Flexible power supply system and method based on photovoltaic energy storage

Abstract

This application discloses a flexible power supply system and method based on photovoltaic energy storage, which uses artificial intelligence technology based on deep learning to perform time series analysis on the historical power generation and historical weather data of photovoltaic power stations, and captures the time series dynamic change patterns of historical power generation and historical weather data in the global time domain through local time series feature extraction and feature time series propagation aggregation mechanism, and mines the nonlinear correlation between power generation and weather conditions through multi-scale interactive response analysis of the two, so as to perform decoding prediction of photovoltaic power generation. In this way, the accuracy of prediction can be effectively improved, and complex and changeable weather conditions can be better adapted to meet the needs of modern smart grids for accurate prediction and efficient management.

Description

Technical Field

[0001] The present application relates to the field of intelligent prediction, and more specifically, to a flexible power supply system and method based on photovoltaic energy storage. Background Art

[0002] As the world pays more and more attention to renewable energy, photovoltaic power generation, as a clean and renewable energy form, has become increasingly important. However, the high uncertainty of photovoltaic power generation, especially the significant impact of weather conditions (such as light intensity, cloud cover, temperature, etc.), has brought great challenges to the dispatching and energy storage management of power systems.

[0003] The invention patent with publication number CN116436077A discloses a flexible power supply system and method based on photovoltaic energy storage, which predicts the power generation of photovoltaic power stations and the power consumption of users by obtaining information such as photovoltaic power station information, user information, and transmission loss, so as to realize the energy storage and transmission management of photovoltaic power stations. Specifically, in terms of the prediction of power generation of photovoltaic power stations, the scheme takes the weighted average of the power generation of each photovoltaic power station under similar weather conditions in the same month as the predicted value of power generation under the same weather conditions in the same month. However, due to the complexity and variability of weather conditions, there may be large differences in weather conditions even in the same month. The prior art only considers the monthly factor and uses a simple statistical model to predict power generation, which is too general and vague. In addition, as the service life of photovoltaic power stations increases, the decline of equipment performance further reduces the accuracy of power generation prediction based on the same historical month.

[0004] Therefore, an optimized flexible power supply system and method based on photovoltaic energy storage is expected. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of the present application provides a flexible power supply system and method based on photovoltaic energy storage, which uses artificial intelligence technology based on deep learning to perform time series analysis on the historical power generation and historical weather data of photovoltaic power stations, and through local time series feature extraction and feature time series propagation aggregation mechanism, respectively captures the time series dynamic change patterns of historical power generation and historical weather data in the global time domain, and through multi-scale interactive response analysis of the two, excavates the nonlinear correlation between power generation and weather conditions, so as to perform decoding prediction of photovoltaic power generation. In this way, the accuracy of the prediction can be effectively improved, better adapt to complex and changeable weather conditions, and meet the needs of modern smart grids for accurate prediction and efficient management.

[0006] According to one aspect of the present application, a flexible power supply system based on photovoltaic energy storage is provided, which includes: a photovoltaic power station acquisition module, which is used to acquire photovoltaic power station information in a certain area in real time, and the photovoltaic power station information includes power supply location and power supply amount; a user end acquisition module, which is used to acquire user information in the certain area, and the user information includes power consumption and power consumption location; a photovoltaic power generation prediction module, which is used to predict power generation according to the historical power generation and historical weather data of each photovoltaic power station; a power consumption prediction module, which is used to predict power consumption according to the user's historical monthly power consumption and historical working day and non-working day power consumption; a pre-reserve determination module, which is used to determine the pre-reserve of each photovoltaic power station according to the prediction results of the power generation of each photovoltaic power station and the prediction results of the power consumption, characterized in that the photovoltaic power generation prediction module includes:

[0007] A historical data acquisition unit, used to acquire historical power generation and historical weather data of the photovoltaic power station;

[0008] A historical data time series encoding unit, used for performing local time series feature extraction and feature propagation aggregation on the historical power generation and historical weather data of the photovoltaic power station to obtain a time series significant aggregation propagation representation vector of power generation and a time series significant aggregation propagation representation vector of weather data;

[0009] An interactive response analysis unit, used for performing interactive response analysis on the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector to obtain a power generation-weather local-global interactive response matrix;

[0010] The power generation prediction unit is used to generate a predicted decoded value of the power generation of the photovoltaic power station based on the power generation-weather local-global interactive response matrix.

[0011] According to another aspect of the present application, a flexible power supply method based on photovoltaic energy storage is provided, which includes:

[0012] Real-time acquisition of photovoltaic power station information within a certain area, wherein the photovoltaic power station information includes power supply location and power supply amount;

[0013] Acquire user information within the certain area, wherein the user information includes power consumption and power consumption location;

[0014] Predict power generation based on historical power generation and historical weather data for each photovoltaic power station;

[0015] Predict electricity consumption based on the user's historical monthly electricity consumption and historical working day and non-working day electricity consumption;

[0016] The pre-storage capacity of each photovoltaic power station is determined according to the prediction result of the power generation of each photovoltaic power station and the prediction result of the power consumption.

[0017] Compared with the prior art, the flexible power supply system and method based on photovoltaic energy storage provided by the present application uses artificial intelligence technology based on deep learning to perform time series analysis on the historical power generation and historical weather data of photovoltaic power stations, and captures the time series dynamic change patterns of historical power generation and historical weather data in the global time domain through local time series feature extraction and feature time series propagation aggregation mechanism, and mines the nonlinear correlation between power generation and weather conditions through multi-scale interactive response analysis of the two, so as to perform decoding prediction of photovoltaic power generation. In this way, the accuracy of prediction can be effectively improved, and it can better adapt to complex and changeable weather conditions, and meet the needs of modern smart grids for accurate prediction and efficient management. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 is a block diagram of a flexible power supply system based on photovoltaic energy storage according to an embodiment of the present application;

[0020] Figure 2 A schematic diagram of data flow of a flexible power supply system based on photovoltaic energy storage according to an embodiment of the present application;

[0021] Figure 3 A block diagram of a photovoltaic power generation prediction module in a flexible power supply system based on photovoltaic energy storage according to an embodiment of the present application;

[0022] Figure 4 It is a flow chart of a flexible power supply method based on photovoltaic energy storage according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0025] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0026] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0027] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0028] The existing technology only considers the monthly factor and uses a simple statistical model to predict power generation, which is too general and vague. In addition, as the service life of photovoltaic power stations increases, the performance of the equipment declines, which further reduces the accuracy of power generation prediction based on the same month in history. Therefore, an optimized flexible power supply system and method based on photovoltaic energy storage is expected.

[0029] In the technical solution of the present application, a flexible power supply system based on photovoltaic energy storage is proposed. Figure 1 It is a block diagram of a flexible power supply system based on photovoltaic energy storage according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of a flexible power supply system based on photovoltaic energy storage according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the flexible power supply system 300 based on photovoltaic energy storage includes: a photovoltaic power station acquisition module 310, which is used to acquire the information of photovoltaic power stations in a certain area in real time, and the photovoltaic power station information includes the power supply location and power supply amount; a user-end acquisition module 320, which is used to acquire the user information in the certain area, and the user information includes the power consumption and the power consumption location; a photovoltaic power generation prediction module 330, which is used to predict the power generation according to the historical power generation and historical weather data of each photovoltaic power station; a power consumption prediction module 340, which is used to predict the power consumption according to the user's historical monthly power consumption and historical working day and non-working day power consumption; a pre-reserve determination module 350, which is used to determine the pre-reserve of each photovoltaic power station according to the prediction results of the power generation of each photovoltaic power station and the prediction results of the power consumption.

[0030] In particular, the photovoltaic power station acquisition module 310 is used to obtain photovoltaic power station information in a certain area in real time, and the photovoltaic power station information includes power supply location and power supply amount. It should be understood that due to the complexity and variability of weather conditions, weather conditions may vary greatly even in the same month. Real-time data can improve the accuracy of power generation prediction, especially considering the impact of weather changes on photovoltaic power generation.

[0031] In particular, the user-side acquisition module 320 is used to obtain user information in the certain area, and the user information includes power consumption and power consumption location. It should be understood that the power consumption patterns of users in different regions may be different, and understanding the power consumption location can help identify regional demand differences. By analyzing power consumption, power distribution can be optimized to avoid grid overload or power shortage. By real-time monitoring and analysis of these data, power management can be optimized and the reliability and efficiency of power supply can be improved.

[0032] In particular, the photovoltaic power generation prediction module 330 is used to predict the power generation according to the historical power generation and historical weather data of each photovoltaic power station. Figure 3As shown, the photovoltaic power generation prediction module 330 includes: a historical data acquisition unit 331, which is used to acquire the historical power generation and historical weather data of the photovoltaic power station; a historical data time series encoding unit 332, which is used to perform local time series feature extraction and feature propagation aggregation on the historical power generation and historical weather data of the photovoltaic power station to obtain a power generation time series significant aggregation propagation representation vector and a weather data time series significant aggregation propagation representation vector; an interactive response analysis unit 333, which is used to perform interactive response analysis on the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector to obtain a power generation-weather local-global interactive response matrix; a power generation prediction unit 334, which is used to generate a predicted decoded value of the power generation of the photovoltaic power station based on the power generation-weather local-global interactive response matrix.

[0033] Specifically, the historical data acquisition unit 331 is used to acquire the historical power generation and historical weather data of the photovoltaic power station. In the technical solution of the present application, in order to ensure the richness and diversity of data, the time range of the historical power generation and historical weather data of the photovoltaic power station covers at least one year.

[0034] Specifically, the historical data time series encoding unit 332 is used to perform local time series feature extraction and feature propagation aggregation on the historical power generation and historical weather data of the photovoltaic power station to obtain a power generation time series significant aggregation propagation representation vector and a weather data time series significant aggregation propagation representation vector. Specifically, first, the historical power generation of the photovoltaic power station is input into the power generation sequence encoder based on the Bi-LSTM model to obtain a sequence of implicitly associated feature vectors of the local time series of the power generation; considering that the time span of the historical power generation and the historical weather data is relatively long, in order to capture the time series variation law of the historical power generation and historical weather data in more detail, the present application adopts the local time series feature extraction technology to perform time series analysis on the two. Specifically, for the historical power generation, the present application adopts a power generation sequence encoder based on the Bi-LSTM model to perform sequence encoding on the power generation data in each local time domain respectively, and uses the bidirectional long short-term memory architecture of the Bi-LSTM model to simultaneously capture the forward and backward temporal dependencies of the historical power generation data, fully explore the temporal context information of the power generation in each local time domain, and extract its local temporal correlation feature representation to obtain a sequence of implicit correlation feature vectors of the local temporal sequence of power generation. It is worth mentioning that the bidirectional long short-term memory network (Bi-LSTM) is an improved recurrent neural network (RNN) that can capture contextual information from sequence data. Bi-LSTM improves the ability to understand time series data by processing data simultaneously in two directions of the sequence. Next, each weather data in the historical weather data of the photovoltaic power station is one-hot encoded to obtain a time series of one-hot encoded vectors of weather data; since the historical weather data has various types, such as cloudy, sunny, overcast, rainy, strong sunshine, fog, haze, etc., in order to ensure that the model can accurately identify and process different types of weather conditions, the present application further performs one-hot encoding on each weather data in the historical weather data of the photovoltaic power station, so that each weather category corresponds to a unique binary vector to maintain the clarity of the data structure, thereby obtaining a time series of one-hot encoded vectors of weather data. It is worth mentioning that one-hot encoding is an encoding method that converts categorical variables into numerical form, which is commonly used in machine learning and data processing. Its main purpose is to represent each category as a binary vector so that the model can effectively process these category information.Then, the time series of the weather data unique hot encoding vector is input into the weather sequence encoder based on the 1D-CNN model to obtain the sequence of implicitly associated feature vectors of the local time series of weather data; that is, the time series of the weather data unique hot encoding vector is divided into multiple local time domains using the data processing means for extracting local time series features, and the 1D-CNN model is used as the weather sequence encoder to perform time series encoding on the weather data unique hot encoding vector in each local time domain, and a sliding convolution operation is performed on the time series using a one-dimensional convolution kernel to capture the characteristic change law of the weather data in the local time domain, thereby obtaining a sequence of implicitly associated feature vectors of the local time series of weather data. In an embodiment of the present application, the historical weather data and the historical power generation are divided into local time domains with the month as the time unit. One-dimensional convolutional neural network (1D-CNN) is a deep learning model specifically used for processing one-dimensional sequence data. It extracts local features through convolution operations and is suitable for tasks such as time series analysis and text data processing. Subsequently, the sequence of implicitly associated feature vectors of the local time series of power generation and the sequence of implicitly associated feature vectors of the local time series of weather data are respectively subjected to feature time series propagation aggregation to obtain the significant aggregation propagation representation vector of the power generation time series and the significant aggregation propagation representation vector of the weather data time series; in order to obtain the time series dynamic change pattern of historical weather data and historical power generation in the global time domain, the sequence of implicitly associated feature vectors of the local time series of power generation and the sequence of implicitly associated feature vectors of the local time series of weather data are further subjected to information time series aggregation processing. It is worth mentioning that, considering that in the process of photovoltaic power generation prediction, data information in different local time domains has different importance for the prediction of power generation, for example, local time domain information with a shorter time interval and a more similar time background to the current time point often has a higher reference value. Therefore, this application adopts an attention mechanism to aggregate local time domain information based on the spatiotemporal transmission characteristics of information.

[0035] In an embodiment of the present application, the process of performing feature time series propagation aggregation on the sequence of the local time series implicit association feature vectors of the power generation and the sequence of the local time series implicit association feature vectors of the weather data to obtain the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector includes: first extracting the current local time series implicit association feature vector from the sequence of the local time series implicit association feature vectors of the power generation, and defining other local time series implicit association feature vectors of the power generation in the sequence of the local time series implicit association feature vectors of the power generation as historical local time series implicit association feature vectors of the power generation to obtain a sequence of historical local time series implicit association feature vectors of the power generation; and calculating the static energy value of each historical local time series implicit association feature vector in the sequence of the historical local time series implicit association feature vectors of the power generation; and calculating the static energy value of each historical local time domain feature vector based on its statistical characteristics, and using this as the inherent attribute of the feature vector in the feature space to represent its static importance. Next, based on the spatial span and time span between the local time series implicit correlation feature vectors of each historical power generation and the local time series implicit correlation feature vectors of the current power generation, the static energy values ​​of the local time series implicit correlation feature vectors of each historical power generation are modulated to obtain the dynamic energy spatiotemporal transfer value of the local time series implicit correlation feature vectors of each historical power generation relative to the local time series implicit correlation feature vectors of the current power generation; that is, based on the time span between each historical local time domain feature vector and the current local time domain feature vector and the spatial span of both in the feature space, the static energy values ​​are spatiotemporally modulated to dynamically adjust the energy intensity of information transfer and obtain the dynamic energy spatiotemporal transfer value of each historical local time domain feature vector, so that the local time domain information that is closer to the current time point and has a more similar time background can obtain a higher weight in the information aggregation process, thereby enhancing the sensitivity to the locality and timeliness of information transfer. Then, each dynamic energy spatiotemporal transfer value is further normalized and masked to screen and strengthen important spatiotemporal transfer information. Finally, based on the dynamic energy spatiotemporal transfer value of each historical power generation local time series implicit correlation feature vector, the sequence of the power generation local time series implicit correlation feature vector is aggregated to obtain the power generation time series significant aggregation propagation representation vector. The local time domain information is aggregated by weighted summation to obtain the power generation time series significant aggregation propagation representation vector with global time series context information and the weather data time series significant aggregation propagation representation vector.

[0036] Among them, the process of calculating the static energy value of each local time series implicit correlation feature vector of historical power generation in the sequence of local time series implicit correlation feature vector of historical power generation includes: calculating the expected value of the fourth power of the difference between each eigenvalue in the local time series implicit correlation feature vector of historical power generation and its eigenmean, and dividing the expected value by the square of the eigenvariance of the local time series implicit correlation feature vector of historical power generation to obtain the kurtosis of the local time series implicit correlation feature vector of historical power generation; calculating the kurtosis of the local time series implicit correlation feature vector of historical power generation and the sum of its maximum eigenvalue to obtain the static energy value.

[0037] More specifically, based on the spatial span and time span between the local time series implicit association feature vectors of each historical power generation amount and the local time series implicit association feature vector of the current power generation amount, the static energy value of the local time series implicit association feature vector of each historical power generation amount is modulated to obtain the dynamic energy spatiotemporal transfer value of the local time series implicit association feature vector of each historical power generation amount relative to the local time series implicit association feature vector of the current power generation amount, including: taking the square value of the number of feature vectors between the local time series implicit association feature vector of the historical power generation amount and the local time series implicit association feature vector of the current power generation amount as the spatial span coefficient, calculating the local time series implicit association feature vector of the historical power generation amount The static energy value of the associated characteristic vector is divided by the spatial span coefficient to obtain the energy spatial attenuation factor; the timestamp difference between the local time series implicit associated characteristic vector of the historical power generation and the local time series implicit associated characteristic vector of the current power generation is calculated, the timestamp difference is rounded down and divided by the preset time attenuation period to obtain the time span coefficient, and the static energy value of the local time series implicit associated characteristic vector of the historical power generation is divided by the natural exponential function value of the time span coefficient to obtain the energy time attenuation factor; the weighted sum of the energy spatial attenuation factor and the energy time attenuation factor of the local time series implicit associated characteristic vector of the historical power generation is calculated to obtain the dynamic energy spatiotemporal transfer value.

[0038] More specifically, based on the dynamic energy spatiotemporal transfer values ​​of the implicitly associated characteristic vectors of each historical local time series of power generation, a process of aggregating the sequence of the implicitly associated characteristic vectors of the local time series of power generation to obtain the significantly aggregated propagation representation vector of the power generation time series includes: inputting the dynamic energy spatiotemporal transfer values ​​of the implicitly associated characteristic vectors of the local time series of historical power generation into an energy transfer gating module based on a mask function to obtain a sequence of dynamic energy spatiotemporal transfer weights; based on the sequence of dynamic energy spatiotemporal transfer weights, calculating the weighted sum of the sequence of the implicitly associated characteristic vectors of the local time series of historical power generation to obtain the aggregated modulation vector of the local time series of historical power generation characteristics; calculating the positional sum of the aggregated modulation vector of the local time series of historical power generation characteristics and the implicitly associated characteristic vector of the local time series of current power generation to obtain the significantly aggregated propagation representation vector of the power generation time series.

[0039] In summary, in the above embodiment, the sequence of implicitly associated feature vectors of the local time series of power generation is subjected to feature time series propagation aggregation to obtain the significant aggregate propagation representation vector of the time series of power generation, including: the sequence of implicitly associated feature vectors of the local time series of power generation is subjected to feature time series propagation aggregation using the following feature time series propagation aggregation formula to obtain the significant aggregate propagation representation vector of the time series of power generation; wherein the formula is:

[0040] X={x1,x2,...,x i , ..., x p}

[0041]

[0042] Where X represents the sequence of implicit correlation feature vectors of the local time series of power generation, x1, x2 and x i They represent the first, second and i-th local time series implicit correlation feature vectors of historical power generation, respectively, and x p represents the implicit correlation feature vector of the local time series of the current power generation, the value of p is the number of feature vectors in the sequence of the implicit correlation feature vector of the local time series of the power generation, and x i (j) represents the eigenvalue of the jth position in the implicit correlation eigenvector of the local time series of the i-th historical power generation, max(x i ), μ i and σ i 4 denote the maximum eigenvalue, eigenmean and square of eigenvariance of the implicit correlation eigenvector of the local time series of the i-th historical power generation, E{·} denotes the expected value of the calculation set, represents the static energy value of the implicit correlation feature vector of the local time series of the i-th historical power generation, Count(x i →x p) represents the number of feature vectors between the i-th historical power generation local time series implicit correlation feature vector and the current power generation local time series implicit correlation feature vector, t p and t i respectively represent the timestamps of the implicit correlation feature vector of the local time series of the current power generation and the implicit correlation feature vector of the local time series of the i-th historical power generation, represents rounding down, L represents the preset time decay period, exp(·) represents the exponential function operation, and e (i→p) They respectively represent the energy space attenuation factor, energy time attenuation factor and dynamic energy space-time transfer value of the implicit correlation feature vector of the local time series of the i-th historical power generation relative to the implicit correlation feature vector of the local time series of the current power generation, a and b are hyperparameters with different weights, mask(·) represents mask processing, sigmoid represents the sigmoid activation function, θ represents the preset mask threshold, and f1 represents the significant aggregation propagation representation vector of the power generation time series.

[0043] Specifically, the interactive response analysis unit 333 is used to perform interactive response analysis on the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector to obtain the power generation-weather local-global interactive response matrix. That is, by performing interactive response analysis on the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector, the correlation and influence relationship between weather conditions and power generation is mined to predict future power generation. In order to achieve this goal, the present application adopts an interactive response analysis mechanism, which first decomposes the weather data time series significant aggregation propagation representation vector into a sequence of weather data local time series feature vectors through a decoupling operation to reveal different dimensions or aspects within it and enhance the model's ability to capture local features. At the same time, the power generation time series significant aggregation propagation representation vector is nonlinearly transformed to adjust its feature dimensions for subsequent feature interaction operations. Next, the correlation matrix between the significant aggregation propagation representation vector of the power generation time series after nonlinear transformation and the local time series feature vectors of each weather data after decoupling is calculated to capture the interaction between feature vectors, reveal the correlation and information flow between features, and thus form a sequence of power generation-weather interaction correlation matrices. Then, the characteristic interaction energy factors of each power generation-weather interaction correlation matrix are further calculated to reveal the intensity and importance of the interaction between feature vectors. The characteristic interaction energy factor sequence is then normalized by the Sigmoid function and converted into attention weights, so as to perform weighted summation on the sequence of power generation-weather interaction correlation matrices, thereby strengthening the most critical feature interactions and performing global integration of local interaction information to obtain the power generation-weather local-global interaction response matrix. In this way, a fine-grained correlation response representation between power generation and weather conditions is provided, providing a more comprehensive and accurate basis for the accurate prediction of future power generation.

[0044] In an embodiment of the present application, an interactive response analysis is performed on the time series significant aggregate propagation representation vector of power generation and the time series significant aggregate propagation representation vector of weather data to obtain a power generation-weather local-global interactive response matrix, including: feature decoupling of the time series significant aggregate propagation representation vector of weather data to obtain a sequence of local time series feature vectors of weather data; calculating the feature interaction energy factor between each local time series feature vector of weather data in the sequence of the time series significant aggregate propagation representation vector of power generation and the local time series feature vector of weather data to obtain a sequence of power generation-weather feature interaction energy factors; based on the sequence of power generation-weather feature interaction energy factors, feature interaction global fusion is performed on the sequence of the time series significant aggregate propagation representation vector of power generation and the local time series feature vector of weather data to obtain the local-global interactive response matrix of power generation-weather.

[0045] Among them, the process of calculating the characteristic interaction energy factor between the power generation time series significant aggregation propagation representation vector and each local time series feature vector of weather data in the sequence of local time series feature vectors of weather data to obtain a sequence of power generation-weather characteristic interaction energy factors includes: performing a nonlinear transformation on the power generation time series significant aggregation propagation representation vector to obtain the power generation time series significant aggregation propagation representation vector after the nonlinear transformation; calculating the interaction correlation matrix between the power generation time series significant aggregation propagation representation vector after the nonlinear transformation and each local time series feature vector of weather data in the sequence of local time series feature vectors of weather data to obtain a sequence of power generation-weather interaction correlation matrices; calculating the characteristic interaction energy factor of each power generation-weather interaction correlation matrix in the sequence of power generation-weather interaction correlation matrices to obtain a sequence of power generation-weather characteristic interaction energy factors. Among them, calculating the interaction matrix between the significant aggregation propagation representation vector of the time series of power generation after the nonlinear transformation and each local time series feature vector of weather data in the sequence of local time series feature vectors of weather data to obtain the sequence of power generation-weather interaction correlation matrix includes: calculating the vector product between the transposed vector of the significant aggregation propagation representation vector of the time series of power generation after the nonlinear transformation and each local time series feature vector of weather data in the sequence of local time series feature vectors of weather data to obtain the sequence of power generation-weather interaction correlation matrix. And, calculating the characteristic interaction energy factor of each power generation-weather interaction correlation matrix in the sequence of power generation-weather interaction correlation matrix to obtain the sequence of power generation-weather characteristic interaction energy factor includes: calculating the square of the difference between the kurtosis of the power generation-weather interaction correlation matrix and its characteristic mean, the characteristic variance of the power generation-weather interaction correlation matrix and the weighted sum value between the bias terms, and then adding the characteristic variance of the power generation-weather interaction correlation matrix and the bias term and dividing by the weighted sum value to obtain the power generation-weather characteristic interaction energy factor.

[0046] More specifically, based on the sequence of power generation-weather characteristic interaction energy factors, the process of performing feature interaction global fusion on the sequence of power generation time series significant aggregation propagation representation vectors and the sequence of local time series feature vectors of weather data to obtain the power generation-weather local-global interaction response matrix includes: inputting the sequence of power generation-weather characteristic interaction energy factors into the Sigmoid function to obtain a sequence of normalized power generation-weather characteristic interaction energy factors; using the sequence of normalized power generation-weather characteristic interaction energy factors as a sequence of weights, calculating the position-weighted sum of the sequence of the power generation-weather interaction correlation matrix to obtain the power generation-weather local-global interaction response matrix.

[0047] In summary, in the above embodiment, the interactive response analysis of the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector is performed to obtain the power generation-weather local-global interactive response matrix, including: processing the weather data time series significant aggregation propagation representation vector and the power generation time series significant aggregation propagation representation vector with the following associated interaction formula to obtain the power generation-weather local-global interactive response matrix, wherein the associated interaction formula is:

[0048] f 1t =Wf1+b

[0049] Decouple(f2)={f 21 ,f 22 ,...,f 2i ,...,f 2n}

[0050]

[0051] E norm,12i =sigmoid(E 12i )

[0052] M=∑ i M 12i E norm,12i

[0053] Wherein, f1 is the significant aggregate propagation representation vector of the power generation time series, f2 is the significant aggregate propagation representation vector of the weather data time series, W is the weight matrix, b is the bias vector, and f 1t is the vector representing the significant aggregation propagation of power generation time series after nonlinear transformation, Decouple(·) represents the decoupling process, and f 21 、f 22 、f 2i 、f 2n Respectively represent the first, second, i-th and n-th local time series feature vectors of weather data in the sequence of local time series feature vectors of weather data, (·) T represents the transpose of a vector, Represents matrix multiplication operation, M 12i represents the i-th power generation-weather interaction matrix, that is, the interaction matrix between the power generation time series significant aggregation propagation representation vector and the i-th weather data local time series feature vector, E 12i represents the characteristic interaction energy factor of the i-th power generation-weather interaction matrix, k, μ and σ 2denote the kurtosis, characteristic mean and characteristic variance of the i-th power generation-weather interaction matrix, λ is the bias term, E{·} denotes the expected value of the calculation set, sigmoid denotes the sigmoid activation function, E norm,12i represents the i-th normalized power generation-weather characteristic interaction energy factor, and M represents the power generation-weather local-global interaction response matrix.

[0054] Specifically, the power generation prediction unit 334 is used to generate a predicted decoded value of the power generation of the photovoltaic power station based on the power generation-weather local-global interactive response matrix. In a specific example of the present application, the power generation-weather local-global interactive response matrix is ​​input into a decoder-based power generation prediction module to obtain a predicted decoded value of the power generation of the photovoltaic power station.

[0055] In a preferred example, inputting the power generation-weather local-global interactive response matrix into a decoder-based power generation prediction module to obtain a predicted decoded value of the power generation of the photovoltaic power station includes:

[0056] Calculating the sum of the absolute values ​​of each eigenvalue of the power generation-weather local-global interaction response matrix to obtain a first power generation-weather local-global interaction response and a modulation value, and calculating the square root of the sum of the squares of each eigenvalue of the power generation-weather local-global interaction response matrix to obtain a second power generation-weather local-global interaction response and a modulation value;

[0057] After point-wise subtraction of the power generation-weather local-global interaction response matrix and the second power generation-weather local-global interaction response and modulation value, point-wise multiplication is performed with the number of eigenvalues ​​of the power generation-weather local-global interaction response matrix and the inverse of the first power generation-weather local-global interaction response and modulation value, and the inverse of each eigenvalue is taken to obtain a first power generation-weather local-global interaction response phase conversion matrix;

[0058] After point-wise subtraction of the power generation-weather local-global interaction response matrix from the first power generation-weather local-global interaction response and modulation value, point-wise multiplication is performed with the square root of the number of eigenvalues ​​of the power generation-weather local-global interaction response matrix and the inverse of the second power generation-weather local-global interaction response and modulation value, and the inverse of each eigenvalue is taken to obtain a second power generation-weather local-global interaction response phase conversion matrix; and

[0059] Performing a dot multiplication of the second power generation-weather local-global interactive response phase conversion matrix with a weighted hyperparameter, and subtracting the dot multiplication result from the first power generation-weather local-global interactive response phase conversion matrix to obtain an optimized power generation-weather local-global interactive response matrix; and

[0060] The optimized power generation-weather local-global interactive response matrix is ​​input into a decoder-based power generation prediction module to obtain a predicted decoded value of the power generation of the photovoltaic power station.

[0061] Here, the power generation-weather local-global interaction response matrix, denoted as M, is optimally expressed as:

[0062]

[0063]

[0064] n=W×H

[0065] Where M is the power generation-weather local-global interaction response matrix, m i is the i-th eigenvalue of the power generation-weather local-global interaction response matrix, α is the first power generation-weather local-global interaction response and modulation value, and β is the second power generation-weather local-global interaction response and modulation value, n is the number of eigenvalues ​​of the power generation-weather local-global interaction response matrix, M1 is the first power generation-weather local-global interaction response phase transition matrix, M2 is the second power generation-weather local-global interaction response phase transition matrix, ω is a weighted hyperparameter, is the point subtraction, ⊙ is the point multiplication, and M' is the optimized power generation-weather local-global interaction response matrix.

[0066] That is, considering that the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector represent the power generation time series propagation aggregation characteristics and the weather data time series propagation aggregation characteristics respectively, when they are passed through the feature local-global interactive response module based on feature decoupling, the time series correlation feature distribution pattern based on the source time series distribution will also have local-global time series-decoding prediction interactive response differences. Therefore, it is expected to improve the detailed interactive response semantic aggregation expression effect of the power generation-weather local-global interactive response matrix based on the correlation pattern representation difference.

[0067] Therefore, in the preferred example, the difference of the difference and modulation representation of the eigenvalue of the power generation-weather local-global interaction response matrix relative to the overall feature set of the power generation-weather local-global interaction response matrix is ​​used as the semantic change intensity information, and a phase-like transformation corresponding to the position-based intensity modulation is performed through different and modulation representations, so that the aggregation enhancement of semantic change phase perception can enhance the axial aggregation receptive field along the feature aggregation direction by performing a spatial translation operation based on alternating stacking under the set scale equilibrium of the power generation-weather local-global interaction response matrix, thereby improving the perception effect of the aggregated semantics of the power generation-weather local-global interaction response matrix on the detailed semantic changes, so as to improve the expression effect of the power generation-weather local-global interaction response matrix, and improve the accuracy of the predicted decoding value of the power generation of the photovoltaic power station obtained by inputting it into the decoder-based power generation prediction module.

[0068] It is worth mentioning that in other examples of the present application, the power generation can be predicted based on the historical power generation and historical weather data of each photovoltaic power station through the following steps, for example: first, the historical power generation and historical weather data of each photovoltaic power station are input; statistical methods and visualization tools (such as scatter plots, time series graphs) are used to analyze the relationship between power generation and various weather factors; the correlation between power generation and weather variables is calculated to identify the most influential features; a suitable model is selected according to the characteristics of the data and the prediction target, such as a deep learning model (such as LSTM, GRU), and the trained model is deployed to the production environment for real-time or periodic prediction; the prediction performance of the model is continuously monitored, and the model is regularly updated to adapt to new data and changing environments.

[0069] In particular, the power consumption prediction module 340 and the pre-reserve determination module 350 are used to predict power consumption according to the user's historical monthly power consumption and historical working day and non-working day power consumption; and determine the pre-reserve of each photovoltaic power station according to the prediction results of the power generation of each photovoltaic power station and the prediction results of the power consumption. That is, by performing multi-scale interactive response analysis on the user's historical monthly power consumption and historical working day and non-working day power consumption, the nonlinear correlation between the power generation and weather conditions is excavated, so as to perform decoding prediction of photovoltaic power generation. In this way, the accuracy of the prediction can be effectively improved, and the complex and changeable weather conditions can be better adapted to meet the needs of modern smart grids for accurate prediction and efficient management. Furthermore, analyzing the pre-reserve of photovoltaic power stations can provide important information and support for energy management, which is helpful to optimize power supply.

[0070] As described above, the flexible power supply system 300 based on photovoltaic energy storage according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a flexible power supply algorithm based on photovoltaic energy storage. In one possible implementation, the flexible power supply system 300 based on photovoltaic energy storage according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the flexible power supply system 300 based on photovoltaic energy storage can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the flexible power supply system 300 based on photovoltaic energy storage can also be one of the many hardware modules of the wireless terminal.

[0071] Alternatively, in another example, the flexible power supply system 300 based on photovoltaic energy storage and the wireless terminal may also be separate devices, and the flexible power supply system 300 based on photovoltaic energy storage may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0072] Furthermore, a flexible power supply method based on photovoltaic energy storage is also provided.

[0073] Figure 4 FIG. 1 is a flow chart of a flexible power supply method based on photovoltaic energy storage according to an embodiment of the present application. Figure 4 As shown, according to the flexible power supply method based on photovoltaic energy storage according to the embodiment of the present application, the steps include: S1, real-time acquisition of photovoltaic power station information within a certain area, the photovoltaic power station information including power supply location and power supply amount; S2, acquisition of user information within the certain area, the user information including power consumption and power consumption location; S3, prediction of power generation according to the historical power generation of each photovoltaic power station and historical weather data; S4, prediction of power consumption according to the user's historical monthly power consumption and historical working day and non-working day power consumption; S5, determination of the pre-storage amount of each photovoltaic power station according to the prediction results of the power generation of each photovoltaic power station and the prediction results of the power consumption.

[0074] In summary, the flexible power supply method based on photovoltaic energy storage according to the embodiment of the present application is explained, which uses artificial intelligence technology based on deep learning to perform time series analysis on the historical power generation and historical weather data of the photovoltaic power station, and captures the time series dynamic change patterns of historical power generation and historical weather data in the global time domain through local time series feature extraction and feature time series propagation aggregation mechanism, and through multi-scale interactive response analysis of the two, the nonlinear correlation between power generation and weather conditions is excavated, so as to perform decoding prediction of photovoltaic power generation. In this way, the accuracy of the prediction can be effectively improved, and the complex and changeable weather conditions can be better adapted to meet the needs of modern smart grids for accurate prediction and efficient management.

[0075] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A flexible power supply system based on photovoltaic energy storage, comprising: A photovoltaic power station acquisition module is used to acquire information about photovoltaic power stations in a certain area in real time, and the photovoltaic power station information includes the power supply location and power supply amount; a user terminal acquisition module is used to acquire user information in the certain area, and the user information includes power consumption and power consumption location; a photovoltaic power generation prediction module is used to predict power generation based on the historical power generation and historical weather data of each photovoltaic power station; a power consumption prediction module is used to predict power consumption based on the user's historical monthly power consumption and historical working day and non-working day power consumption; a pre-reserve determination module is used to determine the pre-reserve of each photovoltaic power station based on the prediction results of the power generation of each photovoltaic power station and the prediction results of the power consumption, characterized in that the photovoltaic power generation prediction module includes: A historical data acquisition unit, used to acquire historical power generation and historical weather data of the photovoltaic power station; A historical data time series encoding unit, used for performing local time series feature extraction and feature propagation aggregation on the historical power generation and historical weather data of the photovoltaic power station to obtain a time series significant aggregation propagation representation vector of power generation and a time series significant aggregation propagation representation vector of weather data; An interactive response analysis unit, used for performing interactive response analysis on the power generation time series significant aggregation propagation representation vector and the weather data time series significant aggregation propagation representation vector to obtain a power generation-weather local-global interactive response matrix; A power generation prediction unit, used to generate a predicted decoded value of power generation of the photovoltaic power station based on the power generation-weather local-global interactive response matrix; Wherein, the interactive response analysis unit includes: A feature decoupling subunit, used for performing feature decoupling on the weather data time series significant aggregation propagation representation vector to obtain a sequence of weather data local time series feature vectors; A characteristic interaction energy calculation subunit, used for calculating the characteristic interaction energy factors between the power generation time series significant aggregation propagation representation vector and each weather data local time series feature vector in the sequence of the weather data local time series feature vector to obtain a sequence of power generation-weather characteristic interaction energy factors; The feature interaction global fusion subunit is used to perform feature interaction global fusion on the sequence of the power generation time series significant aggregation propagation representation vector and the sequence of the local time series feature vector of the weather data based on the sequence of the power generation-weather feature interaction energy factors to obtain the power generation-weather local-global interaction response matrix.

2. The flexible power supply system based on photovoltaic energy storage according to claim 1 is characterized in that: The historical data time series encoding unit comprises: A historical power generation local time series encoding subunit, used for inputting the historical power generation of the photovoltaic power station into a power generation sequence encoder based on a Bi-LSTM model to obtain a sequence of implicitly associated feature vectors of the power generation local time series; A one-hot encoding subunit, used for performing one-hot encoding on each weather data in the historical weather data of the photovoltaic power station to obtain a time series of one-hot encoding vectors of the weather data; A weather data local time series encoding subunit, used for inputting the time series of the weather data one-hot encoding vector into a weather sequence encoder based on a 1D-CNN model to obtain a sequence of weather data local time series implicit correlation feature vectors; The characteristic time series propagation aggregation subunit is used to perform characteristic time series propagation aggregation on the sequence of implicitly associated characteristic vectors of the local time series of power generation and the sequence of implicitly associated characteristic vectors of the local time series of weather data, respectively, to obtain the significant aggregation propagation representation vector of the power generation time series and the significant aggregation propagation representation vector of the weather data time series.

3. The flexible power supply system based on photovoltaic energy storage according to claim 2 is characterized in that: The characteristic timing propagation aggregation subunit includes: The time node defines a secondary subunit, which is used to extract the current local time series implicit correlation feature vector from the sequence of the local time series implicit correlation feature vectors of the power generation, and define other local time series implicit correlation feature vectors of the power generation in the sequence of the local time series implicit correlation feature vectors of the power generation as the local time series implicit correlation feature vectors of the historical power generation to obtain the sequence of the local time series implicit correlation feature vectors of the historical power generation; A static energy calculation secondary subunit is used to calculate the static energy value of each historical power generation local time series implicit correlation feature vector in the sequence of the historical power generation local time series implicit correlation feature vector; The dynamic energy spatiotemporal transfer value calculation secondary subunit is used to modulate the static energy value of each local time series implicit correlation feature vector of the historical power generation based on the space span and time span between each local time series implicit correlation feature vector of the historical power generation and the local time series implicit correlation feature vector of the current power generation to obtain the dynamic energy spatiotemporal transfer value of each local time series implicit correlation feature vector of the historical power generation relative to the local time series implicit correlation feature vector of the current power generation; The feature aggregation secondary subunit is used to aggregate the sequence of implicitly associated feature vectors of the local time series of power generation based on the dynamic energy spatiotemporal transfer value of the implicitly associated feature vectors of each historical power generation local time series to obtain the significant aggregation propagation representation vector of the power generation time series.

4. The flexible power supply system based on photovoltaic energy storage according to claim 3 is characterized in that: The static energy calculation secondary subunit is used for: Calculate the expected value of the fourth power of the difference between each eigenvalue and its eigenmean in the local time series implicit correlation eigenvector of the historical power generation, and divide the expected value by the square of the eigenvariance of the local time series implicit correlation eigenvector of the historical power generation to obtain the kurtosis of the local time series implicit correlation eigenvector of the historical power generation; The static energy value is obtained by calculating the sum of the kurtosis of the implicit correlation eigenvector of the local time series of the historical power generation and its maximum eigenvalue.

5. The flexible power supply system based on photovoltaic energy storage according to claim 4 is characterized in that: The dynamic energy spatiotemporal transfer value calculation secondary subunit is used to: Taking the square value of the number of eigenvectors between the local time series implicit correlation eigenvector of the historical power generation and the local time series implicit correlation eigenvector of the current power generation as the spatial span coefficient, the static energy value of the local time series implicit correlation eigenvector of the historical power generation is calculated and divided by the spatial span coefficient to obtain the energy spatial attenuation factor; Calculate the timestamp difference between the local time series implicit correlation feature vector of the historical power generation and the local time series implicit correlation feature vector of the current power generation, round down the timestamp difference and divide it by a preset time decay period to obtain a time span coefficient, and divide the static energy value of the local time series implicit correlation feature vector of the historical power generation by the natural exponential function value of the time span coefficient to obtain an energy time decay factor; The weighted sum of the energy space decay factor and the energy time decay factor of the implicit correlation characteristic vector of the local time series of the historical power generation is calculated to obtain the dynamic energy space-time transfer value.

6. The flexible power supply system based on photovoltaic energy storage according to claim 5 is characterized in that: The feature aggregation secondary subunit is used to: Inputting the dynamic energy spatiotemporal transfer value of each local time series implicitly associated feature vector of the historical power generation into the energy transfer gating module based on the mask function to obtain a sequence of dynamic energy spatiotemporal transfer weights; Based on the sequence of dynamic energy spatiotemporal transfer weights, a weighted sum of the sequence of implicitly associated characteristic vectors of the local time series of historical power generation is calculated to obtain a characteristic transfer aggregation modulation vector of the local time series of historical power generation; The position-wise sum of the local time series feature transfer aggregation modulation vector of the historical power generation and the local time series implicit correlation feature vector of the current power generation is calculated to obtain the significant aggregation propagation representation vector of the power generation time series.

7. The flexible power supply system based on photovoltaic energy storage according to claim 6 is characterized in that: The characteristic interaction energy calculation subunit is used to: Performing a nonlinear transformation on the power generation time series significant aggregation propagation representation vector to obtain a nonlinearly transformed power generation time series significant aggregation propagation representation vector; Calculating the interactive correlation matrix between the significant aggregate propagation representation vector of the power generation time series after the nonlinear transformation and each local time series feature vector of the weather data in the sequence of the local time series feature vector of the weather data to obtain a sequence of power generation-weather interactive correlation matrices; The characteristic interaction energy factor of each power generation-weather interaction correlation matrix in the sequence of power generation-weather interaction correlation matrices is calculated to obtain the sequence of power generation-weather characteristic interaction energy factors.

8. The flexible power supply system based on photovoltaic energy storage according to claim 7 is characterized in that: The feature interaction global fusion subunit is used to: Inputting the sequence of power generation-weather characteristic interaction energy factors into a Sigmoid function to obtain a sequence of normalized power generation-weather characteristic interaction energy factors; The sequence of normalized power generation-weather characteristic interaction energy factors is used as a sequence of weights, and the position-weighted sum of the sequence of power generation-weather interaction correlation matrices is calculated to obtain the power generation-weather local-global interaction response matrix.

9. The flexible power supply system based on photovoltaic energy storage according to claim 8 is characterized in that: The power generation prediction unit is used to: The power generation-weather local-global interactive response matrix is ​​input into a decoder-based power generation prediction module to obtain a predicted decoded value of the power generation of the photovoltaic power station.

10. A flexible power supply method based on photovoltaic energy storage, used in the flexible power supply system based on photovoltaic energy storage as claimed in any one of claims 1 to 9, comprising: Real-time acquisition of photovoltaic power station information within a certain area, wherein the photovoltaic power station information includes power supply location and power supply amount; Acquire user information within the certain area, wherein the user information includes power consumption and power consumption location; Predict power generation based on historical power generation and historical weather data for each photovoltaic power station; Predict electricity consumption based on the user's historical monthly electricity consumption and historical working day and non-working day electricity consumption; The pre-storage capacity of each photovoltaic power station is determined according to the prediction result of the power generation of each photovoltaic power station and the prediction result of the power consumption.

Images