Power load data processing method and device, electronic equipment and storage medium

Abstract

The invention discloses a power load data processing method and device, electronic equipment and a storage medium, and relates to the field of power. The method comprises the following steps: acquiring original power load data of a plurality of power grid sub-regions; extracting modal feature information in the original power load data under different time scales through a feature extraction model; and inputting the modal feature information into a power load data prediction model, and predicting the power load data of the plurality of power grid sub-regions based on the modal feature information through the power load data prediction model to obtain predicted power load data. According to the invention, the technical problem of low prediction accuracy of power load data is solved.

Description

Technical Field

[0001] The present invention relates to the field of electric power, and in particular to a method, device, electronic equipment and storage medium for processing electric power load data. Background Art

[0002] The power system is the cornerstone of modern economic and social operations, and its safety, stability and economic operation are crucial to national development. Power regulation is the core link of power system management. If the power load data of each sub-region in the power grid can be predicted in advance, the allocation and dispatch of power resources can be effectively carried out, the instability of the power grid caused by the imbalance of supply and demand can be prevented, the waste of electricity can be reduced, and the operating efficiency of the power system can be improved.

[0003] However, since the relevant technology fails to fully consider the multi-scale characteristics of power load changes, there is a problem of low accuracy in power load data prediction.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method, device, electronic device and storage medium for processing power load data, so as to at least solve the technical problem of low accuracy in power load data prediction.

[0006] According to one aspect of an embodiment of the present invention, a method for processing power load data is provided, comprising: collecting original power load data of multiple power grid sub-areas; extracting modal feature information at different time scales in the original power load data through a feature extraction model; inputting the modal feature information into a power load data prediction model, and predicting the power load data of multiple power grid sub-areas based on the modal feature information through the power load data prediction model to obtain predicted power load data.

[0007] In one embodiment of the present application, modal feature information at different time scales is extracted from original power load data through a feature extraction model, including: preprocessing the original power load data to obtain preprocessed data; fitting the preprocessed data using the feature extraction model to obtain a residual sequence; decomposing the residual sequence using a multiple time series decomposition technique to obtain modal feature information at different time scales, and the different time scales include at least two of the following: day, week, month, and year.

[0008] In one implementation of the present application, the original power load data is preprocessed to obtain preprocessed data, including: using a firefly algorithm to identify abnormal data in the original power load data; removing the abnormal data from the original power load data to obtain preprocessed data.

[0009] In one embodiment of the present application, modal feature information is input into a power load data prediction model, and the power load data of multiple power grid sub-areas are predicted based on a prediction data set by the power load data prediction model to obtain predicted power load data, including: integrating modal feature information at different time scales to obtain a prediction data set; inputting the prediction data set into the power load data prediction model, and predicting the power load data of multiple power grid sub-areas by the power load data prediction model to obtain predicted power load data.

[0010] In one embodiment of the present application, original power load data of multiple power grid sub-areas are collected, including: dividing the power grid coverage area to obtain multiple power grid sub-areas; determining collection frequencies of the multiple power grid sub-areas based on the power consumption characteristics of different groups in the multiple power grid sub-areas; determining a data collection strategy based on the collection frequencies of the multiple power grid sub-areas; and collecting original power load data of the multiple power grid sub-areas based on the data collection strategy.

[0011] In one implementation of the present application, the feature extraction model is a model obtained by integrating an autoregressive module, a moving average module, and an autoregressive moving average module.

[0012] In one embodiment of the present application, the order of the feature extraction model is obtained by adjusting the order of the initial feature extraction model based on the autocorrelation coefficient and the partial autocorrelation coefficient, wherein the autocorrelation coefficient and the partial autocorrelation coefficient are determined based on samples in the time series, and the parameters of the feature extraction model are obtained by adjusting the parameters of the initial feature extraction model based on the minimum information criterion.

[0013] According to another aspect of an embodiment of the present invention, there is also provided a device for processing power load data, including: an acquisition module for acquiring original power load data of multiple power grid sub-areas; an extraction module for extracting modal feature information at different time scales in the original power load data through a feature extraction model; and a prediction module for inputting the modal feature information into a power load data prediction model, and predicting the power load data of multiple power grid sub-areas based on the modal feature information through the power load data prediction model to obtain predicted power load data.

[0014] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein when the program is running, the electronic device is controlled to execute the method for processing power load data.

[0015] According to another aspect of an embodiment of the present invention, there is further provided a computer-readable storage medium, comprising a stored executable program, wherein when the executable program runs, the device where the storage medium is located is controlled to execute the method for processing power load data.

[0016] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and the computer program implements the method for processing power load data when executed by a processor.

[0017] In an embodiment of the present invention, original power load data of multiple power grid sub-areas are collected; modal feature information at different time scales in the original power load data is extracted by a feature extraction model; the modal feature information is input into a power load data prediction model, and the power load data of multiple power grid sub-areas are predicted based on the modal feature information by the power load data prediction model to obtain predicted power load data. It is easy to notice that by collecting original power load data of multiple power grid sub-areas and extracting modal feature information at different time scales by a feature extraction model, the modal feature information at different time scales can capture the multi-scale dynamic characteristics of power load data, and predicting power load data by a power load data prediction model based on modal feature information at different time scales can achieve the technical effect of improving the accuracy of power load data prediction, thereby solving the technical problem of low accuracy of power load data prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0019] Figure 1 is a flow chart of a method for processing power load data according to an embodiment of the present invention;

[0020] Figure 2 is a flow chart of an optional method for processing power load data according to an embodiment of the present invention;

[0021] Figure 3 It is a schematic diagram of an optional method of processing data using a feature extraction model according to an embodiment of the present invention.

[0022] Figure 4 is a schematic diagram of a device for processing power load data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] Example 1

[0026] According to an embodiment of the present invention, an embodiment of a method for processing power load data is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] Figure 1 is a flow chart of a method for processing power load data according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0028] Step S102, collecting original power load data of multiple power grid sub-areas.

[0029] The grid sub-area in the above steps is a local area within the grid divided according to factors such as geography, administration, economy or power load characteristics.

[0030] The original power load data in the above steps is power consumption data collected from the power system, which usually includes power consumption in different time periods, specifically, it can be hourly power consumption or daily power consumption, but is not limited thereto.

[0031] In an optional embodiment, a collection device is installed at a key node in each sub-area, and the original power load data is collected through the collection device, wherein the key nodes include but are not limited to substations, large user equipment connected to the power grid, etc., and the collection device can be a smart meter, a data collection terminal, etc., but is not limited thereto. The collection device can collect the original power load data in real time, or it can collect the original power load data at a preset frequency through the collection device.

[0032] When collecting raw power load data at a preset frequency, the data collection frequency can be set according to the power load characteristics of the sub-area. For example, for industrial areas with fast load changes, a higher frequency of collection may be required, such as once every 5 minutes; while for residential areas with stable load changes, a lower frequency of collection may be used, such as once every hour. It should be noted that the above collection frequency is only an example, and the specific value of the collection frequency in actual applications may not be limited to this.

[0033] Step S104: extracting modal feature information at different time scales from the original power load data through a feature extraction model.

[0034] The feature extraction model in the above steps is a model for identifying and extracting key features from the original power load data, which can analyze the time series characteristics in the original power load data and extract modal feature information at different time scales. The feature extraction model can be ARMA (Autoregressive Moving Average model), but is not limited to this.

[0035] The modal feature information in the above steps is the features under different modes identified from the original power load data. For the power load data, the modal feature information may include: short-term fluctuation information such as daily peak and trough power consumption, medium-term fluctuation information such as weekly or monthly changes in power consumption patterns, and long-term trend information such as annual power consumption growth.

[0036] The time scale in the above steps refers to daily, weekly, monthly, quarterly, annual, etc. In power load data, features of different time scales have different effects on load prediction. For example, intraday fluctuations may be affected by user behavior, while seasonal changes may be affected by factors such as weather and holidays. Considering features of different time scales helps improve the accuracy of power load data prediction.

[0037] In an optional embodiment, the ARMA model can be selected as a feature extraction model. The ARMA model identifies the long-term trend and periodic pattern in the original power load data through the AR (Autoregressive) part, and captures short-term random fluctuations through the MA (Moving Average) part. Among them, the long-term trend can be a long-term load growth pattern or a load reduction pattern, the periodic pattern can be a change in power demand within a daily, weekly, monthly, etc. periodic pattern, and the short-term random fluctuation is an unpredictable random fluctuation in the short term. Then, feature information at different time scales is extracted from the captured long-term trends, periodic patterns, and short-term random fluctuations.

[0038] Step S106, inputting the modal characteristic information into the power load data prediction model, and predicting the power load data of multiple power grid sub-areas based on the modal characteristic information through the power load data prediction model to obtain predicted power load data.

[0039] The power load data prediction model in the above steps is used to predict the power load data of each sub-region in the power grid within a certain period of time in the future. The power load data prediction model comprehensively considers multiple modal characteristics of historical power consumption data and can predict power load data more accurately.

[0040] The predicted power load data in the above steps is the data obtained by estimating the power demand of the power system in the future period using the power load data prediction model. By obtaining highly accurate predicted power load data, power companies can adjust power supply in advance according to the predicted power load data, ensure the balance of power supply and demand, and improve the stability and economy of the power grid system operation.

[0041] In an optional embodiment, the modal feature information at different time scales at the current moment can be integrated to obtain a feature vector containing the modal feature information at different time scales. The feature vector is input into the power load data prediction model, and the power load data prediction model predicts the power load data of multiple power grid sub-areas in the future to obtain predicted power load data. The predicted power load data of multiple power grid sub-areas are integrated to form a comprehensive prediction result of the power grid power load data, and the comprehensive prediction result can be used as a decision-making basis for power grid resource allocation, thereby optimizing power grid resource allocation and ensuring the balance of supply and demand within the power grid.

[0042] In an embodiment of the present invention, original power load data of multiple power grid sub-areas are collected; modal feature information at different time scales in the original power load data is extracted by a feature extraction model; the modal feature information is input into a power load data prediction model, and the power load data of multiple power grid sub-areas are predicted by the power load data prediction model to obtain predicted power load data. It is easy to notice that by collecting original power load data of multiple power grid sub-areas and extracting modal feature information at different time scales by a feature extraction model, the modal feature information at different time scales can capture the multi-scale dynamic characteristics of power load data, and predicting power load data by a power load data prediction model based on modal feature information at different time scales can achieve the technical effect of improving the accuracy of power load data prediction, thereby solving the technical problem of low accuracy of power load data prediction.

[0043] In one embodiment of the present application, modal feature information at different time scales is extracted from original power load data through a feature extraction model, including: preprocessing the original power load data to obtain preprocessed data; fitting the preprocessed data using the feature extraction model to obtain a residual sequence; decomposing the residual sequence using a multiple time series decomposition technique to obtain modal feature information at different time scales, and the different time scales include at least two of the following: day, week, month, and year.

[0044] The residual sequence in the above steps can represent the difference between the predicted value and the actual value of the modal feature information of the feature extraction model, and contains the change information of the original power load data. The modal feature information can be obtained by decomposing the residual sequence.

[0045] The multiple time series decomposition technology in the above steps refers to the technology of decomposing a time series data into multiple components. The multiple components may include three components: long-term trend, seasonal pattern and short-term random fluctuation. The long-term trend refers to the long-term change direction of the time series. The seasonal pattern reflects the periodic pattern, such as the regular changes of days, weeks, months and years. The short-term random fluctuation is the unpredictable random change in the short term.

[0046] In an optional embodiment, the original power load data is preprocessed, and the preprocessing may include data cleaning, data conversion, normalization, etc., and the data cleaning removes or corrects abnormal values, missing values, and duplicate values ​​to ensure the integrity and accuracy of the data. The data conversion makes the data present better stability, which is convenient for the processing of the feature extraction model, and the data is converted into a unified scale through normalization. After the original power load data is preprocessed, the preprocessed data is obtained.

[0047] When the feature extraction model is used to fit the preprocessed data, the maximum likelihood estimation method can be used to optimize the parameters of the feature extraction model so that the residual sum of squares between the model prediction value and the actual observation value is minimized. Once the feature extraction model is fitted, the residual sequence can be obtained by subtracting the predicted value of the feature extraction model from the actual observation value, and the residual sequence can be used for further feature extraction.

[0048] Then, when the residual sequence is decomposed using multiple time series decomposition techniques, the long-term trend, seasonal pattern and remaining short-term random fluctuations in the residual sequence are identified. For power load data, the daily, weekly, monthly and annual periodic patterns are the most critical modal feature information. The periodic patterns as modal feature information can be extracted separately. For example, the daily scale feature information can include the daily peak and trough of electricity consumption, the weekly scale feature information can reflect the difference in electricity consumption between weekends and weekdays, the monthly scale feature information can involve the monthly changes in electricity consumption patterns, and the annual scale feature information can reveal seasonal changes in electricity consumption.

[0049] In one implementation of the present application, the original power load data is preprocessed to obtain preprocessed data, including: using a firefly algorithm to identify abnormal data in the original power load data; removing the abnormal data from the original power load data to obtain preprocessed data.

[0050] The Firefly Algorithm (FA) in the above steps is a meta-heuristic optimization algorithm based on the behavior of fireflies. The attraction and brightness between fireflies are related to the distance between them, and fireflies with higher brightness have stronger attraction to fireflies with lower brightness. The data points in the original power load data can be regarded as fireflies. If the fitness of the data point, that is, the brightness of the firefly, is lower than the average level of the group, the data point can be considered as an outlier.

[0051] In an optional embodiment, each data point in the original power load data is regarded as a firefly, and it is assumed that the brightness of each firefly, that is, the fitness of the data point, is initially equal. Then a brightness attenuation rule is defined to determine that the brightness decreases with increasing distance. Since the distance between fireflies is related to the brightness of the fireflies, the brightness of the fireflies can be characterized by the average distance between the data point and other data points.

[0052] The distance between data points is calculated as follows:

[0053]

[0054] Among them, d p，q For data point x p and data point x q The distance between them, assuming the initial number of clusters is 2, c pr For data point x p The cluster center, c qr For data point x q The cluster center of , r is the number of iterations, and M is the total number of iterations.

[0055] The average distance between a data point and other data points can be determined based on the distance between the data points, and then the brightness of the fireflies corresponding to the data points can be determined, and data points with brightness lower than the group average can be screened out. Data points below the group average are significantly different from most other data points, which may be due to measurement errors, extreme events or other abnormal factors. Therefore, these data points are identified as outliers.

[0056] The identified outliers will be removed from the original power load data to prevent the outliers from affecting the accuracy of data processed by the feature extraction model.

[0057] In one embodiment of the present application, modal feature information is input into a power load data prediction model, and power load data of multiple power grid sub-areas are predicted by the power load data prediction model to obtain predicted power load data, including: integrating modal feature information at different time scales to obtain a prediction data set; inputting the prediction data set into the power load data prediction model, and predicting power load data of multiple power grid sub-areas based on the prediction data set by the power load data prediction model to obtain predicted power load data.

[0058] The prediction data set in the above steps is a data set obtained by integrating modal feature information at different time scales.

[0059] In an optional embodiment, the modal feature information at different time scales can be combined using data fusion technology, and the data fusion technology can be weighted average, principal component analysis, etc., but is not limited thereto. After combining the modal feature information at different time scales, a feature vector that can be input into the power load data prediction model is obtained, and the feature vector can contain modal features at multiple time scales. The feature vectors are organized into a prediction data set, and the prediction data set can include corresponding feature vectors of multiple power grid sub-areas.

[0060] The prediction data set is input into the trained power load data prediction model, which generates predictions of future power loads of multiple power grid sub-regions based on the input feature information to obtain predicted power load data. The prediction of future power loads of each power grid sub-region can include predictions at multiple time points, such as predicting power loads for the next day, week, or month.

[0061] In one embodiment of the present application, original power load data of multiple power grid sub-areas are collected, including: dividing the power grid coverage area to obtain multiple power grid sub-areas; determining collection frequencies of the multiple power grid sub-areas based on the power consumption characteristics of different groups in the multiple power grid sub-areas; determining a data collection strategy based on the collection frequencies of the multiple power grid sub-areas; and collecting original power load data of the multiple power grid sub-areas based on the data collection strategy.

[0062] The electricity consumption characteristics in the above steps are the electricity consumption behavior patterns of the user group within a certain period of time, including the total amount of electricity demand, the law of demand changes over time, the peak and valley values ​​of demand, etc.

[0063] The data collection strategy in the above steps includes characteristics such as data collection frequency. When formulating the data collection strategy, the electricity consumption characteristics of user groups in different regions need to be taken into account.

[0064] In an optional embodiment, the grid coverage area is divided based on factors such as geographical location, administrative boundaries, and user types, and the entire grid coverage area is subdivided into multiple relatively homogeneous sub-areas to obtain multiple grid sub-areas, each of which may include one or more substations, distribution networks, and power user groups.

[0065] Analyze the electricity consumption characteristics of each sub-area, such as seasonal changes, daytime load curves, peak demand time, etc., and determine the data collection frequency of each sub-area based on these characteristics. For example, for commercial areas with large fluctuations in electricity demand, higher-frequency data collection may be required to capture its rapidly changing patterns; while for residential areas with relatively stable electricity demand, a lower collection frequency can be used. In addition, the total collection range, data collection limits, repeated collection range, etc. can be determined based on the electricity consumption characteristics, and then the data collection frequency can be calculated. The calculation formula is as follows:

[0066]

[0067] Among them, N is the data collection frequency, f is the total collection range, s is the basic collection unit difference, a is the data collection limit, h is the repeated collection range, and c is the directional data collection.

[0068] Then, the data collection frequencies of multiple sub-areas are integrated to formulate a global data collection strategy, and then the power load data is actually collected according to the data collection strategy.

[0069] In one implementation of the present application, the feature extraction model is a model obtained by integrating an autoregressive module, a moving average module, and an autoregressive moving average module.

[0070] The autoregression module in the above steps is used for time series analysis. In the autoregression model, the value at the current time point is a linear combination of the values ​​at several past time points and a random error term, where the value at the current time point is the response variable and the values ​​at several past time points are the independent variables.

[0071] The moving average module in the above steps is used for time series analysis and data series smoothing. The basic idea is to generate a new series by calculating the average of several consecutive values ​​in the data series. This new series is usually smoother and reduces random fluctuations or noise in the original data, making it easier to identify trends or patterns in the data.

[0072] The autoregressive moving average module in the above steps is used for time series analysis, which combines the characteristics of the autoregressive module and the moving average module, and also includes an optional difference link to process non-stationary time series data.

[0073] In an optional embodiment, the definition formula of the autoregressive module is:

[0074]

[0075] Among them, y t is the current numerical term, t is the current time, μ is the constant term, p is the autoregressive order, y i is the autocorrelation coefficient, y t-1 is the numerical term at time t-1, e t is the error term at time t, and i is the i-th time point counting back from time point t.

[0076] The definition formula of the moving average module is:

[0077]

[0078] Among them, y t is the current numerical term, t is the current time, μ is the constant term, e t is the error term at time t, q is the moving average order, i is the i-th time point from time point t, θ i is the moving average coefficient, e t-1 is the error term at time t-1.

[0079] The definition formula of the autoregressive moving average module is:

[0080]

[0081] Among them, y t is the current numerical term, t is the current time, μ is the constant term, p is the autoregressive order, i is the i-th time point from time point t, and y i is the autocorrelation coefficient, y t-1 is the numerical term at time t-1, e t is the error term at time t, q is the moving average order, θ i is the moving average coefficient, e t-1 is the error term at time t-1.

[0082] In one embodiment of the present application, the order of the feature extraction model is obtained by adjusting the order of the initial feature extraction model based on the autocorrelation coefficient and the partial autocorrelation coefficient, wherein the autocorrelation coefficient and the partial autocorrelation coefficient are determined based on samples in the time series, and the parameters of the feature extraction model are obtained by adjusting the parameters of the initial feature extraction model based on the minimum information criterion.

[0083] The initial feature extraction model in the above steps is a model that combines an autoregressive module, a moving average module, and an autoregressive moving average module, but does not adjust the order and parameters.

[0084] The autocorrelation coefficient (ACF) in the above steps is used to measure the correlation between values ​​at different time points in a time series, and can calculate the degree of linear dependence between a point in the series and a point in its past.

[0085] The partial autocorrelation coefficient (PACF) in the above steps is an extension of the autocorrelation coefficient, which is used to measure the correlation between a point in a sequence and a point in its past, but removes the correlation effect of all intermediate points between the two points in the sequence.

[0086] The minimum information criterion (Akaike Information Criterion, AIC) in the above steps is a metric used to evaluate the relative quality of statistical feature extraction models. AIC combines the likelihood function that represents the degree of match between the feature extraction model and the data, and the penalty term of the model complexity to balance the fit and complexity of the feature extraction model. The smaller the AIC value, the better the model is considered, because it means that the feature extraction model can both fit the data well and avoid overfitting.

[0087] The order in the above steps is used to characterize how many values ​​or errors of past time points the feature extraction model uses to predict the value of the current time point.

[0088] The above steps are coefficient values ​​in the feature extraction model, which may include but are not limited to autoregressive coefficients, moving average coefficients, and constant terms.

[0089] In an optional embodiment, the calculation formula of the autocorrelation coefficient is as follows:

[0090]

[0091] in, is the autocorrelation coefficient of time lag k, where k is the time lag, that is, the time interval between the current value and the historical value used for comparison in the time series, x t is the observed value at time point t, x t-k is the observed value at time point tk, is the mean value of the series, and n is the length of the time series, that is, the total number of observations.

[0092] Then, the partial autocorrelation coefficient can be calculated through multivariate regression analysis based on the autocorrelation coefficient, and then the order q of the moving average module and the order p of the autoregressive module can be determined respectively based on the truncation characteristics of the autocorrelation coefficient and the partial autocorrelation coefficient. The truncation behavior of the partial autocorrelation coefficient is more obvious, so the order q of the moving average module can be determined based on the truncation characteristics of the partial autocorrelation coefficient, and the partial autocorrelation coefficient will not be suddenly truncated after the qth order, so the order p of the autoregressive module is still determined based on the autocorrelation coefficient. According to the order q of the moving average module and the order p of the autoregressive module, the orders of the moving average module and the autoregressive module in the feature extraction model are adjusted to obtain the order of the feature extraction model.

[0093] When adjusting the parameters of the initial feature extraction model based on the minimum information criterion, it is necessary to determine the parameter combination that minimizes AIC through parameter search technology, where the parameter search technology may include but is not limited to gradient descent, quasi-Newton method, genetic algorithm, etc. The formula of AIC is as follows:

[0094]

[0095] Where AIC(p,q) is the value of AIC, p is the order of the autoregressive module, q is the order of the moving average module, n is the sample size of the parameter, represents the variance of the fitted residual, and ln is the logarithmic calculation.

[0096] After determining the parameter combination that minimizes the AIC, the parameters of the initial feature extraction model are adjusted based on the parameter combination to obtain the parameters of the feature extraction model.

[0097] A preferred embodiment is described below. Figure 2 is a flow chart of an optional method for processing power load data according to an embodiment of the present invention, such as Figure 2 As shown, the method may include the following steps:

[0098] Step S202: Collect user electricity consumption data.

[0099] Among them, the user's electricity consumption data is the original power load data mentioned above.

[0100] Step S204: distinguish the acquisition frequencies according to the acquisition area characteristics.

[0101] Analyze the electricity consumption characteristics of the collection area and determine the data collection frequency of the collection area.

[0102] Step S206: removing abnormal values ​​in the electricity consumption data.

[0103] Use the firefly algorithm to identify and remove abnormal data in electricity consumption data.

[0104] Step S208, input the training model to obtain parameters.

[0105] The historical electricity consumption data is used to train the association analysis model, that is, the power load data prediction model mentioned above, to determine the parameters of the association analysis model and obtain the trained association analysis model.

[0106] Step S210: constructing an association analysis model based on the multiple time series decomposition model.

[0107] A multiple time series decomposition model, that is, the feature extraction model mentioned above, is used to extract the modal feature information in the current electricity consumption data. A prediction data set of the association analysis model is constructed based on the modal feature information. The prediction data set is input into the association analysis model to predict future power load data.

[0108] Figure 3 is a schematic diagram of an optional method of processing data using a feature extraction model according to an embodiment of the present invention, such as Figure 3 As shown, the collected power load sequence, that is, the original power load data mentioned above, is first preprocessed, and then the preprocessed data is input into the feature extraction model to extract modal feature information. Among them, the construction process of the feature extraction model includes creating a moving average module, creating an autoregressive module, and creating an autoregressive moving average module, wherein the moving average module contains a moving average model, the autoregressive module contains an autoregressive model, and the autoregressive moving average module contains an autoregressive moving average model, and then an optimization calculation is performed, that is, the parameters of the moving average module, the end of the autoregressive module and the entire feature extraction model are optimized to obtain better parameters.

[0109] Example 2

[0110] According to an embodiment of the present invention, an embodiment of a device for processing power load data is provided, and the device can execute the method for processing power load data provided in the above-mentioned embodiment 1. The specific implementation method and preferred application scenario are the same as those of the above-mentioned embodiment 1, and will not be repeated here.

[0111] Figure 4 is a schematic diagram of a device for processing power load data according to an embodiment of the present invention. Figure 4 As shown:

[0112] A collection module 40 is used to collect raw power load data of multiple power grid sub-areas;

[0113] An extraction module 42, used to extract modal feature information at different time scales from the original power load data through a feature extraction model;

[0114] The prediction module 44 is used to input the modal characteristic information into the power load data prediction model, and predict the power load data of multiple power grid sub-areas based on the modal characteristic information through the power load data prediction model to obtain predicted power load data.

[0115] The extraction module includes: a preprocessing unit, which is used to preprocess the original power load data to obtain preprocessed data; a fitting unit, which is used to fit the preprocessed data using a feature extraction model to obtain a residual sequence; a decomposition unit, which is used to decompose the residual sequence using a multiple time series decomposition technology to obtain modal feature information at different time scales, and the different time scales include at least two of the following: day, week, month, and year.

[0116] The preprocessing unit is also used to identify abnormal data in the original power load data using the firefly algorithm; and remove the abnormal data from the original power load data to obtain preprocessed data.

[0117] The prediction module includes: an integration unit, which is used to integrate the modal feature information at different time scales to obtain a prediction data set; a prediction unit, which is used to input the prediction data set into the power load data prediction model, and predict the power load data of multiple power grid sub-areas based on the prediction data set through the power load data prediction model to obtain predicted power load data.

[0118] The collection module includes: a division unit, which is used to divide the grid coverage area into multiple grid sub-areas; a first determination unit, which is used to determine the collection frequency of multiple grid sub-areas based on the electricity consumption characteristics of different groups in the multiple grid sub-areas; a second determination unit, which is used to determine the data collection strategy based on the collection frequency of the multiple grid sub-areas; and a collection unit, which is used to collect the original power load data of the multiple grid sub-areas based on the data collection strategy.

[0119] The feature extraction model in the extraction module is a model obtained by integrating the autoregressive module, the moving average module, and the autoregressive moving average module.

[0120] The order of the feature extraction model in the extraction module is obtained by adjusting the order of the initial feature extraction model based on the autocorrelation coefficient and the partial autocorrelation coefficient, wherein the autocorrelation coefficient and the partial autocorrelation coefficient are determined based on samples in the time series, and the parameters of the feature extraction model are obtained by adjusting the parameters of the initial feature extraction model based on the minimum information criterion.

[0121] Example 3

[0122] According to an embodiment of the present invention, there is also provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method for processing power load data in embodiment 1 is executed when the program is running.

[0123] Example 4

[0124] An embodiment of the present application further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the power load data processing method in each embodiment of the present invention.

[0125] Example 5

[0126] An embodiment of the present application further provides a computer program product, including a computer program, which implements the method for processing power load data in various embodiments of the present invention when executed by a processor.

[0127] Example 6

[0128] An embodiment of the present application also provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method for processing power load data in each embodiment of the present invention is implemented.

[0129] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0130] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0131] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0132] 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 units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0133] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0134] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0135] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for processing power load data, characterized in that: include: Collect raw power load data from multiple power grid sub-areas; Extracting modal feature information at different time scales from the original power load data through a feature extraction model; The modal characteristic information is input into a power load data prediction model, and the power load data of the plurality of power grid sub-areas are predicted based on the modal characteristic information by the power load data prediction model to obtain predicted power load data.

2. The method for processing power load data according to claim 1, characterized in that: The modal feature information at different time scales in the original power load data is extracted by a feature extraction model, including: Preprocessing the original power load data to obtain preprocessed data; Using the feature extraction model to perform fitting processing on the preprocessed data to obtain a residual sequence; The residual sequence is decomposed by using a multiple time series decomposition technique to obtain modal feature information at different time scales, where the different time scales include at least two of the following: day, week, month, and year.

3. The method for processing power load data according to claim 2, characterized in that: Preprocessing the original power load data to obtain preprocessed data includes: Using a firefly algorithm to identify abnormal data in the raw power load data; The abnormal data is removed from the original power load data to obtain the preprocessed data.

4. The method for processing power load data according to claim 1, characterized in that: Inputting the modal characteristic information into a power load data prediction model, and predicting the power load data of the plurality of power grid sub-areas by using the power load data prediction model to obtain predicted power load data, including: Integrating the modal feature information at different time scales to obtain a prediction data set; The prediction data set is input into the power load data prediction model, and the power load data of the plurality of power grid sub-areas are predicted based on the prediction data set by the power load data prediction model to obtain predicted power load data.

5. The method for processing power load data according to claim 1, characterized in that: Collect raw power load data from multiple grid sub-areas, including: Dividing the grid coverage area to obtain a plurality of grid sub-areas; Determining the collection frequencies of the multiple power grid sub-areas based on the power consumption characteristics of different groups in the multiple power grid sub-areas; determining a data collection strategy based on the collection frequencies of the plurality of power grid sub-areas; The original power load data of multiple power grid sub-areas are collected based on the data collection strategy.

6. The method for processing power load data according to claim 1, characterized in that: The feature extraction model is a model obtained by integrating an autoregressive module, a moving average module, and an autoregressive moving average module.

7. The method for processing power load data according to claim 6, characterized in that: The order of the feature extraction model is obtained by adjusting the order of the initial feature extraction model based on the autocorrelation coefficient and the partial autocorrelation coefficient, wherein the autocorrelation coefficient and the partial autocorrelation coefficient are determined based on samples in the time series, and the parameters of the feature extraction model are obtained by adjusting the parameters of the initial feature extraction model based on the minimum information criterion.

8. A device for processing power load data, characterized in that: include: A collection module, used to collect raw power load data of multiple power grid sub-areas; An extraction module, used for extracting modal feature information at different time scales from the original power load data through a feature extraction model; The prediction module is used to input the modal characteristic information into the power load data prediction model, and predict the power load data of the multiple power grid sub-areas based on the modal characteristic information through the power load data prediction model to obtain predicted power load data.

9. An electronic device, characterized in that: include: A memory storing an executable program; A processor is used to run the program, wherein the program, when running, executes the method for processing power load data as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the storage medium is located is controlled to execute the method for processing power load data according to any one of claims 1 to 7.

11. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the method for processing power load data according to any one of claims 1 to 7.

Images