New energy power limit prediction method and device, electronic system and storage medium

Abstract

The invention provides a new energy power limit prediction method and device, an electronic system and a storage medium, and belongs to the technical field of power limit prediction.The method comprises the steps that multi-source heterogeneous data is processed, and a first feature vector is obtained; performing dimension reduction processing on the first feature vector to obtain a second feature vector; inputting the second feature vector into a first target model, and performing power limiting event judgment on the power system in the prediction time period; and in response to the fact that the output result of the first target model is power limitation, inputting the power feature data corresponding to the prediction time period into the second target model to obtain a power limitation prediction result. According to the new energy power limit prediction method and device, the electronic system and the storage medium provided by the invention, the accuracy of power limit prediction can be improved.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of power restriction prediction, and more specifically, relates to a new energy power restriction prediction method and device, an electronic system, and a storage medium. Background Art

[0002] In modern power systems, the integration of new energy sources increases the complexity and uncertainty of the power grid. Power curtailment events may lead to economic losses and energy waste. Therefore, accurate prediction of power curtailment events and their extent is of great significance for power grid scheduling and planning. Traditional methods have limitations in dealing with complex nonlinear relationships and cannot fully utilize multi-source data for accurate prediction. Summary of the invention

[0003] The purpose of the present disclosure is to provide a new energy power curtailment prediction method and device, an electronic system, and a storage medium to improve the accuracy of power curtailment prediction.

[0004] According to a first aspect of an embodiment of the present disclosure, a method for predicting power cuts on new energy sources is provided, comprising: Processing multi-source heterogeneous data to obtain the first eigenvector; Performing dimensionality reduction processing on the first eigenvector to obtain a second eigenvector; Inputting the second characteristic vector into the first target model to determine a power restriction event for the power system in the forecast time period; In response to the output result of the first target model being power restriction, the power characteristic data corresponding to the prediction time period is input into the second target model to obtain a power restriction prediction result.

[0005] A second aspect of the embodiments of the present disclosure provides a new energy power restriction prediction device, comprising: A data processing module, used for processing multi-source heterogeneous data to obtain a first eigenvector; A data dimension reduction module, used for performing dimension reduction processing on the first feature vector to obtain a second feature vector; A power restriction judgment module, used for inputting the second feature vector into the first target model to judge the power restriction event of the power system in the prediction time period; The power restriction prediction module is used to input the power characteristic data corresponding to the prediction time period into the second target model in response to the output result of the first target model being power restriction, so as to obtain the power restriction prediction result.

[0006] According to a third aspect of an embodiment of the present disclosure, an electronic system is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the above-mentioned new energy power restriction prediction method when executing the computer program.

[0007] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned new energy power restriction prediction method are implemented.

[0008] The beneficial effects of the new energy power curtailment prediction method and device, electronic system, and storage medium provided by the embodiments of the present disclosure are: the embodiments of the present disclosure extract the key first eigenvector through the fine processing of multi-source heterogeneous data, effectively integrate the power data from different sources and in different formats, and enhance the information density and availability of the data. Subsequently, the second eigenvector is obtained through dimensionality reduction processing, which further simplifies the data dimension, improves the calculation efficiency, and retains key information. On this basis, the first target model is used to judge the power curtailment event of the power system in the prediction time period, and the preliminary identification of the power curtailment risk is achieved. Once it is determined to be a power curtailment, the second target model is used to combine the power characteristic data corresponding to the prediction time period to make a more accurate prediction of the degree of power curtailment. The embodiments of the present disclosure make full use of the advantages of multi-source data, improve the accuracy and reliability of power curtailment prediction, provide strong decision-making support for power grid dispatching and planning, and help reduce the economic losses and energy waste caused by power curtailment events. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a flow chart of a new energy power restriction prediction method provided in an embodiment of the present disclosure; Figure 2 A schematic diagram of a flow chart of a new energy power curtailment prediction method provided in another embodiment of the present disclosure; Figure 3 A structural block diagram of a new energy power restriction prediction device provided in an embodiment of the present disclosure; Figure 4 A schematic block diagram of an electronic system provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0011] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.

[0012] In order to make the purpose, technical solutions and advantages of the present disclosure more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0013] Please refer to Figure 1 , Figure 1 A schematic flow chart of a new energy power restriction prediction method provided in an embodiment of the present disclosure, the method comprising: S101: Process multi-source heterogeneous data to obtain a first eigenvector.

[0014] In this embodiment, multi-source heterogeneous data refers to data from different data sources and having different data formats, structures, and characteristics.

[0015] In this embodiment, the multi-source heterogeneous data related to the power curtailment event includes, but is not limited to: Meteorological data: temperature, humidity, wind speed, rainfall, etc.; Meteorological data can affect electricity demand and renewable energy generation efficiency in many ways. Extreme weather can also have an impact on power facilities, and is an environmental factor that cannot be ignored in power rationing forecasts.

[0016] Economic change data: GDP, CPI, industrial production index, etc.; Economic change data reflects the level of economic development, prices and production activity, and is closely related to electricity demand. Economic growth or fluctuations can directly or indirectly affect the scale of electricity consumption.

[0017] Power load data: short-term load, tie line load, total load demand, etc.; Power load data reflects the electricity demand conditions at different levels of the power system, and is of key significance for predicting power rationing and power allocation.

[0018] New energy output data: photovoltaic load, wind power load, hydropower load, etc.; New energy output data has different volatility, and accurately grasping its changes will help to rationally arrange power dispatch and integrate energy resources.

[0019] Grid operation data: non-market output, total active power / actual power generation, etc.; Grid operation data reflects the power generation status of the grid under special circumstances and in actual operation, and can help ensure the stability of power supply and deal with possible imbalances in power supply and demand.

[0020] In this embodiment, the data comes from multiple different channels or systems, such as meteorological data, power load data, etc.; the data formats, data types and data structures are different, such as numerical data, text data, time series data, etc. In order to convert multi-source heterogeneous data into a unified format and form that can be used for subsequent analysis and prediction, and provide a basis for subsequent prediction, this embodiment processes multi-source heterogeneous data.

[0021] In this embodiment, processing multi-source heterogeneous data may include: Data cleaning: remove duplicate data, missing data, and abnormal data; fill in missing data, such as using mean, median, interpolation, etc.; smooth or correct abnormal data.

[0022] Data conversion: convert different types of data into a unified data type, such as converting text data into numeric data; standardize or normalize the data so that data with different characteristics have the same dimension and range.

[0023] Data merging: Merge data from different sources to form a complete data set.

[0024] Feature extraction: Extract features related to power system brownout events from the data.

[0025] After the above data processing steps, the data from different data sources are converted into a unified format and feature representation. These processed features are combined to form a feature vector, namely the first feature vector. The first feature vector integrates the key information in multi-source heterogeneous data and provides a basis for subsequent analysis and prediction.

[0026] S102: Perform dimensionality reduction processing on the first eigenvector to obtain a second eigenvector.

[0027] In this embodiment, the first feature vector may contain a large number of features, and there may be redundant information between these features. For example, in the first feature vector extracted from multi-source heterogeneous data such as meteorological data, economic data, and power load data, some features may be highly correlated. Dimensionality reduction processing can remove this redundant information, making the data more concise and reducing the amount of calculation and storage costs. At the same time, dimensionality reduction can highlight the main features in the data, which helps to improve the training efficiency and prediction accuracy of subsequent models.

[0028] In this embodiment, dimensionality reduction can be performed through principal component analysis (PCA), specifically: First, each feature in the first eigenvector is standardized so that they have the same dimension and range. Secondly, the covariance matrix between the features in the first eigenvector is calculated. The covariance matrix reflects the linear correlation between the features. Thirdly, the covariance matrix is ​​subjected to eigenvalue decomposition to solve the eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​represent the size of the variance in the direction of the eigenvector, while the eigenvectors indicate these directions. Thirdly, according to the size of the eigenvalues, the eigenvectors corresponding to the first k largest eigenvalues ​​are selected as the principal components. These principal components can retain the information in the original data to the greatest extent. Finally, the first eigenvector is projected onto the selected principal component to obtain the second eigenvector after dimensionality reduction.

[0029] The second eigenvector obtained after dimensionality reduction retains the main information in the first eigenvector, which is closely related to the power curtailment event.

[0030] S103: Input the second eigenvector into the first target model to determine a power limit event for the power system in the forecast time period.

[0031] In this embodiment, the first target model can be a trained machine learning model, such as logistic regression, support vector machine, random forest, gradient boosting tree, etc. The function of the first target model is to judge the power curtailment event of the power system in the prediction time period according to the input second feature vector. The first target model can establish a mapping relationship between the feature vector and the power curtailment event by learning the feature patterns related to the power curtailment event in the historical data, so as to predict the new input data.

[0032] In this embodiment, after the first target model calculates and analyzes the input second feature vector, it can output a judgment result on whether a power curtailment event will occur in the power system during the forecast time period. This result can be a binary classification result, such as "power curtailment" or "no power curtailment". If the first target model judges as "power curtailment", it means that according to the current characteristic conditions, the power system is more likely to experience a power curtailment event during the forecast time period; if it is judged as "no power curtailment", it means that the power system is unlikely to experience a power curtailment event during the forecast time period.

[0033] S104: In response to the output result of the first target model being power restriction, inputting the power characteristic data corresponding to the prediction time period into the second target model to obtain a power restriction prediction result.

[0034] In this embodiment, the power characteristic data corresponding to the forecast time period is the data characteristics closely related to the power restriction situation after screening and extraction. These power characteristic data may include short-term load, tie line load, non-market output, photovoltaic load, total load demand, hydropower load, wind power load, total active power and other characteristics.

[0035] These specific power characteristic data are input into the second target model. Compared with the second characteristic vector input into the first target model, the power characteristic data input here is more focused on the operation of the power system itself, and is a more targeted input when the first target model determines that there is power restriction.

[0036] In this embodiment, the second target model aims to provide a specific power limit prediction result based on the first target model's judgment of power limit, which may include the time range of power limit, the degree of power limit (such as the power size or proportion of power limit), and the areas that may be affected.

[0037] For different types of second objective models, the working process will be different. For example, if it is a model based on time series analysis, it may analyze the time series pattern of historical power data, combine it with the characteristics of current input power data, and predict the power shortage in the future.

[0038] In practical applications, the second target model needs to be trained. The training process includes collecting historical power-limiting data, extracting relevant features, building a training set, selecting appropriate algorithms (such as linear regression, support vector regression, random forest regression, neural network, etc.) for training, and adjusting model parameters to optimize prediction performance. When the first target model outputs the "power limit" result, the second target model will be activated. At this time, the power data features corresponding to the predicted time period are input into the model, and the model will predict the severity and duration of the power limit event based on its learned knowledge.

[0039] After the model outputs the power outage prediction results, these results need to be interpreted and understood. For example, if the prediction results show that the power outage will last for 4 hours and the load will be reduced by 200 megawatts, the power system operator can use this information to formulate a dispatch plan, adjust the power generation output, or take other measures to mitigate the impact of the power outage.

[0040] It can be concluded from the above that this embodiment extracts the key first eigenvector through the fine processing of multi-source heterogeneous data, effectively integrates the power data from different sources and in different formats, and enhances the information density and availability of the data. Subsequently, the second eigenvector is obtained through dimensionality reduction processing, which further simplifies the data dimension, improves the calculation efficiency, and retains the key information. On this basis, the first target model is used to judge the power limit event of the power system in the forecast time period, and the preliminary identification of the power limit risk is achieved. Once it is determined to be a power limit, the second target model is used to combine the power characteristic data corresponding to the forecast time period to make a more accurate prediction of the degree of power limit. This embodiment makes full use of the advantages of multi-source data, improves the accuracy and reliability of power limit prediction, provides strong decision-making support for power grid dispatching and planning, and helps to reduce the economic losses and energy waste caused by power limit events.

[0041] In one embodiment of the present disclosure, multi-source heterogeneous data is processed to obtain a first feature vector, including: Cleaning multi-source heterogeneous data to obtain multiple first feature data; Performing feature conversion on the plurality of first feature data to obtain a plurality of second feature data; Multiple second feature data are concatenated to obtain a first feature vector.

[0042] In this embodiment, in order to efficiently and in real time process the collected large-scale multi-source heterogeneous data, such as meteorological data, economic data, power load data and power grid operation data, this embodiment adopts the Spark big data parallel framework for concurrent data processing.

[0043] First, data cleaning is performed, including handling missing values, outliers, and noise.

[0044] For missing values, you can choose to delete data with a small proportion of missing values, or fill them using the mean, median, nearest neighbor, or interpolation method.

[0045] For filling missing values, this embodiment can use the average value to fill, and the missing value is:

[0046] in, Represents missing values, It is represented as the characteristic value of the jth data, and n is represented as the number of non-missing values ​​of the characteristic.

[0047] For outliers, this embodiment can use the box plot method (IQR) to identify and replace them with values ​​within the upper and lower quartiles or delete them directly. Outliers are defined as values ​​less than or greater than The value of .

[0048] Where IQR is the interquartile range, Q1 and Q3 are the first quartile and the third quartile respectively.

[0049] After cleaning the multi-source heterogeneous data, multiple first feature data are obtained.

[0050] Next, feature conversion is performed on the plurality of first feature data, and features of different data sources are uniformly converted into a mathematical vector format.

[0051] In this embodiment, RobustScaler can be used to perform feature standardization to reduce the impact of outliers. The formula is:

[0052] in, It represents the standardized eigenvalue, x represents the original eigenvalue of multi-source heterogeneous data, and median(x) represents the middle value (median) of the eigenvalue x. Expressed as the interquartile range of the eigenvalue x.

[0053] Finally, multiple second feature data are concatenated to obtain a first feature vector.

[0054] The feature vector of data source A is V a , the feature vector of data source B is V b and the characteristic vector of data source m is V m , then the concatenated feature vector is T={V a , V b , …, V m}, forming an N-dimensional eigenvector T (i.e., the first eigenvector).

[0055] The first eigenvector data obtained can be stored in a distributed storage system (HDFS) for subsequent machine learning model training or analysis. This method can efficiently process large-scale multi-source heterogeneous data in parallel, ensure data quality, and provide a reliable foundation for subsequent model training and prediction.

[0056] It can be concluded from the above that this embodiment adopts the Spark big data parallel framework, which can efficiently process large-scale meteorological, economic, power load and power grid operation data in real time. Through data cleaning, missing values, outliers and noise are properly handled, such as filling missing values ​​with average values ​​and box plot method to handle outliers, to ensure data quality. Feature conversion unifies the features of different data sources into mathematical vector format, and uses RobustScaler to reduce the impact of outliers. Finally, it is spliced ​​into the first feature vector and stored in a distributed storage system. This process efficiently processes data in parallel, ensures data reliability, and lays a solid foundation for subsequent machine learning model training and prediction.

[0057] In one embodiment of the present disclosure, performing dimensionality reduction processing on the first eigenvector to obtain the second eigenvector includes: The first feature vector is input into the encoder for compression to obtain the second feature vector.

[0058] In this embodiment, since the features of different data sources are usually in the form of high-dimensional feature vectors, too high a dimension may have an adverse effect on model training. To solve this problem, this embodiment uses a deep autoencoder to reduce the dimension of features. The deep autoencoder can extract implicit relationships and patterns in the data, reduce the number of features while retaining important information. Through the deep autoencoder neural network, the original N-dimensional feature vector T can be reduced in dimension to obtain a reduced-dimensional vector with a dimension of M, where . This process helps to discover implicit feature relationships in the data.

[0059] A deep autoencoder is a special type of neural network consisting of multiple layers of neural networks. The encoder is composed of multiple layers of neural networks and compresses high-dimensional features into a low-dimensional latent space by learning the nonlinear relationship of the data.

[0060] The encoder compresses the input data into a low-dimensional latent space (Latent Space), maps the original N-dimensional feature vector T to a low-dimensional feature vector Z with a dimension of M, and maps the input feature vector T to the latent space representation Z.

[0061]

[0062] in is the encoder weight matrix, is the encoder bias vector, It is represented as the activation function.

[0063] It can be concluded from the above that this embodiment uses a deep autoencoder to reduce the dimensionality of the first feature vector. Since the features of different data sources often exist in the form of high-dimensional vectors, too high a dimension will be detrimental to model training, and the deep autoencoder can effectively solve this problem. It can extract implicit relationships and patterns in the data, retain important information while reducing the number of features, and reduce the original N-dimensional feature vector to an M-dimensional vector. As a special network composed of multi-layer neural networks, its encoder can learn nonlinear relationships between data and compress high-dimensional features into low-dimensional potential space, which helps to discover implicit feature relationships in the data and lays the foundation for subsequent model training optimization.

[0064] In one embodiment of the present disclosure, it also includes: Inputting the second feature vector into the decoder to reconstruct the second feature vector to obtain a third feature vector; The encoder is updated based on the second eigenvector and the third eigenvector.

[0065] In this embodiment, after the encoder compresses the first feature vector into the second feature vector, the purpose of introducing the decoder is to try to recover a representation as close to the original data as possible from the low-dimensional second feature vector. The decoder is also composed of a multi-layer neural network, which receives the second feature vector as input, and gradually maps the low-dimensional second feature vector back to the high-dimensional space through the learned parameters and network structure to generate the third feature vector.

[0066] In this embodiment, the representation in the latent space is reconstructed back to the original data space through the decoder, and the dimension reduction vector in the latent space is Reconstruct back to the original data space, and get .

[0067]

[0068] in, is represented as the third eigenvector, Denoted as the weight matrix of the decoder, is represented as the encoder bias vector, is represented as the activation function, and Z is represented as the second eigenvector (reduced dimensionality vector ).

[0069] The autoencoder minimizes the relationship between the input data T and its reconstruction The error between them is used to optimize its parameters. The error loss function is the mean squared error (MSE):

[0070] This process aims to learn an effective low-dimensional representation of the data while maintaining the accuracy of the reconstructed data. The autoencoder optimizes its parameters by minimizing the error between the input data and its reconstruction, thereby reducing the dimensionality of the data.

[0071] From the above, we can conclude that the decoder can recover a representation close to the original data from the low-dimensional second eigenvector, generate the third eigenvector by mapping it back to the high-dimensional space, and continuously optimize the encoder parameters using the mean square error loss function. This helps to learn an effective low-dimensional representation of the data, maintain the accuracy of the reconstructed data while reducing the dimensionality, provide better features for subsequent power-limiting predictions, improve model performance, reduce computational and storage costs, better cope with the challenges brought by high-dimensional data, and lay the foundation for accurately predicting power-limiting events.

[0072] In one embodiment of the present disclosure, the second characteristic vector is input into the first target model, and a power restriction event is determined for the power system in the forecast time period, including: Inputting the second eigenvectors into multiple binary classification models respectively to obtain the probability of occurrence of power curtailment events corresponding to the multiple binary classification models; The power outage event judgment of the power system in the forecast time period is carried out based on probability.

[0073] In this embodiment, different binary classification model algorithms are first selected, such as logistic regression, support vector machine, random forest, gradient boosting tree, etc., and the selected multiple binary classification models are trained and verified to select the model with the best performance. The dimension reduction vector Z after dimensionality reduction processing is input into the trained multiple binary classification models to predict power curtailment events, and the prediction results of different models are integrated to improve the prediction accuracy of the model, and each model will output the probability of a power curtailment event.

[0074] In one embodiment of the present disclosure, a power curtailment event is determined for a power system in a predicted time period based on probability, including: Based on the first formula, a power restriction event is determined for the power system in the forecast time period; The first formula is:

[0075] Among them, P represents the judgment result of the power system power outage event in the forecast time period, N represents the number of binary classification models, and P i It represents the probability of power curtailment in the power system predicted by the i-th binary classification model.

[0076] In this embodiment, firstly, for each binary classification model, the probability P of a power-cutting event occurring is obtained by analyzing and calculating the second feature vector input. i ,Each binary classification model will output a probability value reflecting the degree of prediction of the ,occurrence of power restriction events by the model.

[0077] Then sum the probability values ​​of all N binary classification models, that is, This summation process integrates the prediction results of each model. In practical applications, different numbers of binary classification models can be selected according to specific circumstances to improve the accuracy and stability of the prediction.

[0078] Finally, the sum is divided by the number of models N to obtain the final judgment result P. This averaging process can reduce the uncertainty and error of a single model and improve the reliability of the overall prediction.

[0079] From the above, it can be concluded that this embodiment can integrate the advantages of different models by analyzing and calculating the probability of the second eigenvector of the input through multiple binary classification models, and then summing and averaging. It can adapt to complex power systems and fully consider the impact of various factors on power outages. Reduce the uncertainty and error of a single model and improve the overall prediction reliability and stability. Select different quantity models according to specific circumstances to improve prediction accuracy. Provide a more accurate basis for the planning and scheduling of power systems and reduce the economic losses and adverse effects caused by power outages.

[0080] In one embodiment of the present disclosure, a power curtailment event is determined for a power system in a predicted time period based on probability, including: In response to the probability being greater than the first threshold, it is determined to be a power-cutting event.

[0081] In this embodiment, after multiple binary classification models process the input second feature vector, the probability P of a power outage occurring in the power system predicted by each binary classification model is obtained. i Then the comprehensive judgment result probability P is calculated according to the first formula.

[0082] If the probability of a power-rationing event occurring output by multiple binary classification models is greater than a first threshold (such as 0.5, the threshold can be adjusted according to actual conditions), it is judged as a power-rationing event. The first threshold here is a pre-set value, which represents the standard for judging whether a power-rationing event occurs.

[0083]

[0084] It can be concluded from the above that this embodiment uses the prediction probability of multiple models for comprehensive calculation. When this comprehensive probability reaches a certain level, it indicates that the possibility of power restriction is high, thereby making a judgment on the power restriction event. This can avoid possible misjudgment of a single model, improve the accuracy and reliability of the judgment, and provide timely and effective decision-making basis for the operation and management of the power system.

[0085] In one embodiment of the present disclosure, in response to the output result of the first target model being power restriction, inputting power characteristic data corresponding to the prediction time period into the second target model to obtain a power restriction prediction result includes: Select an appropriate regression model algorithm, such as a random forest regression model. Use the probability of power outages predicted by each binary classification model as a feature, and train a random forest regression model based on features such as short-term load, tie line load, non-market output, photovoltaic load, total load demand, hydropower load, wind power load, and total active power. Train and validate multiple random forest regression models and select the model with the best performance.

[0086] In response to the first target model outputting a power curtailment event, the probability of the power curtailment event and the power characteristic data are input into a random forest regression model to obtain a power curtailment prediction result.

[0087] In this embodiment, when the output result of the first target model is power curtailment, the probability of power curtailment and the above-mentioned power characteristic data are input into the trained random forest regression model. The model performs regression prediction based on the relationship between the learned features and the power curtailment results, and outputs the power curtailment prediction result. This result may include specific information such as the degree of power curtailment (such as the size of the power curtailment) and the duration of the power curtailment, providing decision-making basis for the operators and relevant departments of the power system so that they can take corresponding measures to deal with the power curtailment situation.

[0088] From the above, it can be concluded that when the first target model determines that there is power curtailment, the random forest regression model is used to predict power curtailment. The power curtailment probability predicted by the binary classification model is combined with a variety of power feature data training models to ensure the richness and comprehensiveness of the model input information. By training and verifying multiple random forest regression models, the best performance is selected to improve the accuracy and reliability of the prediction. The final output power curtailment prediction results include information such as the degree and duration of power curtailment, which provides a clear decision-making basis for power system operators and relevant departments, facilitating timely and effective measures to deal with power curtailment.

[0089] Please refer to Figure 2 , Figure 2 A flowchart of a new energy power curtailment prediction method provided by another embodiment of the present disclosure.

[0090] First, data collection is carried out to collect multi-source heterogeneous data, including meteorological data, economic data, power load data, new energy output data, and power grid operation data. Then, these multi-source heterogeneous data are processed, including operations such as cleaning, conversion, merging, and feature extraction, to obtain the first feature vector. Then feature selection is performed to select key features for subsequent analysis. The first feature vector is then subjected to dimensionality reduction processing to obtain the second feature vector. After that, binary classification model training is carried out, and multiple binary classification model algorithms such as logistic regression and support vector machine are used for training and verification. By integrating the prediction results of multiple binary classification models, it is determined whether the power system in the predicted time period is power-restricted. If it is determined to be power-restricted, a suitable regression model, such as a random forest regression model, is selected, and the power-restriction probability predicted by the binary classification model and a variety of power feature data are used to train the regression model, and then the power restriction is predicted and the power restriction prediction result is output; if it is determined to be unrestricted, the output power restriction is 0. The entire process provides a comprehensive and accurate method for predicting power restriction in the power system through multi-step data processing and model application, which helps power system operators and relevant departments to prepare countermeasures in advance and reduce the adverse effects of power restriction.

[0091] Corresponding to the new energy power restriction prediction method in the above embodiment, Figure 3 This is a structural block diagram of a new energy power restriction prediction device provided by an embodiment of the present disclosure. For the sake of convenience, only the parts related to the embodiment of the present disclosure are shown. Figure 3 The new energy power restriction prediction device 20 includes: a data processing module 21 , a data dimension reduction module 22 , a power restriction judgment module 23 , and a power restriction prediction module 24 .

[0092] The data processing module 21 is used to process multi-source heterogeneous data to obtain a first feature vector; A data dimension reduction module 22, used for performing dimension reduction processing on the first eigenvector to obtain a second eigenvector; A power restriction judgment module 23 is used to input the second characteristic vector into the first target model to judge the power restriction event of the power system in the prediction time period; The power restriction prediction module 24 is used to input the power characteristic data corresponding to the prediction time period into the second target model in response to the output result of the first target model being power restriction, so as to obtain a power restriction prediction result.

[0093] In one embodiment of the present disclosure, the data processing module 21 is specifically used for: Cleaning multi-source heterogeneous data to obtain multiple first feature data; Performing feature conversion on the plurality of first feature data to obtain a plurality of second feature data; Multiple second feature data are concatenated to obtain a first feature vector.

[0094] In one embodiment of the present disclosure, the data dimension reduction module 22 is specifically used to: The first feature vector is input into the encoder for compression to obtain the second feature vector.

[0095] In one embodiment of the present disclosure, the data dimension reduction module 22 is further used to: Inputting the second feature vector into the decoder to reconstruct the second feature vector to obtain a third feature vector; The encoder is updated based on the second eigenvector and the third eigenvector.

[0096] In one embodiment of the present disclosure, the power limit determination module 23 is specifically used to: Inputting the second eigenvectors into multiple binary classification models respectively to obtain the probability of occurrence of power curtailment events corresponding to the multiple binary classification models; The power outage event judgment of the power system in the forecast time period is carried out based on probability.

[0097] In one embodiment of the present disclosure, the power limit determination module 23 is further configured to: Based on the first formula, a power restriction event is determined for the power system in the forecast time period; The first formula is:

[0098] Among them, P represents the judgment result of the power system power outage event in the forecast time period, N represents the number of binary classification models, and P i It represents the probability of power curtailment in the power system predicted by the i-th binary classification model.

[0099] In one embodiment of the present disclosure, the power limit determination module 23 is further configured to: In response to the probability being greater than the first threshold, it is determined to be a power-cutting event.

[0100] See also Figure 4 , Figure 4 Schematic block diagram of an electronic system provided by an embodiment of the present disclosure. Figure 4 The electronic system 300 in the embodiment shown may include: one or more processors 301, one or more input devices 302, one or more output devices 303 and one or more memories 304. The processors 301, input devices 302, output devices 303 and memories 304 communicate with each other via a communication bus 305. The memory 304 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. The processor 301 is configured to call the program instructions to execute the functions of the modules in the above-mentioned device embodiments, such as Figure 3 The functions of modules 21 to 24 are shown.

[0101] It should be understood that in the embodiment of the present disclosure, the processor 301 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0102] The input device 302 may include a touch panel, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and the output device 303 may include a display (LCD, etc.), a speaker, etc.

[0103] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A portion of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.

[0104] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present disclosure can execute the implementation methods described in the first and second embodiments of the new energy power restriction prediction method provided in the embodiments of the present disclosure, and can also execute the implementation methods of the electronic system described in the embodiments of the present disclosure, which will not be repeated here.

[0105] In another embodiment of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by the processor, all or part of the processes in the above-mentioned embodiment method are implemented, and the computer program can also be completed by instructing the relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0106] The computer-readable storage medium may be an internal storage unit of the electronic system of any of the aforementioned embodiments, such as a hard disk or memory of the electronic system. The computer-readable storage medium may also be an external storage device of the electronic system, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic system. Furthermore, the computer-readable storage medium may also include both an internal storage unit of the electronic system and an external storage device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic system. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0107] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0108] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the electronic systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0109] In the several embodiments provided in the present application, it should be understood that the disclosed electronic systems 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. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces or units, or it can be an electrical, mechanical or other form of connection.

[0110] 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 embodiments of the present disclosure.

[0111] In addition, each functional unit in each embodiment of the present disclosure 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.

[0112] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present disclosure, and these modifications or replacements should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A new energy power curtailment prediction method, characterized in that: include: Processing multi-source heterogeneous data to obtain the first eigenvector; Performing dimensionality reduction processing on the first eigenvector to obtain a second eigenvector; Inputting the second characteristic vector into the first target model to determine a power restriction event for the power system in the forecast time period; In response to the first target model outputting a power restriction result, the power characteristic data corresponding to the prediction time period is input into the second target model to obtain a power restriction prediction result.

2. The new energy power restriction prediction method according to claim 1, characterized in that: The processing of multi-source heterogeneous data to obtain a first feature vector includes: Cleaning the multi-source heterogeneous data to obtain a plurality of first feature data; Performing feature conversion on the plurality of first feature data to obtain a plurality of second feature data; The plurality of second feature data are concatenated to obtain a first feature vector.

3. The new energy power restriction prediction method according to claim 1, characterized in that: The performing dimensionality reduction processing on the first feature vector to obtain a second feature vector includes: The first feature vector is input into an encoder for compression to obtain the second feature vector.

4. The new energy power restriction prediction method according to claim 3, characterized in that: Also includes: Inputting the second feature vector into a decoder to reconstruct the second feature vector to obtain a third feature vector; The encoder is updated based on the second eigenvector and the third eigenvector.

5. The new energy power restriction prediction method according to claim 1, characterized in that: The step of inputting the second characteristic vector into the first target model and determining a power restriction event for the power system in the forecast time period includes: Inputting the second feature vector into a plurality of binary classification models respectively to obtain the probability of occurrence of power curtailment events corresponding to the plurality of binary classification models; A power outage event is determined for the power system in the predicted time period based on the probability.

6. The new energy power restriction prediction method according to claim 5, characterized in that: The determining of a power restriction event for the power system in the predicted time period based on the probability includes: Based on the first formula, a power restriction event is determined for the power system in the forecast time period; The first formula is: Among them, P represents the judgment result of the power system power outage event in the forecast time period, N represents the number of binary classification models, and P i It represents the probability of power curtailment in the power system predicted by the i-th binary classification model.

7. The new energy power restriction prediction method according to claim 5, characterized in that: The determining of a power restriction event for the power system in the predicted time period based on the probability includes: In response to the probability being greater than a first threshold, it is determined to be a power-cutting event.

8. A new energy power restriction prediction device, characterized in that: include: A data processing module, used for processing multi-source heterogeneous data to obtain a first eigenvector; A data dimension reduction module, used for performing dimension reduction processing on the first feature vector to obtain a second feature vector; A power restriction judgment module, used for inputting the second feature vector into the first target model to judge the power restriction event of the power system in the prediction time period; The power restriction prediction module is used to input the power characteristic data corresponding to the prediction time period into the second target model in response to the output result of the first target model being power restriction, so as to obtain the power restriction prediction result.

9. An electronic system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images