New energy power generation power prediction system and method

Abstract

The invention relates to the technical field of new energy power generation power prediction, particularly provides a new energy power generation power prediction system and method, and solves the problems of long-term and short-term dependency between data, slow model convergence speed and relatively high input space dimension. The system comprises a data module, a prediction model construction module and a test evaluation module. The method comprises the following steps: acquiring real-time environment sensing data, weather forecast and historical power generation power data, and combining to form a fusion data set; inputting the data set into a new energy power generation power prediction model, and predicting a time period and power generation power; the new energy power generation power prediction model performs dimension reduction on data features, constructs a convolutional neural network according to the data features and timestamp information, captures spatial features of the data features, and captures a long and short term dependency relationship by using a long and short term memory network layer. According to the method, the new energy power generation power prediction model is optimized, so that the new energy power generation power prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of renewable energy power generation prediction, and in particular to a renewable energy power generation prediction system and method. Background Art

[0002] The new energy power generation prediction system is one of the key technologies to achieve safe and economical operation of the new energy power system. With the transformation of the global energy structure and the proposal of the "dual carbon" strategic goal, the development and utilization of new energy such as wind energy and solar energy are increasing, and the prediction of their power generation has become particularly important. The power generation of new energy has significant intermittent and random characteristics, which poses a huge challenge to the dispatching and operation of the power system and the safety and stability of the power grid. Therefore, the development of an accurate and efficient new energy power generation prediction system is of great significance to improving the operating efficiency and reliability of the new energy power system.

[0003] Prior art 1, Chinese patent, application number: CN202311540463.2 provides a new energy power generation prediction method, system, equipment and medium to solve the technical problems of the existing new energy power generation prediction method, such as high model complexity, large amount of calculation, and poor accuracy. The method comprises: obtaining new energy power generation related data; inputting the new energy power generation related data into a pre-trained new energy power generation prediction model, and outputting the predicted new energy power generation; wherein the new energy power generation prediction model is constructed based on an RBF neural network, and the training data of the new energy power generation prediction model is based on wavelet denoising data. The data preprocessing technology of wavelet threshold denoising is adopted to reduce the influence of data noise on the prediction results. Through model optimization and adjustment, although the prediction accuracy is further improved, the real-time prediction of new energy power generation is realized, which provides strong support for the stable operation of the power system; however, the short-term dependency between data is not taken into account, which affects the stability of the model.

[0004] Prior art 2, Chinese patent, application number: CN202410144489.3 includes the following steps: data collection and processing, collecting weather forecast data for the area where the target power station is located, collecting historical power generation data of the target power station; environmental perception and dynamic data fusion, deploying environmental perception units, real-time monitoring of environmental changes around the power station, using dynamic data fusion technology, combining real-time environmental perception data with weather forecasts and historical power generation data to form comprehensive data; using the above comprehensive data, pre-training a large model based on deep learning, which can learn the complex relationship between renewable energy power generation and meteorological elements, and perform short-term power generation prediction. Although the prediction accuracy of renewable energy power generation has been significantly improved by combining the advanced GRU neural network model and comprehensive data processing, the large amount of calculation has slowed down the convergence of the model.

[0005] Prior art three, Chinese patent, application number: CN202311640746.4 discloses a prediction method, device, equipment and medium for renewable energy power generation, by obtaining historical unit data of renewable energy power plants; clustering renewable energy power plants based on historical unit data to generate multiple unit partitions of renewable energy power plants; extracting features from historical operation data and historical meteorological data of each unit partition to obtain power generation feature data of the unit partition; training a preset LSTM-CNN model based on power generation feature data until the preset LSTM-CNN model reaches a preset convergence condition to obtain a power generation prediction model for renewable energy power plants; using the power generation prediction model, predicting the power generation data of renewable energy power plants in the future target time period according to the real-time meteorological data of renewable energy power plants. Although the accuracy of the prediction of renewable energy power generation capacity is improved, the input space dimension is high, which affects the stability of the model.

[0006] At present, the existing technologies 1, 2 and 3 do not take into account the short-term dependency between data, have large amount of calculation, slow model convergence speed and high input space dimension. Therefore, the present invention provides a new energy power generation prediction system and method. Summary of the invention

[0007] The main purpose of the present invention is to provide a new energy power generation prediction system and method to solve the problems existing in the prior art that the short-term dependency between data is not taken into account, the amount of calculation is large, the model convergence speed is slow, and the input space dimension is high.

[0008] In order to achieve the above object, the present invention adopts the following technical scheme:

[0009] In one aspect of the present invention, a new energy power generation prediction system is provided, comprising:

[0010] The data module is responsible for collecting historical meteorological data, real-time data, and historical power generation data, and performing pre-processing operations such as cleaning, denoising, and formatting on the collected data, and sending the pre-processed data to the prediction model building module;

[0011] The prediction model building module is responsible for integrating historical meteorological data, real-time data and historical power generation data to form a fused data set; preprocessing the fused data set to extract data features; using dimensionality reduction technology to compress data features, and building a convolutional neural network based on the extracted dimensionality reduction data features and timestamp information. The convolutional neural network will capture spatial features, use the long-short-term memory network layer to capture long-short-term dependencies, and adjust the prediction model building module parameters through optimization algorithms;

[0012] The test and evaluation module is responsible for analyzing the prediction result error of the renewable energy power generation prediction model and making adjustments if the error exceeds the threshold.

[0013] As a further improvement of the present invention, the data module comprises:

[0014] The data collection submodule is responsible for connecting to the real-time data interface of the meteorological department, collecting historical meteorological data of the target power station area, and deploying environmental sensing units in the target power station area to monitor the changes in the surrounding environment of the target power station in real time, and implement real-time data collection for the target power station. The data collection includes meteorological elements such as temperature, humidity, wind direction and wind speed, as well as environmental parameters such as changes in vegetation coverage; collecting historical power generation data of the target power station; and transmitting the collected historical meteorological data, real-time data and historical power generation data to the data processing center;

[0015] The fusion data set submodule is responsible for preprocessing the historical meteorological data, real-time data set and historical power generation data in the data processing center, including data cleaning and standardization; extracting target features from the preprocessed historical meteorological data and real-time data set historical power generation data, including meteorological elements and power generation output at different timestamps; performing time synchronization processing on the features, aligning all historical meteorological data, real-time data sets and historical power generation data on the time axis to form a fusion data set;

[0016] Among them, historical meteorological data, real-time data sets, and historical power generation data may have incompletely synchronized data in time, and the difference method is used to fill in the missing time points;

[0017] The evaluation data set submodule is responsible for evaluating the fused data set, setting the threshold of the impact of different features on power generation prediction, comparing the impact of different feature fusions on power generation prediction, and dynamically adjusting the new energy power generation prediction model if the threshold is exceeded;

[0018] As a further improvement of the present invention, the fusion data set submodule comprises:

[0019] The missing timestamp unit is responsible for time synchronization processing of features, aligning all historical meteorological data, real-time data sets and historical power generation data on the time axis, marking the timestamps that do not exist in the time axis of historical meteorological data, real-time data sets and historical power generation data and marking them as missing timestamps;

[0020] The timestamp temperature value unit is responsible for performing a linear difference between the two data before and after the time point to calculate the temperature value of the timestamp;

[0021] The supplementary timestamp unit is responsible for setting a threshold for the data change trend by comparing the data change trend before filling. If the data change trend is greater than the threshold, the filled data is unqualified and the parameters are readjusted; if the data change trend is less than the threshold, the filled data is qualified, and the filled data is stored again in the data processing center, and the relevant timestamp information is updated.

[0022] As a further improvement of the present invention, the prediction model building module comprises:

[0023] The feature extraction submodule is responsible for extracting feature data from the fused data set, performing dimensionality reduction on the data, extracting the main input components, and reducing the data dimension; training the new energy power generation prediction model based on the fused data set after dimensionality reduction;

[0024] The prediction module submodule is responsible for the prediction model of renewable energy power generation. It divides the preset time period into multiple sub-periods. Each sub-period corresponds to a time step. The time step includes every hour. In each sub-period, the power generation of the prediction result of the previous time step is used as input for the next round of prediction.

[0025] The prediction error submodule is responsible for setting the prediction error threshold and monitoring the prediction error of the prediction model building module in real time. If the prediction error is greater than the prediction error threshold, the prediction error will be analyzed and the prediction model building module will be dynamically adjusted.

[0026] As a further improvement of the present invention, the feature extraction submodule comprises:

[0027] The convolutional neural network unit is responsible for receiving the reduced-dimensional data features and timestamp information of the new energy power generation prediction model, and building a convolutional neural network based on dynamic time series; the convolutional neural network design includes an input layer, a feature extraction convolution layer, a feature optimization selection pooling layer, and a long short-term memory network layer;

[0028] The dependency unit is responsible for transmitting data features from the input layer to the convolution layer. The convolution layer learns different features through the convolution kernel, captures spatial features, generates feature atlases, and sends the feature atlases to the pooling layer. The pooling layer averages the feature atlases, corrects the average pooled feature atlases, and corrects the data features to the same timestamp. It captures long-term dependencies in the time series.

[0029] When the vector value of the dependency relationship of the memory unit in the input gate of the long short-term memory network layer is close to 1, the element will be updated, and when it is close to 0, the element will remain unchanged; when the vector value of the dependency relationship of the memory unit in the forget gate is close to 1, the element will be retained, and if it is close to 0, the element will be abandoned; when the vector value of the dependency relationship of the memory unit in the output gate is close to 1, the element will be output, and if it is close to 0, the unit will not be output; wherein, element is a memory element in the memory unit, which is used to memorize the vector value of the dependency relationship;

[0030] The feature fusion unit is responsible for mapping the long-term dependencies output by the LSTM network layer to the final prediction results, fusing the features extracted by the convolution layer and the LSTM network layer, and using the output of the convolution layer as the input of the LSTM network layer; the fused features are used to train the new energy power generation prediction model.

[0031] As a further improvement of the present invention, the prediction model building module comprises:

[0032] The prediction value submodule is responsible for predicting the power generation within a preset time period using the pre-trained new energy power generation prediction model, and forming a prediction value based on the learning results of historical prediction data;

[0033] The identification submodule is responsible for comparing the predicted value of the renewable energy power generation prediction model with the actual power generation, calculating the prediction error, analyzing the calculated prediction error, and identifying the main sources of the error, which mainly include improper settings of the renewable energy power generation prediction model, numerical quality problems, or changes in the external environment.

[0034] The optimization module is responsible for adjusting the parameters of the renewable energy power generation prediction model according to the error analysis results, and optimizing the weight and structure of the renewable energy power generation prediction model.

[0035] As a further improvement of the present invention, the sub-module is optimized, comprising:

[0036] The feature extraction unit is responsible for performing principal component analysis on the error analysis results, extracting features that are strongly correlated with the generated power, and inputting the relevant features into the adaptive weight device;

[0037] The adjustment unit is responsible for the adaptive weight device to dynamically adjust the weight according to the features that have the greatest impact on the power generation among the relevant features; the adaptive weight device is integrated into the new energy power generation prediction model, and the historical data is used for training. During the training process, the adaptive weight device dynamically adjusts the weight according to the performance of the features;

[0038] The analysis and comparison unit is responsible for comparing the predicted value with the actual power generation after the model training is completed, calculating the prediction error, and adjusting the parameters of the adaptive weight according to the error analysis results.

[0039] As a further improvement of the present invention, the test evaluation module comprises:

[0040] Error average submodule, prediction accuracy submodule, calculation evaluation submodule,

[0041] The error average submodule is responsible for calculating the average error of the power generation of each time mode in the past n days based on the time period and power generation prediction statistical deviation.

[0042] The prediction accuracy submodule is responsible for judging that the power generation prediction is accurate if the absolute difference between the power generation and the actual value is less than or equal to 2, otherwise it is inaccurate;

[0043] The calculation and evaluation submodule is responsible for evaluating and verifying the power generation accuracy and root mean square error, average error and mean absolute error of the power generation forecast and the revision field; among them, the power generation forecast accuracy is the percentage of the power generation forecast error |Fi-Oi| less than or equal to 2, and the positive and negative errors in the average error calculation formula offset each other.

[0044] Another aspect of the present invention provides a method for predicting power generation from renewable energy, comprising:

[0045] Collect weather forecast data for the target power station area, including temperature, humidity, wind direction and wind speed, and collect historical power generation data for the target power station, including power output at different timestamps; integrate environmental perception and dynamic data, deploy environmental perception units, monitor environmental changes around the power station in real time, and combine real-time environmental perception data, weather forecasts and historical power generation data to form a fused data set;

[0046] Based on the fusion data set, the new energy power generation prediction model is pre-trained, and the preset time period and power generation prediction are performed. The prediction error of the new energy power generation prediction model is monitored in real time, and the parameters of the new energy power generation prediction model are adjusted in real time according to the prediction error.

[0047] The optimized new energy power generation prediction model is used to predict the new energy power generation, and the prediction accuracy and performance of the new energy power generation prediction model are evaluated by comparing it with the actual power generation prediction data.

[0048] The present invention significantly improves the accuracy and stability of new energy power generation prediction by combining multiple methods such as longitudinal matching of historical data, selection of meteorological characteristics and optimization of SVM model parameters, and outlier detection based on Boosting integrated framework. For example, the method based on longitudinal matching of historical data realizes power prediction for the next few hours or days by excluding distorted data, with small prediction error, simple and practical. The outlier detection method based on Boosting integrated framework effectively solves the problem of low detection accuracy of traditional single model on unbalanced data, and improves the accuracy of outlier detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0050] Figure 1 This is a module block diagram of the new energy power generation prediction system provided in Example 1 of the present invention;

[0051] Figure 2 This is a block diagram of a data module provided in Embodiment 2 of the present invention;

[0052] Figure 3 This is a block diagram of the fusion data set submodule provided in Example 3 of the present invention;

[0053] Figure 4 This is a block diagram of a prediction model construction module provided in Example 4 of the present invention;

[0054] Figure 5 This is a block diagram of a feature extraction submodule provided in Embodiment 5 of the present invention;

[0055] Figure 6 This is a block diagram of a prediction model construction module provided in Example 6 of the present invention;

[0056] Figure 7 This is a block diagram of the optimization submodule provided in Embodiment 7 of the present invention;

[0057] Figure 8 This is a block diagram of a test evaluation module provided in Embodiment 8 of the present invention;

[0058] Fig. 9 This is a flow chart of the new energy power generation prediction method provided in Example 9 of the present invention;

[0059] Fig.10 A block diagram of an electronic device provided by the present invention;

[0060] Fig.11 A block diagram of a computer-readable storage medium provided for the present invention. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present invention will be described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0062] In the following, the terms "first", "second", etc. are used only for convenience of description and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0063] In the present invention, unless otherwise clearly specified and limited, the term "connection" should be understood in a broad sense, for example, "connection" can be a fixed mechanical connection, or a detachable mechanical connection, or integrated; or, "connection" can be a direct connection, or an indirect connection through an intermediate medium. In addition, unless otherwise clearly specified and limited, the term "coupling" should be understood in a broad sense, for example, "coupling" can be a direct electrical connection, such as physical contact and electrical conduction between two components, and can also be understood as electrical connection between different components in a circuit structure through physical lines such as copper foil or wires on a printed circuit board (PCB) that can transmit electrical signals to transmit electrical signals; or, "coupling" can be an indirect electrical connection between two components through an intermediate medium; or, "coupling" can be an electrical connection between two components in an air-spaced / non-contact manner, for example, two components are electrically connected by capacitive coupling to transmit electrical signals.

[0064] In the embodiments of the present invention, directional terms such as "up", "down", "left" and "right" may include but are not limited to being defined relative to the orientation of the components schematically placed in the drawings. It should be understood that these directional terms may be relative concepts, which are used for relative description and clarification, and may change accordingly according to changes in the orientation of the components in the drawings.

[0065] The embodiments of the present invention can be used for smart grid construction, new energy power station operation optimization, energy storage system optimization, power system optimization decision-making, etc.

[0066] The key technical features of the embodiments of the present invention are as follows: The problems existing in the prior art, such as not taking into account the short-term dependency between data, large amount of calculation, slow model convergence speed, high input space dimension, etc., are solved by adjusting the parameters of the new energy power generation prediction model through the optimization algorithm. This embodiment reduces the amount of calculation and improves the accuracy and adaptability of the prediction model through deep learning and optimization algorithms, thereby realizing accurate prediction of new energy power generation.

[0067] Embodiment 1:

[0068] like Figure 1 As shown, the embodiment of the present invention provides a new energy power generation power prediction system, comprising:

[0069] Data module 1 is responsible for collecting historical meteorological data, real-time data and historical power generation data, and performing pre-processing operations such as cleaning, denoising and formatting on the collected data, and sending the pre-processed data to prediction model building module 2;

[0070] Prediction model building module 2 is responsible for integrating historical meteorological data, real-time data and historical power generation data to form a fused data set; preprocessing the fused data set to extract data features; using dimensionality reduction technology to compress data features, and building a convolutional neural network based on the extracted dimensionality reduction data features and timestamp information. The convolutional neural network will capture spatial features, use the long-short-term memory network layer to capture long-term and short-term dependencies, and adjust the parameters of the prediction model building module through the optimization algorithm;

[0071] The test evaluation module 3 is responsible for analyzing the prediction result error of the renewable energy power generation prediction model and making adjustments if the error exceeds the threshold.

[0072] In the above embodiment, data module 1 is responsible for collecting historical meteorological data, real-time data and historical power generation data, and performs pre-processing operations such as cleaning, denoising and formatting on the collected data, and sends the pre-processed data to prediction model construction module 2; ensuring the quality and consistency of the input data, and providing a reliable data basis for subsequent prediction models. Prediction model construction module 2 uses historical meteorological data, real-time data and historical power generation data to train the new energy power generation prediction model, and adjusts the parameters of the new energy power generation prediction model through optimization algorithms. This embodiment improves the accuracy and adaptability of the prediction model through deep learning and optimization algorithms, thereby achieving accurate prediction of new energy power generation. Test evaluation module 3 performs prediction result error analysis on the new energy power generation prediction model, and makes adjustments if the error exceeds the threshold. This embodiment ensures the accuracy and reliability of the prediction results through error analysis and model adjustment, thereby improving the overall performance of new energy power prediction.

[0073] Embodiment 2:

[0074] like Figure 2 As shown, based on Example 1, the data module 1 provided in this embodiment of the present invention specifically includes:

[0075] The data collection submodule 11 is responsible for connecting to the real-time data interface of the meteorological department, collecting historical meteorological data of the target power station area, and deploying environmental sensing units in the target power station area to monitor the changes in the surrounding environment of the target power station in real time, and implement real-time data collection for the target power station, where the data collection includes meteorological elements such as temperature, humidity, wind direction and wind speed, as well as environmental parameters such as changes in vegetation coverage; collecting historical power generation data of the target power station; and transmitting the collected historical meteorological data, real-time data and historical power generation data to the data processing center;

[0076] The fusion data set submodule 12 is responsible for preprocessing the historical meteorological data, real-time data set and historical power generation data in the data processing center, including data cleaning and standardization; extracting target features from the preprocessed historical meteorological data and real-time data set historical power generation data, including meteorological elements and power generation output at different timestamps; performing time synchronization processing on the features, aligning all historical meteorological data, real-time data sets and historical power generation data on the time axis to form a fusion data set;

[0077] Among them, historical meteorological data, real-time data sets, and historical power generation data may have incompletely synchronized data in time, and the difference method is used to fill in the missing time points;

[0078] The evaluation data set submodule 13 is responsible for evaluating the fused data set, setting the threshold of the impact of different features on the power generation prediction, comparing the impact of different feature fusions on the power generation prediction, and dynamically adjusting the new energy power generation prediction model if the threshold is exceeded;

[0079] In the above embodiment, the data acquisition submodule 11 is a data interface connected to the meteorological department in real time, collects historical meteorological data of the target power station area, and deploys an environmental perception unit in the target power station area to monitor the changes in the surrounding environment of the target power station in real time, and implements real-time data collection for the target power station. These data include meteorological elements such as temperature, humidity, wind direction and wind speed, as well as environmental parameters such as changes in vegetation coverage. In addition, the module also collects historical power generation data of the target power station and transmits these data to the data processing center. Through the method of this embodiment, the data acquisition submodule 11 can provide comprehensive meteorological and environmental data, providing a basis for subsequent data processing and analysis. The fusion data set submodule 12 pre-processes historical meteorological data, real-time data sets and historical power generation data, including data cleaning and standardization. By extracting features, such as meteorological elements and power generation output at different timestamps, and performing time synchronization processing, all historical meteorological data, real-time data sets and historical power generation data are aligned on the time axis to form a fused data set. In addition, the module also uses the difference method to fill in the missing time points to ensure the integrity and consistency of the data. Through these processes, the fusion data set submodule 12 can provide a high-quality fusion data set to provide support for subsequent analysis and prediction. The evaluation data set submodule 13 evaluates the fused data set, sets a threshold for the impact of different features on the power generation prediction, and compares the impact of different feature fusions on the power generation prediction. If the threshold is exceeded, the new energy power generation prediction model is dynamically adjusted. Through the method of this embodiment, the evaluation data set submodule 13 can ensure the accuracy and reliability of the prediction model, thereby improving the accuracy of the power generation prediction.

[0080] Embodiment 3:

[0081] like Figure 3 As shown, based on Example 2, the fusion data set submodule 12 provided in this embodiment of the present invention specifically includes:

[0082] The missing timestamp unit 121 is responsible for time synchronization processing of the features, aligning all historical meteorological data, real-time data sets and historical power generation data on the time axis, marking the timestamps that do not exist in the historical meteorological data, real-time data sets and historical power generation data on the time axis and marking them as missing timestamps;

[0083] The timestamp temperature value unit 122 is responsible for performing a linear difference between the two data before and after the time point to calculate the temperature value of the timestamp;

[0084] The supplementary timestamp unit 123 is responsible for setting a threshold for the data change trend by comparing the data change trend before filling. If the data change trend is greater than the threshold, the filled data is unqualified and the parameters are readjusted; if the data change trend is less than the threshold, the filled data is qualified, and the filled data is stored again in the data processing center and the relevant timestamp information is updated.

[0085] In this embodiment, the missing timestamp unit 121 is responsible for time synchronization processing of the features to ensure that all historical meteorological data, real-time data sets and historical power generation data are aligned on the time axis; all timestamps that do not exist on the time axis will be marked as missing timestamps;

[0086] The timestamp temperature value unit 122 is responsible for performing a linear difference based on two known data points before and after to calculate the temperature value corresponding to the timestamp;

[0087] Assume that the data before the missing timestamp is time point A(X 1 , Y 1 ), the missing timestamp is followed by time point B(X 2 , Y 2 ), then the value C(X, Y) corresponding to any point between points A and B can be calculated as:

[0088]

[0089] Complex interpolation model, when predicting temperature data, build the following four models to enhance the data fitting ability:

[0090] Polynomial interpolation (high-order polynomials)

[0091] Y poly =c 0 +c 1 (XX 1 )+c 2 (XX 1 ) 2 +c 3 (XX 1 ) 3 +c 4 (XX 1 ) 4 +…+c n (XX 1 ) n

[0092] Among them, c 0 , c 1 , c 2 , ..., c n It is the coefficient obtained by fitting historical data.

[0093] Piecewise linear interpolation

[0094]

[0095] Fourier series expansion

[0096]

[0097] where a n and b n is the Fourier coefficient, T is the period length, X 0 The base time point.

[0098] Stochastic Process Modeling

[0099] Y stochastic =Y deterministic +∈,∈~N(0,σ 2 )

[0100] Y stochastic is the combined result of the above models, and ∈ is the random error term.

[0101] The final temperature forecast value is obtained by weighted combination of the above different models:

[0102] Y final =αY poly +βY piecewise +γY Fourier +δY stochastic +k(XX 1 )(XX 2 )(X-μ)

[0103] Among them, α, β, γ, and δ are weight coefficients, which are obtained through historical data regression analysis or cross-validation;

[0104] k(XX 1 )(XX 2 )(X-μ) is a nonlinear term used to better capture complex data patterns.

[0105] In the above embodiment, the missing timestamp unit 121 aligns the historical meteorological data, the real-time data set and the historical power generation data on the time axis through time synchronization processing, and marks the time point that does not exist on the time axis, which is marked as a missing timestamp. This embodiment ensures the time consistency of the data and provides a basis for subsequent data analysis and processing. The timestamp temperature value unit 122 calculates the temperature value of the missing time point through linear difference. Specifically, it uses two data points (point A and point B) before and after the time point for interpolation calculation. The method of this embodiment can effectively fill in the missing data and improve the continuity and integrity of the data. The supplementary timestamp unit 123 sets a threshold to judge the eligibility of the filled data by comparing the data change trend before filling. If the data change trend is greater than the threshold, the parameters are readjusted; if it is less than the threshold, the filled data is qualified, and the filled data is stored again in the data processing center, and the relevant timestamp information is updated. This embodiment ensures the accuracy and reliability of the filled data and avoids errors caused by improper data filling.

[0106] Embodiment 4:

[0107] like Figure 4 As shown, based on Example 1, the prediction model building module 2 provided in this embodiment of the present invention specifically includes:

[0108] The feature extraction submodule 21 is responsible for extracting feature data from the fused data set, performing dimensionality reduction processing on the data, extracting the main input components, and reducing the data dimension; training the new energy power generation prediction model based on the fused data set after dimensionality reduction processing;

[0109] The prediction module submodule 22 is responsible for the prediction model of the power generation of new energy. The preset time period is divided into multiple sub-periods, each sub-period corresponds to a time step, and the time step includes every hour. In each sub-period, the power generation of the prediction result of the previous time step is used as input for the next round of prediction;

[0110] The prediction error submodule 23 is responsible for setting the prediction error threshold and monitoring the prediction error of the prediction model building module in real time. If the prediction error is greater than the prediction error threshold, the prediction error will be analyzed and the prediction model building module will be dynamically adjusted.

[0111] In the above embodiment, the feature extraction submodule 21 extracts the main input components and reduces the data dimension by performing feature data extraction and dimensionality reduction processing on the fusion data set. This embodiment helps to improve the efficiency of data processing and the training speed of the model, and can also reduce the complexity of the model, making it easier to understand and maintain. The prediction module submodule 22 divides the preset time period into multiple sub-periods, each sub-period corresponds to a time step, and the time step includes every hour. In each sub-period, the power generation result of the prediction result of the previous time step is used as input for the next round of prediction. The method of this embodiment can improve the accuracy and real-time performance of the prediction, because it takes into account the dynamic changes of the time series data, so that the model can better capture the time series characteristics of photovoltaic power generation. After the training of the new energy power generation model is completed, the prediction error submodule 23 sets the prediction error threshold and monitors the prediction error of the new energy power generation model in real time. If the prediction error is greater than the threshold, the prediction error will be analyzed and the new energy power generation model will be dynamically adjusted. The method of this embodiment can improve the robustness and adaptability of the model, ensuring that when the prediction error exceeds expectations, the model can be adjusted in time to improve the prediction accuracy.

[0112] Embodiment 5:

[0113] like Figure 5 As shown, based on Example 4, the feature extraction submodule 21 provided in this embodiment of the present invention specifically includes:

[0114] The convolutional neural network unit 211 is responsible for receiving the data features and timestamp information after dimension reduction in the new energy power generation prediction model, and constructing a convolutional neural network based on dynamic time series; the convolutional neural network design includes an input layer, a feature extraction convolution layer, a feature optimization selection pooling layer, and a long short-term memory network layer;

[0115] The dependency unit 212 is responsible for transmitting the data features from the input layer to the convolution layer. The convolution layer learns different features through the convolution kernel, captures spatial features, generates feature atlases, and sends the feature atlases to the pooling layer. The pooling layer averages the feature atlases, corrects the average pooled feature atlases, corrects the data features to the same timestamp, and captures long-term dependencies in the time series.

[0116] When the vector value of the dependency relationship of the memory unit in the input gate of the long short-term memory network layer is close to 1, the element will be updated, and when it is close to 0, the element will remain unchanged; when the vector value of the dependency relationship of the memory unit in the forget gate is close to 1, the element will be retained, and if it is close to 0, the element will be abandoned; when the vector value of the dependency relationship of the memory unit in the output gate is close to 1, the element will be output, and if it is close to 0, the unit will not be output; wherein, element is a memory element in the memory unit, which is used to memorize the vector value of the dependency relationship;

[0117] The feature fusion unit 213 is responsible for mapping the long-term dependency relationship output by the long short-term memory network layer to the final prediction result, and fusing the features extracted by the convolution layer and the long short-term memory network layer, and using the output of the convolution layer as the input of the long short-term memory network layer; the fused features are used to train the new energy power generation prediction model.

[0118] In the above embodiment, the convolutional neural network unit 211 is responsible for receiving the data features and timestamp information after dimensionality reduction, and constructing a convolutional neural network based on dynamic time series. The network design includes an input layer, a feature extraction convolution layer, a feature optimization selection pooling layer, and a long short-term memory network layer. The input layer transmits the data features to the convolution layer, and the convolution layer learns different features through the convolution kernel, captures spatial features, generates a feature atlas, and sends the feature atlas to the pooling layer. The pooling layer averages the feature atlas, and corrects the average pooled feature atlas, corrects the data features to the same timestamp, and captures the long-term dependencies in the time series. The dependency unit 212 is responsible for processing the input gate, forget gate, and output gate of the long short-term memory network layer. The feature fusion unit 213 is responsible for mapping the long-term dependencies output by the long short-term memory network layer to the final prediction result, and fuses the features extracted by the convolution layer and the long short-term memory network layer. The output of the convolution layer is used as the input of the long short-term memory network layer, and the fused features are used to train the new energy power generation prediction model.

[0119] Embodiment 6:

[0120] like Figure 6 As shown, based on Example 1, the prediction model building module 2 provided in this embodiment of the present invention specifically includes the following modules:

[0121] The prediction value submodule 24 is responsible for predicting the power generation within a preset time period using the pre-trained new energy power generation prediction model, and forming a prediction value based on the learning results of the historical prediction data;

[0122] The identification submodule 25 is responsible for comparing the predicted value of the new energy power generation prediction model with the actual power generation, calculating the prediction error, analyzing the calculated prediction error, and identifying the main source of the error, which mainly includes factors such as improper setting of the new energy power generation prediction model, numerical quality problems, or changes in the external environment;

[0123] The optimization module 26 is responsible for adjusting the parameters of the renewable energy power generation prediction model according to the error analysis results, and optimizing the weight and structure of the renewable energy power generation prediction model.

[0124] In the above embodiment, the prediction value submodule 24 is responsible for predicting the power generation within a preset time period using the pre-trained new energy power generation prediction model. According to the learning results of the historical prediction data, the prediction value is formed. Through the learning of the pre-trained model and historical data, the future power generation can be predicted more accurately, providing basic data for subsequent error analysis and model optimization. The identification submodule 25 is responsible for comparing the predicted value of the new energy power generation prediction model with the actual power generation, and calculating the prediction error. The calculated prediction error is analyzed to identify the main source of the error, which mainly includes factors such as improper setting of the new energy power generation prediction model, numerical quality problems or changes in the external environment. By comparing the predicted value with the actual value, the accuracy of the prediction can be quantified, and the key factors affecting the prediction accuracy can be identified, providing a basis for model optimization. The optimization module 26 is responsible for adjusting the parameters of the new energy power generation prediction model according to the error analysis results, and optimizing the weight and structure of the new energy power generation prediction model. By adjusting the model parameters and optimizing the model structure, the accuracy and robustness of the prediction model can be improved, the prediction error can be reduced, and the overall prediction performance can be improved.

[0125] Embodiment 7:

[0126] like Figure 7 As shown, based on Example 6, the optimization submodule 26 provided in this embodiment of the present invention specifically includes the following modules:

[0127] The feature extraction unit 261 is responsible for performing principal component analysis on the error analysis results, extracting features that are strongly correlated with the generated power, and inputting the relevant features into the adaptive weight device;

[0128] The adjustment unit 262 is responsible for the adaptive weight device to dynamically adjust the weight according to the feature that has the greatest impact on the power generation among the relevant features; the adaptive weight device is integrated into the new energy power generation prediction model, and the historical data is used for training. During the training process, the adaptive weight device dynamically adjusts the weight according to the performance of the feature;

[0129] The analysis and comparison unit 263 is responsible for comparing the predicted value with the actual generated power after the model training is completed, calculating the prediction error, and adjusting the parameters of the adaptive weight according to the error analysis result.

[0130] In the above embodiment, the feature extraction unit 261 processes the error analysis results through principal component analysis, extracts features that are strongly related to the generated power, and inputs these features into the adaptive weight device. This embodiment can effectively reduce the complexity of data calculation, extract the features that have the greatest impact on the generated power, thereby improving the accuracy of the subsequent prediction model. The adjustment unit 262 is responsible for dynamically adjusting the weights according to the features that have a great impact on the generated power among the relevant features. Through the adaptive weight device, this embodiment can adjust the model parameters in real time to adapt to the changes in the importance of different features, thereby improving the prediction accuracy and adaptability of the model. After the model training is completed, the analysis and comparison unit 263 compares the predicted value with the actual generated power, calculates the prediction error, and adjusts the parameters of the adaptive weight according to the error analysis results. By comparing the predicted value with the actual value, this embodiment can promptly discover the deficiencies of the model, and optimize the model by adjusting the adaptive weight parameters, thereby improving the accuracy and reliability of the prediction.

[0131] Embodiment 8:

[0132] like Figure 8 As shown, based on Example 1, the test evaluation module 3 provided in this embodiment of the present invention specifically includes:

[0133] The error average submodule 31 is responsible for calculating the average error of the power generation of each time mode in the past n' days according to the time period and the power generation prediction statistical deviation.

[0134] The prediction accuracy submodule 32 is responsible for judging that the prediction of the generated power is accurate if the absolute difference between the generated power and the actual value is less than or equal to 2, otherwise it is inaccurate;

[0135] The calculation and evaluation submodule 33 is responsible for evaluating and checking the power generation accuracy and root mean square error, average error and mean absolute error of the power generation forecast and the revision field; among which, the power generation forecast accuracy is in the power generation forecast error |F i -O i | is less than or equal to 2 percentage, the average error calculation formula positive and negative errors cancel each other out.

[0136] In this embodiment, the error average submodule 31 sets the parameters:

[0137] The corrected temperature value of target station i within the forecast validity period (t=24) hours.

[0138] The temperature value of the target site obtained by interpolation method.

[0139] The correction term, which represents the error between the actual temperature and the interpolated temperature, is calculated using the following formula:

[0140]

[0141] The actual observed temperature of target site i at time j and duration t, ζ j External factors that affect the prediction accuracy, such as climate change, wind speed, humidity, etc., for the jth observation.

[0142] The mean of all external factors over the past n′ days is used as a normalization to make it consistent with the individual external factors ζ j In contrast, the λ adjustment factor is used to control the influence of external factors on the correction term. λ>1 indicates an enhanced influence, and λ<1 indicates a weakened influence.

[0143] Calculation of the corrected target site temperature:

[0144]

[0145] θ is the parameter for temperature attenuation, which controls the magnitude of the correction term's effect on temperature. As the temperature increases, the effect decreases exponentially;

[0146] The prediction accuracy submodule 32 sets the parameters:

[0147] Nr is the number of predictions that meet the accuracy conditions; Nf is the total number of predictions;

[0148] The calculation formula of prediction accuracy T′ is:

[0149]

[0150] m represents the total number of times considered, represents the product of the results of all m time periods to capture the subtle changes of each time period, θ k represents the adjustment factor at each time, which is used to consider the enhancement of external influences at a specific time, σ k It represents the standard deviation of each time, reflecting the stability of the internal prediction of the time, and D represents the cumulative error in different prediction sets, which is calculated as follows: β″ represents the scaling parameter, which is used to control the influence of the error on the prediction accuracy;

[0151] Calculation of root mean square error (RMSE) and mean absolute error (MAE):

[0152]

[0153] F i represents the predicted power generation value for the i-th time, O i represents the actual power generation value of the ith time, E i =|Fi -O i | represents the absolute error of the ith time, It represents the mean of all absolute errors and is used for normalized evaluation. γ′ represents the nonlinear adjustment factor. If γ′>1, the weight of high error prediction is increased to enhance the sensitivity of the model.

[0154]

[0155] T max represents the maximum power generation value in all observation time periods, δ″ represents the parameter controlling the enhancement effect of the mean absolute error, which is responsible for adjusting the impact of large errors;

[0156] The calculation conditions for calculating the evaluation accuracy of the evaluation submodule 33 are:

[0157]

[0158] N accurate It represents the number of qualified predictions that meet the prediction accuracy requirements, α′ represents the adjustment factor that controls the overall accuracy result, affecting the adaptability of the accuracy to external factors, Represents the weighting function, which fine-tunes the overall accuracy based on the actual accuracy, N total represents the total number of predictions, β′ represents the adjustment factor, which is used to control the attenuation effect of the error term on the accuracy. Represents an adjustment function based on the overall error that can increase or decrease the stability of the overall accuracy.

[0159] The calculation formula of average error is:

[0160]

[0161] C i represents the control factor, which is used to adjust the matching degree between the prediction and the observed value according to the actual situation;

[0162] The formula for the final evaluation model is:

[0163]

[0164] δ′ represents the new adjustment coefficient, which is used to affect the additional error term, ∈′ represents the improvement strength of the control error in the final evaluation, φ(E) represents the coefficient of the function adjustment of the total error, which helps to weight the impact of errors of different sizes on the final evaluation, α′ represents the adjustment factor that controls the overall accuracy result, the prediction accuracy T′, β′ represents the adjustment factor, γ′ represents the nonlinear adjustment factor, δ″ represents the parameter that controls the mean absolute error enhancement effect, δ″ represents the parameter that controls the mean absolute error enhancement effect, N′ represents the number of samples when calculating the mean absolute error MAE, Nr is the number of predictions that meet the accuracy conditions, F i represents the predicted power generation value for the i-th time, O i Represents the actual power generation value of the i-th time.

[0165] In the above embodiment, the error average submodule 31 corrects the deviation of the power generation prediction by calculating the error average of the past n days. The error correction method helps to improve the accuracy of the power generation prediction, especially by making real-time adjustments based on historical errors. The prediction accuracy submodule 32 determines the accuracy of the prediction by judging whether the absolute difference between the power generation and the actual value is less than or equal to 2. If the conditions are met, the power generation prediction is considered accurate; otherwise, the prediction is considered inaccurate. This embodiment is simple and direct, and can quickly evaluate the accuracy of the prediction. The calculation and evaluation submodule 33 evaluates and verifies the power generation forecast and the revision field by calculating the power generation accuracy, root mean square error (RMSE), mean error (MAE) and mean absolute error (MAE); the indicators can fully reflect the accuracy and stability of the prediction, which helps to optimize the prediction model and improve the prediction accuracy.

[0166] Embodiment 9:

[0167] like Fig. 9 As shown, based on Embodiment 1 to Embodiment 8, the embodiment of the present invention provides a method for predicting power generation of new energy, comprising the following steps:

[0168] Step S100: Collect weather forecast data of the target power station area, including temperature, humidity, wind direction and wind speed, and collect historical power generation data of the target power station, including power generation output at different timestamps; integrate environmental perception and dynamic data, deploy environmental perception units, monitor environmental changes around the power station in real time, and combine real-time environmental perception data, weather forecast and historical power generation data to form a fused data set;

[0169] Step S200: pre-training a new energy power generation prediction model based on the fused data set, and performing a preset time period and power generation prediction, monitoring the prediction error of the new energy power generation prediction model in real time, and adjusting the parameters of the new energy power generation prediction model in real time according to the prediction error;

[0170] Step S300: The optimized new energy power generation prediction model is used to predict the new energy power generation, and the prediction accuracy and performance of the new energy power generation prediction model are evaluated by comparing the prediction data with the actual power generation prediction data.

[0171] In the above embodiment, in step S100, weather forecast data and historical power generation data of the target power station area are collected, and combined with real-time environmental perception data to form a fused data set; through the fusion of multi-source data, the comprehensiveness and accuracy of the data are improved, and more abundant input information is provided for the prediction model, thereby improving the accuracy and reliability of the prediction model. In step S200, the new energy power generation prediction model is pre-trained according to the fused data set, and the prediction error of the model is monitored in real time, and the model parameters are adjusted according to the error; through the pre-training of the model and the real-time error feedback mechanism, the model parameters are continuously optimized, so that the prediction model can better adapt to the changes in actual power generation, and the accuracy and adaptability of the prediction are improved. In step S300, the optimized new energy power generation prediction model is predicted, and the prediction accuracy and performance of the model are evaluated by comparing with the actual power generation; through the verification of actual data, the practicality and effectiveness of the prediction model are ensured, providing a scientific basis for the operation and management of the power station.

[0172] Fig.10 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present invention is shown.

[0173] The electronic device may include a central processing unit / microprocessor / main control chip, etc. 4; a storage medium 5, coupled to the central processing unit / microprocessor / main control chip, etc. 4, and storing computer executable instructions therein, for performing the steps of each method of an embodiment of the present invention when executed by the processor.

[0174] The central processing unit / microprocessor / main control chip 4 may include but is not limited to, for example, one or more processors or microprocessors.

[0175] The storage medium 5 may include, but is not limited to, for example, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (such as hard disk, floppy disk, solid state drive, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.).

[0176] In addition, the electronic device may also include (but not limited to) a data bus 6, an input / output bus / external bus / device bus 7, a display 8, and input / output devices 9 (eg, keyboard, mouse, speaker, etc.).

[0177] The central processing unit / microprocessor / main control chip etc. 4 can communicate with external devices ( 8 , 9 etc.) through the I / O bus 7 via a wired or wireless network (not shown).

[0178] The storage medium 5 may also store at least one computer executable instruction for executing the various functions and / or method steps in the embodiments described in the present technology when executed by the central processing unit / microprocessor / main control chip 4.

[0179] In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product, wherein one or more computer executable instructions are executed by a processor to perform the various functions and / or method steps in the embodiments described in the present technology.

[0180] Fig.11 A schematic diagram of a computer-readable storage medium according to an embodiment of the present invention is shown.

[0181] like Fig.11 As shown, instructions are stored on the non-transitory computer-readable storage medium 101, and the instructions are, for example, computer-readable instructions 10. When the computer-readable instructions 10 are executed by the processor, the various methods described above can be executed. The non-transitory computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may, for example, include random access memory (RAM) and / or cache memory (cache), etc. The non-transitory non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer-readable storage medium 101 can be connected to a computing device such as a computer, and then, when the computing device runs the computer-readable instructions 10 stored on the computer-readable storage medium 101, the various methods described above can be performed.

[0182] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0183] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0184] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0185] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and contains several instructions for executing all or part of the steps of the various embodiments of the method of the present invention through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program codes.

[0186] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A new energy power generation prediction system, characterized in that: Include: The data module is responsible for collecting historical meteorological data, real-time data, and historical power generation data, and performs cleaning, denoising, and formatting preprocessing operations on the collected data, and sends the preprocessed data to the prediction model building module; The prediction model building module is responsible for integrating historical meteorological data, real-time data, and historical power generation data to form a fused data set; Preprocess the fused data set to extract data features; use dimensionality reduction technology to compress data features, and build a convolutional neural network based on the extracted dimensionality reduction data features and timestamp information; the convolutional neural network will capture spatial features, use the long-short-term memory network layer to capture long-term and short-term dependencies, and adjust the parameters of the prediction model building module through the optimization algorithm; The test evaluation module is responsible for analyzing the prediction result error of the prediction model building module and making adjustments if the error exceeds the threshold; Evaluation formula for the test evaluation module: δ′ represents the new adjustment coefficient, which is used to affect the additional error term, ∈′ represents the improvement strength of the control error in the final evaluation, φ(E) represents the coefficient of the function adjustment of the total error, which helps to weight the impact of errors of different sizes on the final evaluation, α′ represents the adjustment factor that controls the overall accuracy result, the prediction accuracy T′, β′ represents the adjustment factor, γ′ represents the nonlinear adjustment factor, δ″ represents the parameter that controls the mean absolute error enhancement effect, δ″ represents the parameter that controls the mean absolute error enhancement effect, N′ represents the number of samples when calculating the mean absolute error MAE, Nr is the number of predictions that meet the accuracy conditions, F i represents the predicted power generation value for the i-th time, O i Represents the actual power generation value of the i-th time.

2. The new energy power generation prediction system according to claim 1, characterized in that: Data module, including: The data collection submodule is responsible for connecting to the real-time data interface of the meteorological department, collecting historical meteorological data of the target power station area, and deploying environmental sensing units in the target power station area to monitor the changes in the surrounding environment of the target power station in real time and implement real-time data collection for the target power station; The fusion data set submodule is responsible for preprocessing the historical meteorological data, real-time data set and historical power generation data in the data processing center; extracting target features from the preprocessed historical meteorological data and real-time data set historical power generation data; The evaluation dataset submodule is responsible for setting the threshold of the impact of different features on power generation prediction, comparing the impact of different feature fusions on power generation prediction, and dynamically adjusting the new energy power generation prediction model if the threshold is exceeded.

3. The new energy power generation prediction system according to claim 2, characterized in that: The fusion dataset submodule includes: The missing timestamp unit is responsible for time synchronization processing of features, aligning all historical meteorological data, real-time data sets and historical power generation data on the time axis, marking the timestamps that do not exist in the time axis of historical meteorological data, real-time data sets and historical power generation data and marking them as missing timestamps; The timestamp temperature value unit is responsible for performing a linear difference between the two data before and after the time point to calculate the temperature value of the timestamp; The supplementary timestamp unit is responsible for setting a threshold for the data change trend by comparing the data change trend before filling. If the data change trend is greater than the threshold, the filled data is unqualified and the parameters are readjusted; if the data change trend is less than the threshold, the filled data is qualified, and the filled data is stored again in the data processing center, and the relevant timestamp information is updated.

4. The new energy power generation prediction system according to claim 1, characterized in that: Prediction model building module, including: The feature extraction submodule is responsible for reducing the dimension of the fused data, extracting the input components, and reducing the data dimension; training the new energy power generation prediction model based on the fused data set after the dimension reduction process; The prediction module submodule is responsible for the prediction model of renewable energy power generation. It divides the preset time period into multiple sub-periods. Each sub-period corresponds to a time step. The time step includes every hour. In each sub-period, the power generation of the prediction result of the previous time step is used as input for the next round of prediction. The prediction error submodule is responsible for setting the prediction error threshold and monitoring the prediction error of the prediction model building module in real time. If the prediction error is greater than the prediction error threshold, the prediction error will be analyzed and the prediction model building module will be dynamically adjusted.

5. The new energy power generation prediction system according to claim 4, characterized in that: Feature extraction submodule, including: The convolutional neural network unit is responsible for receiving the reduced-dimensional data features and timestamp information of the new energy power generation prediction model, and building a convolutional neural network based on dynamic time series; the convolutional neural network design includes an input layer, a feature extraction convolution layer, a feature optimization selection pooling layer, and a long short-term memory network layer; The dependency unit is responsible for transmitting data features from the input layer to the convolution layer. The convolution layer learns different features through the convolution kernel, captures spatial features, generates feature atlases, and sends the feature atlases to the pooling layer. The pooling layer averages the feature atlases and corrects the average pooled feature atlases to correct the data features to the same timestamp. Capturing long-term dependencies in time series; When the vector value of the dependency relationship of the memory unit in the input gate of the LSTM network layer is close to 1, the element will be updated, and when it is close to 0, the element will remain unchanged; When the vector value of the dependency relationship in the memory unit of the forget gate is close to 1, the element will be retained; if it is close to 0, the element will be abandoned; When the vector value of the dependency relationship in the memory unit in the output gate is close to 1, the element will be output, and if the element is close to 0, the unit will not be output; wherein the element is a memory element in the memory unit, which is used to memorize the vector value of the dependency relationship; The feature fusion unit is responsible for mapping the long-term dependencies output by the LSTM network layer to the final prediction results, fusing the features extracted by the convolution layer and the LSTM network layer, and using the output of the convolution layer as the input of the LSTM network layer; the fused features are used to train the new energy power generation prediction model.

6. The new energy power generation prediction system according to claim 1, characterized in that: Prediction model building module, including: The prediction value submodule is responsible for predicting the power generation within a preset time period using the pre-trained new energy power generation prediction model, and forming a prediction value based on the learning results of historical prediction data; The identification submodule is responsible for comparing the predicted value of the renewable energy power generation prediction model with the actual power generation, calculating the prediction error, analyzing the calculated prediction error, and identifying the source of the error; The optimization module is responsible for adjusting the parameters of the renewable energy power generation prediction model according to the error analysis results, and optimizing the weight and structure of the renewable energy power generation prediction model.

7. The new energy power generation prediction system according to claim 6, characterized in that: Optimization submodules, including: Extract feature sheet, responsible for principal component analysis of error analysis results, extracting features strongly correlated with power generation, and inputting relevant features into adaptive weight device; The adjustment unit is responsible for the adaptive weight device to dynamically adjust the weight according to the features that have the greatest impact on the power generation among the relevant features; the adaptive weight device is integrated into the new energy power generation prediction model, and the historical data is used for training. During the training process, the adaptive weight device dynamically adjusts the weight according to the performance of the features; The analysis and comparison unit is responsible for comparing the predicted value with the actual power generation after the model training is completed, calculating the prediction error, and adjusting the parameters of the adaptive weight according to the error analysis results.

8. The new energy power generation prediction system according to claim 1, characterized in that: Test and evaluation module, including: The error average submodule is responsible for calculating the average error of the power generation of each time mode in the past n days based on the time period and power generation prediction statistical deviation: The prediction accuracy submodule is responsible for judging that the power generation prediction is accurate if the absolute difference between the power generation and the actual value is less than or equal to 2, otherwise it is inaccurate; The calculation and evaluation submodule is responsible for evaluating and verifying the power generation accuracy and root mean square error, average error and mean absolute error of the power generation forecast and the revision field.

9. The new energy power generation prediction system according to claim 8, characterized in that: In the calculation and evaluation submodule, the accuracy of power generation forecast is less than or equal to 2 percentage of power generation forecast error |Fi-Oi|, and the positive and negative errors in the average error calculation formula offset each other.

10. A method for predicting power generation from renewable energy, characterized in that: The following steps are involved: Collect weather forecast data for the target power station area, including temperature, humidity, wind direction and wind speed, and collect historical power generation data for the target power station, including power output at different timestamps; integrate environmental perception and dynamic data, deploy environmental perception units, monitor environmental changes around the power station in real time, and combine real-time environmental perception data, weather forecasts and historical power generation data to form a fused data set; Based on the fusion data set, pre-train the new energy power generation prediction model, and perform preset time period and power generation prediction, monitor the prediction error of the new energy power generation prediction model in real time, and adjust the parameters of the new energy power generation prediction model in real time according to the prediction error; The optimized new energy power generation prediction model is used to predict the new energy power generation, and the prediction accuracy and performance of the new energy power generation prediction model are evaluated by comparing it with the actual power generation prediction data.

Images