaic power generation prediction values.

[0098] In some optional embodiments, multiple first photovoltaic power generation power prediction values ​​and multiple second photovoltaic power generation power prediction values ​​are fused to obtain multiple target photovoltaic power generation power prediction values, including: inputting multiple first photovoltaic power generation power prediction values ​​and multiple second photovoltaic power generation power prediction values ​​into a trained artificial neural network model to obtain multiple target photovoltaic power generation power prediction values.

[0099] In some optional embodiments, the first photovoltaic power generation power prediction value and the second photovoltaic power generation power prediction value are integrated to obtain a target photovoltaic power generation power prediction value, including: screening multiple corresponding photovoltaic power generation power prediction values ​​in the first preset time period among multiple second photovoltaic power generation power prediction values ​​in the second preset time period; obtaining a first coefficient and a second coefficient; the first coefficient is used to reflect the degree of influence of the first photovoltaic power generation power prediction value on the actual power, and the second coefficient is used to reflect the degree of influence of the second photovoltaic power generation power prediction value on the actual power; according to the product of the first coefficient and each first photovoltaic power generation power prediction value, a plurality of first product results are obtained, and according to the product of the second coefficient and each second photovoltaic power generation power prediction value, a plurality of second product results are obtained; according to the sum of each first product result and the corresponding second product result, a plurality of target photovoltaic power generation power prediction values ​​are obtained.

[0100] The ultra-short-term photovoltaic power generation prediction method provided by this embodiment integrates short-term prediction results. The method obtains historical power data and historical meteorological data of a target photovoltaic power station. Based on the historical power data and historical meteorological data, at least one target meteorological influence factor that affects the ultra-short-term photovoltaic power generation prediction is selected from multiple meteorological influence factors corresponding to the historical meteorological data. By analyzing the historical power data and historical meteorological data, the target meteorological influence factor is selected from the numerous meteorological influence factors, avoiding the use of irrelevant or redundant meteorological influence factors. The target meteorological influence factor that truly affects photovoltaic power generation can be obtained more accurately, thereby improving data utilization efficiency and reducing interference from invalid data. The embodiment of the present invention predicts photovoltaic power generation for a first preset time period based on first target meteorological data corresponding to the historical power data and at least one target meteorological influence factor, obtaining multiple first photovoltaic power generation prediction values ​​for the first preset time period. The ultra-short-term photovoltaic power generation prediction can keenly capture details of recent power changes. The embodiment of the present invention predicts photovoltaic power generation for a second preset time period based on second target meteorological data corresponding to at least one target meteorological influence factor, obtaining multiple second photovoltaic power generation prediction values ​​for the second preset time period. The short-term photovoltaic power generation prediction can capture macro power change trends. The embodiment of the present invention fuses multiple first photovoltaic power generation power prediction values ​​with multiple second photovoltaic power generation power prediction values ​​to obtain multiple target photovoltaic power generation power prediction values. The fusion of multiple first photovoltaic power generation power prediction values ​​with multiple second photovoltaic power generation power prediction values ​​can not only keenly capture the details of recent power changes, but also capture the macro power change trends. Compared with related technologies, the embodiment of the present invention can fully leverage the advantages of short-term photovoltaic power generation prediction and ultra-short-term photovoltaic power generation prediction, and make up for the shortcomings of single prediction. In the face of sudden meteorological changes, the trend information of the short-term prediction can provide an overall direction for the ultra-short-term prediction, avoiding large deviations due to data noise or distribution changes. The detailed information of the ultra-short-term prediction can also correct and supplement the short-term prediction, making the prediction results more consistent with actual power changes. The present invention can more comprehensively and accurately reflect the actual changes in photovoltaic power generation, significantly improve the accuracy and stability of ultra-short-term power prediction, and meet the demand for high-precision prediction for real-time dispatch and operation management of power systems.

[0101] In this embodiment, a short-term prediction result fusion ultra-short-term photovoltaic power generation prediction method is provided, which can be used for computer equipment. Figure 2 FIG. 1 is a flow chart of another ultra-short-term photovoltaic power generation prediction method according to an embodiment of the present invention, wherein Figure 2 As shown, the process includes the following steps:

[0102] Step S201 : acquiring historical power data and historical meteorological data of a target photovoltaic power station, and screening at least one target meteorological influencing factor affecting ultra-short-term photovoltaic power generation power prediction from a plurality of meteorological influencing factors corresponding to the historical meteorological data based on the historical power data and the historical meteorological data.

[0103] Specifically, the above step S201 includes:

[0104] Step S2021 : determining correlation coefficients between a plurality of meteorological impact factors corresponding to the historical meteorological data and the historical power data.

[0105] Among them, the Pearson correlation coefficient method is used to determine the correlation coefficients between multiple meteorological influencing factors corresponding to historical meteorological data and historical power data.

[0106] For example, the calculation process of the correlation coefficient is:

[0107]

[0108] Among them, r i is the correlation coefficient between the ith meteorological influencing factor and the historical power data, K is the total number of meteorological influencing factors, P i is the i-th group of historical power data, is the mean value of the historical power data of group i, M i is the historical meteorological data corresponding to the i-th meteorological impact factor, is the mean of the historical meteorological data corresponding to the i-th meteorological impact factor.

[0109] Step S2022 : generating, based on the historical power data and the historical meteorological data, a meteorological factor time series diagram of a plurality of meteorological influencing factors corresponding to the historical meteorological data and a power time series diagram corresponding to the historical power data.

[0110] Step S2023 , selecting at least one target meteorological impact factor from the plurality of meteorological impact factors, the correlation coefficient of which is greater than a first preset value, and the similarity between the changing trend of the meteorological factor time series graph and the changing trend of the power time series graph is greater than a second preset value.

[0111] In an embodiment of the present invention, a method for determining the similarity between the changing trend of the meteorological factor time series graph and the changing trend of the power time series graph is to use a convolutional neural network to extract deep features from the meteorological factor time series graph and the power time series graph respectively, input the extracted features into a fully connected layer, etc., and determine the similarity of the changing trends of the two sequences by calculating the cosine similarity between the feature vectors, etc.

[0112] Step S202: Based on the historical power data and first target meteorological data corresponding to at least one target meteorological influencing factor, the photovoltaic power generation power in the first preset time period is predicted to obtain a plurality of first photovoltaic power generation power prediction values ​​in the first preset time period; the first target meteorological data is the meteorological data in the first preset time period. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0113] Step S203: Based on the second target meteorological data corresponding to the at least one target meteorological influencing factor, the photovoltaic power generation power in the second preset time period is predicted to obtain a plurality of second photovoltaic power generation power prediction values ​​in the second preset time period; the second preset time period is longer than the first preset time period; the second target meteorological data is the meteorological data in the second preset time period. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0114] Step S204 : fusing the plurality of first photovoltaic power generation prediction values ​​and the plurality of second photovoltaic power generation prediction values ​​to obtain a plurality of target photovoltaic power generation prediction values.

[0115] Specifically, the above step S204 includes:

[0116] Step S2041 : Filtering a plurality of corresponding photovoltaic power prediction values ​​in the first preset time period from a plurality of second photovoltaic power prediction values ​​in the second preset time period.

[0117] Step S2042, obtaining a first coefficient and a second coefficient; the first coefficient is used to reflect the degree of influence of the first photovoltaic power generation power prediction value on the actual power, and the second coefficient is used to reflect the degree of influence of the second photovoltaic power generation power prediction value on the actual power.

[0118] Among them, obtaining the first coefficient and the second coefficient includes: inputting multiple groups of historical power training set data and multiple groups of historical meteorological training set data of the target photovoltaic power station into the ultra-short-term photovoltaic power generation power prediction model to obtain multiple first training power prediction values; inputting multiple groups of historical power training set data and multiple groups of historical meteorological training set data of the target photovoltaic power station into the short-term photovoltaic power generation power prediction model to obtain multiple second training power prediction values; using the first training power prediction value and the second training power prediction value as input and the actual power as output, constructing a linear regression equation, solving the linear regression equation, and determining the first coefficient and the second coefficient.

[0119] Step S2043: obtain multiple first product results based on the product of the first coefficient and each first photovoltaic power generation power prediction value, and obtain multiple second product results based on the product of the second coefficient and each second photovoltaic power generation power prediction value.

[0120] Step S2044 , obtaining a plurality of target photovoltaic power prediction values ​​according to the sum of each first multiplication result and the corresponding second multiplication result.

[0121] The ultra-short-term photovoltaic power generation prediction method provided by this embodiment generates a meteorological factor time series diagram and a power time series diagram by determining the correlation coefficient between the historical meteorological influencing factors and the historical power data. It can accurately measure the degree of correlation between the historical meteorological influencing factors and the historical power data in a quantitative manner, effectively exclude meteorological factors that have low or no correlation with photovoltaic power generation, ensure that the selected target meteorological influencing factors have a significant connection with power changes, and improve the accuracy of subsequent predictions. The embodiment of the present invention introduces a first coefficient and a second coefficient to reflect the degree of influence of different predicted values ​​on actual power. By calculating the product of the coefficient and the corresponding predicted value respectively and summing them, the target photovoltaic power generation power prediction value is obtained in a weighted comprehensive manner. The weight can be reasonably allocated according to the degree of influence of different predicted values ​​on actual power, giving full play to the advantages of each predicted value, effectively reducing the prediction error, and improving the accuracy and reliability of the final prediction result, providing more valuable reference data for photovoltaic power station operation, power grid scheduling, etc.

[0122] In this embodiment, a short-term prediction result fusion ultra-short-term photovoltaic power generation prediction method is provided, which can be used for computer equipment. Figure 3 FIG. 1 is a flow chart of another ultra-short-term photovoltaic power generation prediction method based on the fusion of short-term prediction results according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps:

[0123] Step S301: collect historical power data and historical meteorological data of the target photovoltaic power station, perform pre-processing such as missing value interpolation and outlier processing on the historical power data and historical meteorological data, and perform data normalization. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0124] Step S302: Based on the Pearson correlation coefficient method and by drawing a time series diagram of meteorological factors, the target meteorological factors that affect the photovoltaic power generation power forecast are screened. Figure 1 Steps S101 and S102 of the embodiment shown Figure 2 Step S201 of the illustrated embodiment will not be described in detail here.

[0125] Step S303: Use the updated historical power data of 96 time points and the meteorological data at the time to be predicted as model input to construct an ultra-short-term photovoltaic power generation prediction model. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0126] Step S304: Use the ultra-short-term photovoltaic power generation prediction model to predict the ultra-short-term photovoltaic power generation. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0127] Step S305: Use the short-term photovoltaic power generation prediction model to predict the short-term photovoltaic power generation. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0128] Step S306: The short-term power prediction value and the ultra-short-term power prediction value at each moment are combined as the final power prediction result. Figure 1 Steps S104 and S106 of the embodiment shown Figure 2 Step S204 of the illustrated embodiment will not be described in detail here.

[0129] In this embodiment, an ultra-short-term photovoltaic power generation prediction device that integrates short-term prediction results is also provided. The device is used to implement the above-mentioned embodiments and preferred implementation methods, and the details that have been described 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 devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and conceivable.

[0130] This embodiment provides an ultra-short-term photovoltaic power generation prediction device that integrates short-term prediction results. Figure 4 Shown, including:

[0131] The meteorological factor screening module 401 is used to obtain historical power data and historical meteorological data of the target photovoltaic power station, and based on the historical power data and historical meteorological data, screen at least one target meteorological influencing factor that affects the ultra-short-term photovoltaic power generation power forecast from multiple meteorological influencing factors corresponding to the historical meteorological data.

[0132] The ultra-short-term photovoltaic power generation prediction module 402 is used to predict the photovoltaic power generation power in a first preset time period based on the historical power data and the first target meteorological data corresponding to at least one target meteorological influencing factor, and obtain multiple first photovoltaic power generation power prediction values ​​in the first preset time period; the first target meteorological data is the meteorological data of the first preset time period.

[0133] The short-term photovoltaic power generation prediction module 403 is used to predict the photovoltaic power generation power in a second preset time period based on the second target meteorological data corresponding to at least one target meteorological influencing factor, and obtain multiple second photovoltaic power generation power prediction values ​​for the second preset time period; the duration of the second preset time period is greater than the duration of the first preset time period; the second target meteorological data is the meteorological data of the second preset time period.

[0134] The prediction fusion module 404 is configured to fuse the plurality of first photovoltaic power generation prediction values ​​and the plurality of second photovoltaic power generation prediction values ​​to obtain a plurality of target photovoltaic power generation prediction values.

[0135] In some optional embodiments, the meteorological factor screening module 401 includes:

[0136] The correlation coefficient determination unit is used to determine the correlation coefficients between the multiple meteorological impact factors corresponding to the historical meteorological data and the historical power data.

[0137] The sequence diagram determining unit is used to generate a meteorological factor time series diagram of multiple meteorological influencing factors corresponding to the historical meteorological data and a power time series diagram corresponding to the historical power data based on the historical power data and the historical meteorological data.

[0138] The target meteorological factor determination unit is used to select at least one target meteorological influence factor from multiple meteorological influence factors, whose correlation coefficient is greater than a first preset value and whose similarity between the change trend of the meteorological factor time series diagram and the change trend of the power time series diagram is greater than a second preset value.

[0139] In some optional implementations, the ultra-short-term photovoltaic power generation prediction module 402 includes:

[0140] An ultra-short-term model prediction unit is used to input historical power data and first target meteorological data into a trained ultra-short-term photovoltaic power generation prediction model to obtain multiple first photovoltaic power generation power prediction values ​​for a first preset time period; the input of the ultra-short-term photovoltaic power generation prediction model is historical power data and first target meteorological data, and the output of the ultra-short-term photovoltaic power generation prediction model is multiple first photovoltaic power generation power prediction values; the ultra-short-term photovoltaic power generation prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

[0141] In some optional implementations, the short-term photovoltaic power generation prediction module 403 includes:

[0142] The short-term model prediction unit is used to input the second target meteorological data into the trained short-term photovoltaic power generation power prediction model to obtain multiple second photovoltaic power generation power prediction values ​​for the second preset time period; the input of the short-term photovoltaic power generation power prediction model is historical power data and the second target meteorological data, and the output of the short-term photovoltaic power generation power prediction model is multiple second photovoltaic power generation power prediction values; the short-term photovoltaic power generation power prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

[0143] In some optional implementations, the prediction fusion module 404 includes:

[0144] The prediction value screening unit is used to screen a plurality of corresponding photovoltaic power generation prediction values ​​in the first preset time period from a plurality of second photovoltaic power generation prediction values ​​in the second preset time period.

[0145] The coefficient acquisition unit is used to obtain a first coefficient and a second coefficient; the first coefficient is used to reflect the degree of influence of the first photovoltaic power generation power prediction value on the actual power, and the second coefficient is used to reflect the degree of influence of the second photovoltaic power generation power prediction value on the actual power.

[0146] The product calculation unit is used to obtain multiple first product results according to the product of the first coefficient and each first photovoltaic power generation power prediction value, and to obtain multiple second product results according to the product of the second coefficient and each second photovoltaic power generation power prediction value.

[0147] The prediction fusion unit is used to obtain multiple target photovoltaic power generation power prediction values ​​according to the sum of each first product result and the corresponding second product result.

[0148] In some optional embodiments, the ultra-short-term photovoltaic power generation prediction device also includes: a first historical data updating module, used to determine whether the time period corresponding to the historical power data and the first preset time period are on the same day; if the time period corresponding to the historical power data and the first preset time period are not on the same day, then based on the first irradiance data of multiple historical sample days and the second irradiance data of the day to be predicted, a target preset number of first target historical sample days are selected from multiple historical sample days; the historical power data is updated using the first target historical power data of the target preset number of first target historical sample days; the historical power data for photovoltaic power generation power prediction for the first preset time period is the updated historical power data.

[0149] In some optional implementations, the ultra-short-term photovoltaic power generation prediction device further includes:

[0150] The second historical data updating module is used to determine whether the time period corresponding to the historical power data and the second preset time period are on the same day; if the time period corresponding to the historical power data and the second preset time period are not on the same day, then based on the first irradiance data of multiple historical sample days and the third irradiance data of the day to be predicted, a target preset number of second target historical sample days are selected from multiple historical sample days; the historical power data is updated using the second target historical power data of the target preset number of second target historical sample days; the historical power data for photovoltaic power generation power prediction in the second preset time period is the updated historical power data.

[0151] 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.

[0152] The ultra-short-term photovoltaic power generation prediction device that integrates short-term prediction results 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.

[0153] The embodiment of the present invention also provides a computer device having the above Figure 4 The short-term prediction results shown are integrated into the ultra-short-term photovoltaic power prediction device.

[0154] See also Figure 5 , Figure 5 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 5 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other 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 the memory 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. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0155] 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 an application-specific 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.

[0156] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0157] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on 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 located 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.

[0158] 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.

[0159] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0160] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, 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 drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. 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.

[0161] A portion 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 form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. 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 that can be accessed by the computer.

[0162] Although the embodiments of the present invention have been described with reference to 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. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for ultra-short-term photovoltaic power generation prediction based on the fusion of short-term prediction results, characterized in that: The method comprises: Acquiring historical power data and historical meteorological data of a target photovoltaic power station, and selecting, based on the historical power data and the historical meteorological data, at least one target meteorological influencing factor that affects ultra-short-term photovoltaic power generation prediction from a plurality of meteorological influencing factors corresponding to the historical meteorological data; Predicting the photovoltaic power generation power in a first preset time period based on the historical power data and first target meteorological data corresponding to the at least one target meteorological influencing factor to obtain a plurality of first photovoltaic power generation power prediction values ​​for the first preset time period; the first target meteorological data being meteorological data for the first preset time period; Predicting the photovoltaic power generation power in a second preset time period based on second target meteorological data corresponding to the at least one target meteorological influencing factor to obtain a plurality of second photovoltaic power generation power prediction values ​​for the second preset time period; the second preset time period is longer than the first preset time period; and the second target meteorological data is the meteorological data for the second preset time period; The plurality of first photovoltaic power generation power prediction values ​​and the plurality of second photovoltaic power generation power prediction values ​​are merged to obtain a plurality of target photovoltaic power generation power prediction values.

2. The method according to claim 1, characterized in that The step of screening, based on the historical power data and the historical meteorological data, at least one target meteorological influencing factor affecting the ultra-short-term photovoltaic power generation power forecast from a plurality of meteorological influencing factors corresponding to the historical meteorological data comprises: Determining correlation coefficients between a plurality of meteorological impact factors corresponding to the historical meteorological data and the historical power data; generating, based on the historical power data and the historical meteorological data, a meteorological factor time series graph of a plurality of meteorological influencing factors corresponding to the historical meteorological data and a power time series graph corresponding to the historical power data; At least one target meteorological impact factor is selected from the plurality of meteorological impact factors, the correlation coefficient of which is greater than a first preset value, and the similarity between the change trend of the meteorological factor time series graph and the change trend of the power time series graph is greater than a second preset value.

3. The method according to claim 1 or 2, characterized in that The step of predicting the photovoltaic power generation power in a first preset time period based on the historical power data and the first target meteorological data corresponding to the at least one target meteorological influencing factor to obtain a plurality of first photovoltaic power generation power prediction values ​​in the first preset time period includes: The historical power data and the first target meteorological data are input into the trained ultra-short-term photovoltaic power generation power prediction model to obtain multiple first photovoltaic power generation power prediction values ​​for the first preset time period; the input of the ultra-short-term photovoltaic power generation power prediction model is the historical power data and the first target meteorological data, and the output of the ultra-short-term photovoltaic power generation power prediction model is multiple first photovoltaic power generation power prediction values; the ultra-short-term photovoltaic power generation power prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

4. The method according to claim 1 or 2, characterized in that The step of predicting the photovoltaic power generation power in a second preset time period based on the second target meteorological data corresponding to the at least one target meteorological influencing factor to obtain a plurality of second photovoltaic power generation power prediction values ​​in the second preset time period includes: The second target meteorological data is input into the trained short-term photovoltaic power generation power prediction model to obtain multiple second photovoltaic power generation power prediction values ​​for the second preset time period; the input of the short-term photovoltaic power generation power prediction model is the historical power data and the second target meteorological data, and the output of the short-term photovoltaic power generation power prediction model is multiple second photovoltaic power generation power prediction values; the short-term photovoltaic power generation power prediction model is a neural network model constructed based on a bidirectional long short-term memory network and a multi-layer perceptron.

5. The method according to claim 1 or 2, characterized in that The fusing the first photovoltaic power generation power prediction value and the second photovoltaic power generation power prediction value to obtain a target photovoltaic power generation power prediction value includes: Filtering, from the plurality of second photovoltaic power generation prediction values ​​in the second preset time period, a plurality of corresponding photovoltaic power generation prediction values ​​in the first preset time period; Obtaining a first coefficient and a second coefficient; the first coefficient is used to reflect the degree of influence of the first photovoltaic power generation power prediction value on the actual power, and the second coefficient is used to reflect the degree of influence of the second photovoltaic power generation power prediction value on the actual power; Obtain a plurality of first product results by multiplying the first coefficient by each of the first photovoltaic power prediction values, and obtain a plurality of second product results by multiplying the second coefficient by each of the second photovoltaic power prediction values; A plurality of target photovoltaic power generation power prediction values ​​are obtained according to the sum of each first multiplication result and the corresponding second multiplication result.

6. The method according to claim 1 or 2, characterized in that Before predicting the photovoltaic power generation power in a first preset time period based on the historical power data and the first target meteorological data corresponding to the at least one target meteorological influencing factor, the method further includes: Determining whether the time period corresponding to the historical power data and the first preset time period are on the same day; If the time period corresponding to the historical power data is not on the same day as the first preset time period, selecting a target preset number of first target historical sample days from the plurality of historical sample days based on the first irradiance data of the plurality of historical sample days and the second irradiance data of the day to be predicted; The historical power data is updated using the first target historical power data of the first target historical sample days of the target preset number; the historical power data used for photovoltaic power generation power prediction in the first preset time period is the updated historical power data.

7. An ultra-short-term photovoltaic power generation prediction device integrating short-term prediction results, characterized in that: The device comprises: a meteorological factor screening module, configured to obtain historical power data and historical meteorological data of a target photovoltaic power station, and screen, based on the historical power data and the historical meteorological data, at least one target meteorological influencing factor that affects ultra-short-term photovoltaic power generation power forecasting from a plurality of meteorological influencing factors corresponding to the historical meteorological data; an ultra-short-term photovoltaic power generation prediction module, configured to predict the photovoltaic power generation power for a first preset time period based on the historical power data and first target meteorological data corresponding to the at least one target meteorological influencing factor, and obtain a plurality of first photovoltaic power generation power prediction values ​​for the first preset time period; the first target meteorological data being meteorological data for the first preset time period; a short-term photovoltaic power generation prediction module, configured to predict the photovoltaic power generation in a second preset time period based on second target meteorological data corresponding to the at least one target meteorological influencing factor, and obtain a plurality of second photovoltaic power generation prediction values ​​for the second preset time period; the second preset time period is longer than the first preset time period; and the second target meteorological data is the meteorological data for the second preset time period; The prediction fusion module is used to fuse the multiple first photovoltaic power generation power prediction values ​​and the multiple second photovoltaic power generation power prediction values ​​to obtain multiple target photovoltaic power generation power prediction values.

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 ultra-short-term photovoltaic power generation prediction method integrating short-term prediction results 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 ultra-short-term photovoltaic power generation power prediction method integrating short-term prediction results according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the ultra-short-term photovoltaic power generation prediction method by integrating short-term prediction results according to any one of claims 1 to 6.

Images