New energy side reactive power prediction method and device

Abstract

The invention relates to the technical field of power prediction, in particular to a new energy side reactive power prediction method and device, and the method comprises the steps: collecting the real-time meteorological observation data of the position of a target new energy station and the real-time operation condition data of power generation equipment, and carrying out the data processing, so as to obtain standard data; extracting a plurality of high-dimensional key features influencing reactive power from the standard data, performing dimension reduction processing on the plurality of high-dimensional key features to obtain a plurality of low-dimensional key features, selecting the low-dimensional key features meeting a preset requirement to generate a low-dimensional key feature set, and inputting the low-dimensional key feature set into a pre-established reactive power prediction model, and outputting a reactive power prediction result of the target new energy station. According to the method, key characteristics such as meteorological conditions and operation conditions can be used as input of the reactive power prediction model, the method adapts to fluctuation and intermittency characteristics of new energy power generation, the prediction precision of the reactive power is effectively improved, and a real-time and accurate prediction result is provided for a power grid dispatching decision.

Description

Technical Field

[0001] This application relates to the technical field of power prediction, and particularly to a reactive power prediction method and device on the new energy side. Background Art

[0002] In related technologies, traditional reactive power prediction methods include time series method, artificial neural network method, support vector machine method, and grey prediction method, etc. The time series method establishes a time series model to predict future reactive power; the artificial neural network method trains and learns the non-linear relationship between reactive power and related factors through a large amount of data; the support vector machine method transforms the reactive power prediction problem into a non-linear regression problem in a high-dimensional feature space; the grey prediction method establishes a grey model to predict reactive power by generating and processing the original data.

[0003] However, the traditional reactive power prediction methods in related technologies all have their own limitations. When facing the non-linear and unstable characteristics of reactive power supply due to the discontinuity and volatility of new energy power generation, it is difficult to accurately capture the characteristics of reactive power, and they also ignore the influence of external factors such as meteorological conditions and operating conditions on reactive power, making it difficult to effectively guarantee the accuracy of the prediction results, which urgently needs to be solved. Summary of the Invention

[0004] This application provides a reactive power prediction method and device on the new energy side to solve the problems that the traditional reactive power prediction methods in related technologies all have their own limitations, and when facing the non-linear and unstable characteristics of reactive power supply due to the discontinuity and volatility of new energy power generation, it is difficult to accurately capture the characteristics of reactive power, and they also ignore the influence of external factors such as meteorological conditions and operating conditions on reactive power, making it difficult to effectively guarantee the accuracy of the prediction results, etc.

[0005] The first aspect of the embodiments of this application provides a reactive power prediction method on the new energy side, including the following steps: collecting real-time meteorological observation data of the location where the target new energy power station is located and real-time operating conditions data of the power generation equipment, and performing data processing on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment to obtain standard data; extracting multiple high-dimensional key features that affect reactive power from the standard data, and performing dimensionality reduction processing on the multiple high-dimensional key features to obtain multiple low-dimensional key features, so as to select low-dimensional key features that meet preset requirements from the multiple low-dimensional key features to generate a low-dimensional key feature set; inputting the low-dimensional key feature set into a pre-established reactive power prediction model to output the reactive power prediction result of the target new energy power station.

[0006] Optionally, in an embodiment of the present application, the data processing of the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment includes: performing data cleaning on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment to obtain real-time meteorological observation data and real-time operating conditions data of the power generation equipment with abnormal data removed and / or missing data supplemented; performing standardization processing and normalization processing on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment with abnormal data removed and / or missing data supplemented to obtain the standard data.

[0007] Optionally, in an embodiment of the present application, the selecting of the low-dimensional key features that meet the preset requirements from the multiple low-dimensional key features to generate a low-dimensional key feature set includes: constructing a covariance matrix of the multiple low-dimensional key features; decomposing the eigenvalues of the covariance matrix to select the low-dimensional key features that meet the preset requirements according to the eigenvalues.

[0008] Optionally, in an embodiment of the present application, before inputting the low-dimensional key feature set into the pre-established reactive power prediction model, it further includes: obtaining historical reactive power data, historical meteorological data, and historical operating conditions data of multiple new energy power stations to extract historical data features of the multiple new energy power stations from the historical reactive power data, historical meteorological data, and historical operating conditions data; training a pre-constructed hybrid neural network model based on the historical data features to obtain the reactive power prediction model.

[0009] Optionally, in an embodiment of the present application, after outputting the reactive power prediction result of the target new energy power station, it further includes: updating the reactive power prediction model based on an incremental learning strategy and an online learning mechanism using the real-time meteorological data, the real-time operating conditions data of the power generation equipment, and the reactive power prediction result of the target new energy power station, so as to generate a new reactive power prediction result of the target new energy power station using the updated reactive power prediction model.

[0010] In the second aspect of the embodiments of the present application, a reactive power prediction device on the new energy side is provided, including: a data processing module, configured to collect real-time meteorological observation data of the location where the target new energy power station is located and real-time operating conditions data of the power generation equipment, and perform data processing on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment to obtain standard data; a feature extraction module, configured to extract multiple high-dimensional key features affecting reactive power from the standard data, and perform dimensionality reduction processing on the multiple high-dimensional key features to obtain multiple low-dimensional key features, so as to select low-dimensional key features that meet preset requirements from the multiple low-dimensional key features to generate a low-dimensional key feature set; a power prediction module, configured to input the low-dimensional key feature set into a pre-established reactive power prediction model to output a reactive power prediction result of the target new energy power station.

[0011] Optionally, in an embodiment of the present application, the data processing module includes: a data cleaning unit, configured to clean the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment to obtain real-time meteorological observation data and real-time operating conditions data of the power generation equipment with abnormal data removed and / or missing data supplemented; a data standardization unit, configured to perform standardization processing and normalization processing on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment with abnormal data removed and / or missing data supplemented to obtain the standard data.

[0012] Optionally, in an embodiment of the present application, the feature extraction module includes: a construction unit, configured to construct a covariance matrix of the multiple low-dimensional key features; a selection unit, configured to decompose the eigenvalues of the covariance matrix to select the low-dimensional key features that meet the preset requirements according to the eigenvalues.

[0013] Optionally, in an embodiment of the present application, it further includes: an acquisition module, configured to obtain historical reactive power data, historical meteorological data, and historical operating conditions data of multiple new energy power stations before inputting the low-dimensional key feature set into the pre-established reactive power prediction model, so as to extract historical data features of the multiple new energy power stations from the historical reactive power data, historical meteorological data, and historical operating conditions data; a training module, configured to train a pre-constructed hybrid neural network model based on the historical data features to obtain the reactive power prediction model.

[0014] Optionally, in an embodiment of the present application, it further includes: an update module, configured to, after outputting the reactive power prediction result of the target new energy power station, update the reactive power prediction model based on an incremental learning strategy and an online learning mechanism, using the real-time meteorological data, the real-time operating conditions data of the power generation equipment, and the reactive power prediction result of the target new energy power station, so as to generate a new reactive power prediction result of the target new energy power station using the updated reactive power prediction model.

[0015] An embodiment of the third aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the program to implement the reactive power prediction method on the new energy side as described in the above embodiment.

[0016] An embodiment of the fourth aspect of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the program is executed by a processor, it implements the reactive power prediction method on the new energy side as described above.

[0017] An embodiment of the fifth aspect of the present application provides a computer program product, including a computer program, where when the computer program is executed, it is used to implement the reactive power prediction method on the new energy side as described above.

[0018] Embodiments of the present application can obtain a set of key low-latitude features affecting reactive power by collecting real-time meteorological observation data and real-time operating conditions data of power generation equipment at a target new energy power station, and then input it into a reactive power prediction model to output a prediction result of reactive power. Thus, it realizes the preprocessing and feature extraction of multi-source heterogeneous data of the target new energy power station through data fusion technology. When facing reactive power that is non-linear and unstable due to the discontinuity and volatility of new energy power generation, it can also accurately capture the characteristics of reactive power, providing precise data support for the input data of the reactive power prediction model, effectively reducing the time and cost of the model for processing data. Moreover, the anomaly detection algorithm introduced in the reactive power prediction model of the present application can identify abnormal data under extreme weather conditions, improving the robustness of the model. The incremental learning measurement and online learning mechanism adopted can dynamically update model parameters according to real-time data, and can adapt to the volatility and intermittency characteristics of new energy power generation, thereby improving the prediction accuracy of reactive power, providing real-time and accurate prediction results for power grid dispatching decisions, effectively coping with the challenges brought by new energy grid connection, and enhancing the safety and economy of power grid operation. Thus, it solves the problems that traditional reactive power prediction methods in related technologies all have their own limitations, and when facing the characteristics that the supply of reactive power is non-linear and unstable due to the discontinuity and volatility of new energy power generation, it is difficult to accurately capture the characteristics of reactive power, and ignores the influence of external factors such as meteorological conditions and operating conditions on reactive power, and it is difficult to effectively guarantee the accuracy of prediction results, etc.

[0019] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:

[0021] Figure 1 is a schematic diagram of the framework of a new energy side reactive power prediction system based on artificial intelligence according to an embodiment of the present application;

[0022] Figure 2 is a flowchart of a new energy side reactive power prediction method according to an embodiment of the present application;

[0023] Figure 3 is a schematic structural diagram of a new energy side reactive power prediction device according to an embodiment of the present application;

[0024] Figure 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

[0025] Reference Signs:

[0026] 10 - Reactive power prediction device on the new - energy side: 100 - data processing module, 200 - feature extraction module, and 300 - power prediction module; 401 - memory, 402 - processor, and 403 - communication interface. Specific embodiments

[0027] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present application, and should not be construed as a limitation to the present application.

[0028] The method and device for reactive power prediction on the new - energy side according to the embodiments of the present application will be described below with reference to the drawings. In view of the limitations of the traditional reactive power prediction methods in the related art mentioned in the above - mentioned background technology, and when facing the non - linear and unstable characteristics of reactive power supply due to the discontinuity and volatility of new - energy power generation, it is difficult to accurately capture the characteristics of reactive power, and the influence of external factors such as meteorological conditions and operating conditions on reactive power is ignored, making it difficult to effectively guarantee the accuracy of the prediction results. The present application provides a method for reactive power prediction on the new - energy side. In this method, a low - dimensional key feature set affecting reactive power can be obtained by collecting real - time meteorological observation data and real - time operating conditions data of power generation equipment of the target new - energy power station, and then it is input into the reactive power prediction model to output the prediction result of reactive power. Thus, through data fusion technology, pre - processing and feature extraction of multi - source heterogeneous data of the target new - energy power station are realized. When facing the non - linear and unstable reactive power due to the discontinuity and volatility of new - energy power generation, the characteristics of reactive power can also be accurately captured, providing accurate data support for the input data of the reactive power prediction model, effectively reducing the time and cost of the model for processing data. Moreover, the anomaly detection algorithm introduced in the reactive power prediction model of the present application can identify abnormal data under extreme weather conditions, improving the robustness of the model. The incremental learning measurement and online learning mechanism adopted can dynamically update the model parameters according to real - time data, and can adapt to the volatility and intermittency characteristics of new - energy power generation, thereby improving the prediction accuracy of reactive power, providing real - time and accurate prediction results for power grid dispatching decisions, effectively coping with the challenges brought by new - energy grid connection, and enhancing the safety and economy of power grid operation. Thus, the problems in the related art that the traditional reactive power prediction methods all have their own limitations, and when facing the non - linear and unstable characteristics of reactive power supply due to the discontinuity and volatility of new - energy power generation, it is difficult to accurately capture the characteristics of reactive power, and the influence of external factors such as meteorological conditions and operating conditions on reactive power is ignored, making it difficult to effectively guarantee the accuracy of the prediction results, etc. are solved.

[0029] Before explaining the new - energy - side reactive - power prediction method in the embodiments of the present application, first explain the artificial - intelligence - based new - energy - side reactive - power prediction system architecture that uses the new - energy - side reactive - power prediction method in the embodiments of the present application.

[0030] Figure 1 It is a schematic diagram of the framework of the artificial - intelligence - based new - energy - side reactive - power prediction system for an embodiment of the present application. As Figure 1 shown, the artificial - intelligence - based new - energy - side reactive - power prediction system in the embodiments of the present application mainly includes four parts: 1 - data collection module, 2 - data processing module, 3 - model construction module, and 4 - power prediction module.

[0031] Among them, the 1 - data collection and processing module is used to collect real - time meteorological observation data and the operating conditions data of power generation equipment.

[0032] The 2 - data processing module processes and extracts features from the real - time meteorological observation data and the operating conditions data of power generation equipment to obtain a set of low - dimensional key features that affect reactive power.

[0033] The 3 - model construction module is used to collect historical data and construct a reactive - power prediction model based on the historical data.

[0034] The 4 - power prediction module completes the prediction of the new - energy - side reactive power based on the reactive - power prediction model and the set of low - dimensional key features.

[0035] With the development of artificial intelligence, the deep - learning function of artificial intelligence has made breakthroughs in many fields. However, how to apply the deep learning of artificial intelligence to new - energy - side reactive - power prediction still faces many technical challenges. In order to better apply the deep learning of artificial intelligence to new - energy - side reactive - power prediction, the embodiments of the present application propose a new - energy - side reactive - power prediction method, which can effectively utilize the deep - learning function of artificial intelligence.

[0036] Specifically, Figure 2 It is a flowchart of a new - energy - side reactive - power prediction method provided by the embodiments of the present application.

[0037] As Figure 2 shown, the new - energy - side reactive - power prediction method includes the following steps:

[0038] In step S201, collect the real - time meteorological observation data of the location where the target new - energy power station is located and the real - time operating conditions data of the power generation equipment, and perform data processing on the real - time meteorological observation data and the real - time operating conditions data of the power generation equipment to obtain standard data.

[0039] It can be understood that the target new energy power station can be understood as the new energy side in the power system that needs to perform reactive power prediction here.

[0040] In some embodiments, when predicting the reactive power of the new energy power station, in order to ensure the accuracy of the prediction results, the present application can collect data from multiple aspects of the target new energy power station as data support, including but not limited to real-time meteorological observation data of the location where the target new energy power station is located and real-time operating conditions data of the power generation equipment, etc.

[0041] For example, the meteorological information of the target new energy power station includes but not limited to real-time wind speed, real-time wind direction, real-time temperature, real-time light intensity and other aspects of meteorological information at the location where the target new energy power station is located. The real-time operating conditions data of the power generation equipment includes but not limited to the rotational speed of the wind turbine and the angle of the photovoltaic panel, etc.

[0042] Furthermore, considering that there may be certain invalid data or abnormal data in the directly collected data, etc., the embodiments of the present application can also perform data processing on these real-time meteorological observation data and real-time operating conditions data of the power generation equipment to obtain available effective standard data.

[0043] The embodiments of the present application can collect the real-time meteorological observation data of the target new energy power station and the real-time operating conditions data of the power generation equipment as data support, consider various factors affecting reactive power, and through certain data processing on these data, the present application can effectively predict the reactive power of the target new energy power station even in the abnormal working conditions under extreme weather conditions.

[0044] Optionally, in an embodiment of the present application, performing data processing on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment includes: performing data cleaning on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment to obtain real-time meteorological observation data and real-time operating conditions data of the power generation equipment with abnormal data removed and / or missing data supplemented; performing standardization processing and normalization processing on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment with abnormal data removed and / or missing data supplemented to obtain standard data.

[0045] In some embodiments, when the present application performs data processing on the real-time meteorological observation data and the real-time operating conditions data of the power generation equipment, it mainly includes but not limited to two parts: data cleaning and standardization processing and normalization processing.

[0046] Considering that there may be invalid and abnormal data in the real-time meteorological observation data and the real-time operating conditions data of power generation equipment, the embodiments of the present application can first perform data cleaning on these data, so as to remove abnormal data in the real-time meteorological observation data and the operating conditions data of power generation equipment while filling in missing data, and obtain the cleaned valid data.

[0047] For example, the present application can use the box plot method to detect abnormal data in the real-time meteorological observation data and the operating conditions data of power generation equipment, and regard the data exceeding 1.5 interquartile ranges above and below the upper and lower quartiles as outliers and remove them. That is, after sorting a set of data from small to large, it is divided into four equal parts of values. The dividing points of these four equal parts are the 25th percentile (lower quartile, Q1), the 50th percentile (median, Q2), and the 75th percentile (upper quartile, Q3) respectively. Quartiles can describe the distribution and shape of data, especially the degree of dispersion of data. Then, the data points in the real-time meteorological observation data and the operating conditions data of power generation equipment whose distance from the upper quartile (Q3) or the lower quartile (Q1) exceeds a certain 1.5-fold range are removed. Among them, this range is measured by multiples of the interquartile range (IQR, that is, Q3 - Q1). Then, according to the nature of the data, the average value filling strategy is used to fill in the missing data to obtain the cleaned data.

[0048] After obtaining the cleaned data, the embodiments of the present application can perform standardization processing and normalization processing on the cleaned data. For example, the Z-score standardization method is used to perform standardization processing on the cleaned data, and then the Min-Max normalization method is used to perform normalization processing on the standardized data to obtain the standardized data.

[0049] The embodiments of the present application can perform data cleaning on the collected real-time meteorological observation data and the real-time operating conditions data of power generation equipment, which can effectively remove various abnormal data, and the embodiments of the present application can supplement the missing data, and then perform standardization processing and normalization processing on the cleaned data and the supplemented data, which is helpful for the subsequent utilization of data and can reduce the complexity of data processing.

[0050] Step S202: Extract multiple high-dimensional key features that affect reactive power from the standard data, perform dimensionality reduction processing on the multiple high-dimensional key features to obtain multiple low-dimensional key features, and select low-dimensional key features that meet the preset requirements from the multiple low-dimensional key features to generate a set of low-dimensional key features.

[0051] In some embodiments, the present application can also extract a plurality of low-dimensional key features that meet preset requirements and affect reactive power from standard data to generate a set of low-dimensional key features. Thus, data with extremely low influence on reactive power can be discarded, effectively reducing the calculation cost when predicting the reactive power of the target new energy power station, reducing the influence of irrelevant features on the prediction result, and ensuring the accuracy of the prediction result.

[0052] Herein, the preset requirements can be understood as certain requirements that the multiple low-dimensional features in the finally formed set of low-dimensional key features should meet. For example, the correlation between a certain key feature and reactive power is higher than 85%, etc. The specific preset requirements can be set or adjusted by those skilled in the art according to the actual situation. Only exemplary explanations are provided here, without specific limitations.

[0053] Specifically, in the embodiments of the present application, a plurality of high-dimensional key features that affect reactive power can be first extracted from the standard data. For example, using feature engineering methods, such as time series feature extraction algorithms (such as moving average method, autoregressive model, etc.), statistical feature calculation (such as mean, variance, skewness, etc.), and frequency domain feature extraction (such as Fourier transform, wavelet transform, etc.), 20 key features are extracted from the standardized multi-data source data to form a plurality of high-dimensional key features (vectors).

[0054] Then, in order to effectively utilize the key features, the embodiments of the present application can also perform dimensionality reduction processing on these multiple high-dimensional key features to obtain a plurality of low-dimensional key features, and finally select low-dimensional key features that meet certain requirements from these multiple low-dimensional key features to form a set of low-dimensional key features.

[0055] For example, the embodiments of the present application can use the principal component analysis method (PCA) to perform dimensionality reduction processing on the high-dimensional key feature vectors, and select the low-dimensional key features corresponding to the first 5 principal components for combination, thereby obtaining a set of low-dimensional key features.

[0056] The embodiments of the present application can extract a plurality of high-dimensional key features that affect reactive power from the standard data, thereby more comprehensively and accurately considering the influence of various complex factors on reactive power prediction, and then improving the accuracy and reliability of the prediction model, providing a rich data basis for subsequent analysis and prediction; further, the embodiments of the present application can also reduce the high-dimensional key features to low-dimensional key features, thereby removing redundant information in the data, reducing the amount of calculation and computational complexity, and improving efficiency and generalization ability.

[0057] Optionally, in an embodiment of the present application, low-dimensional key features that meet preset requirements are selected from multiple low-dimensional key features to generate a set of low-dimensional key features, including: constructing a covariance matrix of the multiple low-dimensional key features; decomposing the eigenvalues of the covariance matrix to select low-dimensional key features that meet the preset requirements according to the eigenvalues.

[0058] Based on the related descriptions of other embodiments, it can be understood that the present application can select low-dimensional key features that meet preset requirements from multiple low-dimensional key features to generate a set of low-dimensional key features.

[0059] In some embodiments, the present application can, but is not limited to, first construct a covariance matrix between these multiple low-dimensional key features, and then decompose the eigenvalues of the covariance matrix, thereby selecting low-dimensional key features that meet certain requirements to form a set of low-dimensional key features.

[0060] For example, for n low-dimensional key features, the present application can, but is not limited to, construct an n×n covariance matrix to measure the linear correlation between these features. Among them, each element in the matrix represents the covariance between the corresponding two features, reflecting whether their change trends are consistent and the degree of consistency. If the covariance of two features is positive, it means they tend to increase or decrease simultaneously; if the covariance is negative, it means that when one increases, the other tends to decrease; if the covariance is 0, it means there is no linear relationship between the two.

[0061] Performing eigenvalue decomposition on the covariance matrix will obtain a set of eigenvalues and corresponding eigenvectors. The eigenvalue represents the variance magnitude in the direction of the corresponding eigenvector. The larger the variance, the greater the data change in that direction and the more information it contains. The eigenvector represents the change direction of the data in that direction.

[0062] Sort the eigenvalues from largest to smallest, and the corresponding eigenvectors will form the directions of the principal components. Then, in an embodiment of the present application, the first 5 principal components can be selected, that is, the 5 directions with the largest variance. These directions can retain the information of the original data to the greatest extent, represent the main change patterns of the original features, and can capture most of the differences and laws in the data.

[0063] Then, combine the information contained in these 5 principal components to generate a set of low-dimensional key features for tasks such as data analysis and modeling. In this way, while reducing the data dimension, useful information for analysis and modeling can be retained as much as possible, improving the efficiency and performance of the model. It should be noted that the specific rule for selecting the principal components, that is, the low-dimensional key features, can be adjusted by those skilled in the art according to the actual situation. It can be the first 5, the first 10, the first 10% of all low-dimensional key features, etc. The embodiments of the present application are only for illustrative purposes and are not specifically limited.

[0064] In the embodiment of the present application, a covariance matrix between multiple low-dimensional key features can be constructed, and each element in the matrix represents the linear correlation between these features. Then, by decomposing the eigenvalues of the covariance matrix, the low-dimensional key features that meet certain requirements are selected to form a set of low-dimensional key features.

[0065] Step S203: Input the set of low-dimensional key features into a pre-established reactive power prediction model to output the reactive power prediction result of the target new energy power station.

[0066] As a possible implementation manner, after obtaining the set of low-dimensional key features, the embodiment of the present application can input the set of low-dimensional key features into a pre-established reactive power prediction model. The reactive power prediction model processes the low-dimensional key features in the set of low-dimensional key features to obtain the reactive power prediction result of the target new energy power station.

[0067] Among them, the pre-established reactive power prediction model here can be understood as a pre-established prediction model for predicting the reactive power of the target new energy power station.

[0068] For example, the present application can use the pre-established reactive power prediction model to extract the time-series features of reactive power from the set of low-dimensional key features, then perform local feature extraction, and finally output the prediction result of reactive power through a fully connected layer and an activation function.

[0069] Among them, the pre-established reactive power prediction model in the embodiment of the present application also introduces an anomaly detection algorithm. For abnormal working conditions under extreme weather conditions, anomaly point identification and filtering are performed on the reactive power prediction result. The specific process can be but is not limited to the following: For the reactive power prediction result under extreme weather conditions, an anomaly detection algorithm based on the local outlier factor (LOF) is used to analyze and process the prediction result. By calculating the local reachability density of each data point in the prediction result and its neighborhood points, the outlier factor value is obtained. Then, the outlier factor threshold is set to 5, and anomaly data points are identified for the reactive power prediction result. The data points with an outlier factor greater than 5 are marked as abnormal data. According to the anomaly point identification result, a median filtering method is used to filter the reactive power prediction result, and the sliding window size is 10 to eliminate the influence of abnormal data points and obtain an effective reactive power prediction result.

[0070] The embodiment of the present application can input the set of low-dimensional key features into a pre-established reactive power prediction model, and use key features such as meteorological conditions and operating conditions as model inputs, which can improve the accuracy of reactive power prediction, provide real-time and accurate prediction results for power grid dispatching decisions, effectively respond to the challenges brought by new energy grid connection, and improve the safety and economy of power grid operation.

[0071] Optionally, in one embodiment of the present application, before inputting the low-dimensional key feature set into the pre-established reactive power prediction model, it further includes: obtaining the historical reactive power data, historical meteorological data, and historical operating condition data of multiple new energy power stations, so as to extract the historical data features of multiple new energy power stations from the historical reactive power data, historical meteorological data, and historical operating condition data; training the pre-constructed hybrid neural network model based on the historical data features to obtain the reactive power prediction model.

[0072] Based on the relevant descriptions of other embodiments, it can be understood that the present application can input the low-dimensional key feature set into the pre-established reactive power prediction model to obtain the prediction results of the target new energy power station. Next, the construction process of this reactive power prediction model will be described.

[0073] When constructing this reactive power prediction model, the embodiments of the present application can obtain the historical reactive power data, historical meteorological data, and historical operating condition data of multiple new energy power stations, and perform the same data cleaning, data standardization processing, and data normalization processing on these historical data, etc., so as to extract the historical data features of multiple new energy power stations from the standard historical reactive power data, historical meteorological data, and historical operating condition data, and then divide these historical data features into a training set and a test set. Among them, the training set is used to train the initial prediction model, and the test set is used to test the performance and accuracy of the trained prediction model for adjustment and optimization.

[0074] Among them, the reactive power prediction model in the embodiments of the present application can integrate deep learning methods based on artificial intelligence. In recent years, although deep learning has achieved breakthroughs in many fields, applying it to reactive power prediction on the new energy side still faces many technical challenges. For example, aiming at the reactive power characteristics of new energy power generation, how to design a reasonable deep learning network structure and parameters and construct a matching prediction model is a key issue. Secondly, during the model training process, it is necessary to fully mine and integrate the multi-source heterogeneous data of new energy power stations and extract the key features hidden in the data, which puts higher requirements on the data processing and feature engineering capabilities of deep learning models.

[0075] Based on this, the reactive power prediction model in the embodiments of the present application can adopt a hybrid neural network structure combining LSTM and CNN: first, use a 3-layer LSTM network to extract temporal features, and the number of hidden units in each layer of LSTM is 128, 256, and 128 respectively; then send the output of LSTM into a 2-layer one-dimensional CNN network for local feature extraction, the convolution kernel sizes are 3 and 5 respectively, and the number of convolution kernels is 64 and 128 respectively; finally, through the fully connected layer and the activation function, the prediction results are output.

[0076] Then, the model is trained using the training set. During the model training, the embodiments of the present application can, but are not limited to, use the Adam optimizer, set the learning rate to 0.01, the batch size to 64, and the number of iterations to 100, and the model is tested using the test set. Then, the five-fold cross-validation method is adopted to evaluate the model performance, with the mean squared error (MSE) as the evaluation index. After multiple parameter adjustments, a certain reactive power prediction model is finally obtained.

[0077] The embodiments of the present application can adopt deep learning algorithms to construct a non-linear and non-stationary prediction model. Through data fusion technology, multi-source heterogeneous data is preprocessed and feature extracted. Key features such as meteorological conditions and operating conditions are used as model inputs. Combining the introduced anomaly detection algorithm to identify abnormal data under extreme weather conditions can effectively improve the robustness of the reactive power prediction model in the embodiments of the present application.

[0078] Optionally, in an embodiment of the present application, after outputting the reactive power prediction result of the target new energy power station, it further includes: based on the incremental learning strategy and the online learning mechanism, using real-time meteorological data, real-time operating conditions data of the power generation equipment, and the reactive power prediction result of the target new energy power station to update the reactive power prediction model, so as to generate a new reactive power prediction result of the target new energy power station using the updated reactive power prediction model.

[0079] In other embodiments, for abnormal conditions under some extreme weather conditions, the present application also considers the problems of improving the prediction performance and interpretability of the reactive power prediction model. These problems all require exploring the internal mechanism of deep learning to enhance the understandability and credibility of the reactive power prediction model and provide a more reliable basis for power grid dispatching decisions.

[0080] In the embodiments of the present application, an incremental learning measurement and an online learning mechanism can be introduced into the reactive power prediction model. Thus, when using the reactive power prediction model to predict the reactive power of the target new energy power station, the model parameters can also be dynamically updated according to the real-time data of the target new energy power station, adapting to the volatility and intermittency characteristics of new energy power generation, ensuring the continuous optimization and improvement of the prediction performance of the model, and providing real-time and accurate reactive power prediction results for power grid dispatching.

[0081] Specifically, when obtaining the real-time operating data of the power generation equipment of the target new energy power station, the embodiments of the present application can collect information such as power generation power and meteorological conditions in real time through sensors and data acquisition systems installed in the target new energy power station to form a time series data set. For example, the power generation power and meteorological data are recorded every 5 minutes, and the meteorological data includes wind speed, wind direction, temperature, and light intensity.

[0082] Then, the reactive power prediction model updates the model parameters hourly through the introduced incremental learning mechanism, enabling it to adapt to the volatility of new energy power generation. For example, when new data of the target new energy power station is obtained, the embodiments of the present application can use the online gradient descent method to update the weights of the reactive power prediction model to achieve incremental learning.

[0083] Furthermore, to improve the prediction accuracy and robustness of the reactive power prediction model, the embodiments of the present application can also adopt an online learning method, that is, use the real-time data continuously generated by new energy power generation equipment to perform real-time correction and update on the reactive power prediction model. In view of the intermittent characteristics of new energy power generation, in the data preprocessing stage, the embodiments of the present application can use the linear interpolation method to fill in the missing data and use the exponential smoothing method to reduce the abnormal fluctuations of the data, thereby improving the data quality. Continuously monitor the prediction performance of the model, evaluate indicators such as prediction error and stability weekly, and use evaluation indicators such as mean square error and mean absolute error (MAE) to continuously optimize and improve the reactive power prediction model.

[0084] Since the collected meteorological data and the operating condition data of the power generation equipment are both real-time, the reactive power prediction model in the embodiments of the present application will be fine-tuned with the real-time update of the data, and finally achieve the real-time update of the reactive power prediction model. When predicting the reactive power of the target new energy power station next time, the updated reactive power prediction model can be used to obtain a more accurate reactive power prediction result.

[0085] The embodiments of the present application can dynamically update the model parameters using real-time meteorological data, real-time operating condition data of the power generation equipment, and the reactive power prediction results of the target new energy power station, which can enable the reactive power prediction model in the embodiments of the present application to adapt to the volatility and intermittency characteristics of new energy power generation, and can ensure the continuous optimization and improvement of the prediction performance of the model. Furthermore, using the reactive power prediction model after updating the model parameters for the next reactive power prediction can provide real-time and accurate reactive power prediction results for power grid dispatching.

[0086] The reactive power prediction method on the new energy side proposed according to the embodiments of the present application can obtain a set of low-latitude key features affecting reactive power by collecting real-time meteorological observation data and real-time operating conditions data of power generation equipment of the target new energy power station, and then input it into the reactive power prediction model to output the prediction result of reactive power. Thus, it realizes the preprocessing and feature extraction of multi-source heterogeneous data of the target new energy power station through data fusion technology. When facing the reactive power that is non-linear and unstable due to the discontinuity and volatility of new energy power generation, it can also accurately capture the characteristics of reactive power, providing precise data support for the input data of the reactive power prediction model, effectively reducing the time and cost of the model to process data. Moreover, the anomaly detection algorithm introduced in the reactive power prediction model of the present application can identify abnormal data under extreme weather conditions, improving the robustness of the model. The incremental learning measurement and online learning mechanism adopted can dynamically update the model parameters according to real-time data, and can adapt to the volatility and intermittency characteristics of new energy power generation, thereby improving the prediction accuracy of reactive power, providing real-time and accurate prediction results for power grid dispatching decisions, effectively coping with the challenges brought by new energy grid connection, and enhancing the safety and economy of power grid operation. Thus, it solves the problems that the traditional reactive power prediction methods in the related technologies all have their own limitations, and when facing the non-linear and unstable characteristics of reactive power supply due to the discontinuity and volatility of new energy power generation, it is difficult to accurately capture the characteristics of reactive power, and ignores the influence of external factors such as meteorological conditions and operating conditions on reactive power, and it is difficult to effectively guarantee the accuracy of the prediction results.

[0087] Next, a reactive power prediction device on the new energy side proposed according to the embodiments of the present application will be described with reference to the accompanying drawings.

[0088] Figure 3 It is a schematic structural diagram of the reactive power prediction device on the new energy side of the embodiments of the present application.

[0089] As Figure 3 shown, the reactive power prediction device 10 on the new energy side includes: a data processing module 100, a feature extraction module 200, and a power prediction module 300.

[0090] Among them, the data processing module 100 is used to collect real-time meteorological observation data and real-time operating conditions data of power generation equipment at the location of the target new energy power station, and perform data processing on the real-time meteorological observation data and real-time operating conditions data of power generation equipment to obtain standard data.

[0091] A feature extraction module 200 is configured to extract multiple high-dimensional key features that affect reactive power from standard data, and perform dimensionality reduction processing on the multiple high-dimensional key features to obtain multiple low-dimensional key features, so as to select low-dimensional key features that meet preset requirements from the multiple low-dimensional key features to generate a low-dimensional key feature set.

[0092] A power prediction module 300 is configured to input the low-dimensional key feature set into a pre-established reactive power prediction model, so as to output a reactive power prediction result of a target new energy power station.

[0093] Optionally, in an embodiment of the present application, the data processing module 100 includes: a data cleaning unit and a data standardization unit.

[0094] Among them, the data cleaning unit is configured to perform data cleaning on real-time meteorological observation data and real-time operating conditions data of power generation equipment, so as to obtain real-time meteorological observation data and real-time operating conditions data of power generation equipment with abnormal data removed and / or missing data supplemented.

[0095] The data standardization unit is configured to perform standardization processing and normalization processing on the real-time meteorological observation data and real-time operating conditions data of power generation equipment with abnormal data removed and / or missing data supplemented, so as to obtain standard data.

[0096] Optionally, in an embodiment of the present application, the feature extraction module 200 includes: a construction unit and a selection unit.

[0097] Among them, the construction unit is configured to construct a covariance matrix of multiple low-dimensional key features.

[0098] The selection unit is configured to decompose the eigenvalues of the covariance matrix, so as to select low-dimensional key features that meet preset requirements according to the eigenvalues.

[0099] Optionally, in an embodiment of the present application, it further includes: an acquisition module and a training module.

[0100] Among them, the acquisition module is configured to obtain historical reactive power data, historical meteorological data, and historical operating conditions data of multiple new energy power stations before inputting the low-dimensional key feature set into a pre-established reactive power prediction model, so as to extract historical data features of the multiple new energy power stations from the historical reactive power data, historical meteorological data, and historical operating conditions data.

[0101] The training module is configured to train a pre-constructed hybrid neural network model based on the historical data features, so as to obtain a reactive power prediction model.

[0102] Optionally, in an embodiment of the present application, it further includes: an update module, configured to, after outputting the reactive power prediction result of the target new energy power station, update the reactive power prediction model based on the incremental learning strategy and the online learning mechanism, using real-time meteorological data, real-time operating conditions data of the power generation equipment, and the reactive power prediction result of the target new energy power station, so as to generate a new reactive power prediction result of the target new energy power station by using the updated reactive power prediction model.

[0103] It should be noted that the foregoing explanation of the embodiments of the reactive power prediction method on the new energy side also applies to the reactive power prediction device on the new energy side of this embodiment, and will not be elaborated here.

[0104] The reactive power prediction device on the new energy side proposed according to the embodiments of the present application can obtain a low-dimensional key feature set affecting reactive power by collecting real-time meteorological observation data and real-time operating conditions data of the power generation equipment of the target new energy power station, and then input it into the reactive power prediction model to output the prediction result of reactive power. Thus, it realizes the preprocessing and feature extraction of multi-source heterogeneous data of the target new energy power station through data fusion technology. When facing the reactive power that is nonlinear and unstable due to the discontinuity and volatility of new energy power generation, it can also accurately capture the characteristics of reactive power, providing precise data support for the input data of the reactive power prediction model, effectively reducing the time and cost of the model to process data. Moreover, the anomaly detection algorithm introduced in the reactive power prediction model of the present application can identify abnormal data under extreme weather conditions, improving the robustness of the model. The incremental learning measurement and online learning mechanism adopted can dynamically update the model parameters according to real-time data, and can adapt to the volatility and intermittency characteristics of new energy power generation, thereby improving the prediction accuracy of reactive power, providing real-time and accurate prediction results for power grid dispatching decisions, effectively coping with the challenges brought by new energy grid connection, and enhancing the safety and economy of power grid operation. Thus, it solves the problems that the traditional reactive power prediction methods in the related technologies all have their own limitations, and when facing the characteristics that the supply of reactive power is nonlinear and unstable due to the discontinuity and volatility of new energy power generation, it is difficult to accurately capture the characteristics of reactive power, and ignores the influence of external factors such as meteorological conditions and operating conditions on reactive power, and it is difficult to effectively guarantee the accuracy of the prediction results.

[0105] Figure 4 The structural schematic diagram of the electronic device provided for the embodiments of the present application. The electronic device may include:

[0106] A memory 401, a processor 402, and a computer program stored on the memory 401 and executable on the processor 402.

[0107] When the processor 402 executes the program, it implements the new energy side reactive power prediction method provided in the above embodiments.

[0108] Furthermore, the electronic device further includes:

[0109] A communication interface 403, which is used for communication between the memory 401 and the processor 402.

[0110] A memory 401, which is used to store computer programs that can run on the processor 402.

[0111] The memory 401 may include a high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0112] If the memory 401, the processor 402, and the communication interface 403 are implemented independently, the communication interface 403, the memory 401, and the processor 402 can be interconnected through a bus and complete communication with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of simplicity of representation, Figure 4 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0113] Optionally, in a specific implementation, if the memory 401, the processor 402, and the communication interface 403 are integrated on a chip, the memory 401, the processor 402, and the communication interface 403 can complete communication with each other through an internal interface.

[0114] The processor 402 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0115] The embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the new energy side reactive power prediction method as described above.

[0116] The embodiments of the present application further provide a computer program product, including a computer program which can run computer instructions, and when the computer instructions are executed by a processor, the reactive power prediction method on the new energy side provided by the embodiments of the present application is implemented.

[0117] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or N embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0118] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of the present application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically and clearly defined.

[0119] Any process or method description in the flowchart or described in other ways herein can be understood as representing a module, segment or part of code including one or N executable instructions for implementing a customized logic function or process, and the scope of the preferred embodiments of the present application includes additional implementations, where the functions can be executed in a manner that is not shown or discussed in sequence, including in a substantially simultaneous manner according to the functions involved or in a reverse order, which should be understood by those skilled in the art of the embodiments of the present application.

[0120] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definable sequence list of executable instructions for implementing logical functions, which can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion (electronic device) having one or N wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, as the program can be obtained electronically by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then stored in a computer memory.

[0121] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or combinations thereof. In the above-described embodiments, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or a combination of the following techniques known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0122] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the methods of the above-described embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0123] In addition, each functional unit in various embodiments of the present application may be integrated into one processing module, may exist physically alone for each unit, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0124] The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for predicting reactive power on the renewable energy side, characterized in that: The following steps are involved: Collecting real-time meteorological observation data and real-time operating condition data of power generation equipment at the location of the target new energy site, and processing the real-time meteorological observation data and the real-time operating condition data of the power generation equipment to obtain standard data; Extracting multiple high-dimensional key features that affect reactive power from the standard data, and performing dimensionality reduction processing on the multiple high-dimensional key features to obtain multiple low-dimensional key features, so as to select low-dimensional key features that meet preset requirements from the multiple low-dimensional key features to generate a low-dimensional key feature set; The low-dimensional key feature set is input into a pre-established reactive power prediction model to output a reactive power prediction result of the target new energy station.

2. The method according to claim 1, characterized in that The data processing of the real-time meteorological observation data and the real-time operating condition data of the power generation equipment includes: Performing data cleaning on the real-time meteorological observation data and the real-time operating condition data of the power generation equipment to obtain the real-time meteorological observation data and the real-time operating condition data of the power generation equipment with abnormal data removed and / or missing data supplemented; The real-time meteorological observation data and the real-time operating condition data of the power generation equipment from which the abnormal data is removed and / or the missing data is supplemented are standardized and normalized to obtain the standard data.

3. The method according to claim 1, characterized in that The step of selecting low-dimensional key features that meet preset requirements from the plurality of low-dimensional key features to generate a low-dimensional key feature set includes: Constructing a covariance matrix of the plurality of low-dimensional key features; Decomposing the eigenvalues ​​of the covariance matrix to select the low-dimensional key features that meet the preset requirements according to the eigenvalues.

4. The method according to claim 1, characterized in that: Before inputting the low-dimensional key feature set into the pre-established reactive power prediction model, the method further includes: Acquire historical reactive power data, historical meteorological data, and historical operating condition data of a plurality of new energy sites, so as to extract historical data features of the plurality of new energy sites from the historical reactive power data, historical meteorological data, and historical operating condition data; A pre-built hybrid neural network model is trained based on the historical data features to obtain the reactive power prediction model.

5. The method according to claim 1, characterized in that After outputting the reactive power prediction result of the target new energy station, the method further includes: Based on the incremental learning strategy and the online learning mechanism, the reactive power prediction model is updated using the real-time meteorological data, the real-time operating condition data of the power generation equipment and the reactive power prediction results of the target new energy station, so as to generate a new reactive power prediction result of the target new energy station using the updated reactive power prediction model.

6. A reactive power prediction device for a new energy source, characterized in that: include: A data processing module, used to collect real-time meteorological observation data and real-time operating condition data of power generation equipment at the location of the target new energy station, and perform data processing on the real-time meteorological observation data and the real-time operating condition data of the power generation equipment to obtain standard data; A feature extraction module, used to extract multiple high-dimensional key features that affect reactive power from the standard data, and perform dimensionality reduction processing on the multiple high-dimensional key features to obtain multiple low-dimensional key features, so as to select low-dimensional key features that meet preset requirements from the multiple low-dimensional key features to generate a low-dimensional key feature set; The power prediction module is used to input the low-dimensional key feature set into a pre-established reactive power prediction model to output the reactive power prediction result of the target new energy site.

7. The device according to claim 6, characterized in that The data processing module comprises: A data cleaning unit, used for cleaning the real-time meteorological observation data and the real-time operating condition data of the power generation equipment to obtain the real-time meteorological observation data and the real-time operating condition data of the power generation equipment with abnormal data removed and / or missing data supplemented; The data standardization unit is used to standardize and normalize the real-time meteorological observation data and the real-time operating condition data of the power generation equipment from which abnormal data is removed and / or missing data is supplemented, so as to obtain the standard data.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the reactive power prediction method for the new energy side as described in any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the reactive power prediction method for the new energy side as described in any one of claims 1-5.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed, it is used to implement the reactive power prediction method for the new energy side as described in any one of claims 1-5.

Images