Method for constructing power consumption evaluation model of agent power purchase user based on fuzzy weight matrix

Abstract

The present application relates to the technical field of electricity evaluation model, and particularly relates to a proxy electricity purchasing user electricity evaluation model construction method based on a fuzzy weight matrix, which comprises the following steps: obtaining historical electricity consumption; performing pre-interpolation processing on the historical electricity consumption to obtain electricity consumption data at an interpolation time and a collection time; obtaining the electricity stability degree of each user according to the correlation between all historical electricity consumptions of each user in each day and all historical electricity consumptions in other adjacent days, and the difference distribution of the historical electricity consumptions; obtaining the data reliability degree according to the distribution of the interpolation time and the abnormality of the electricity consumption data of each user at the interpolation time, in combination with the electricity stability degree; and correcting the electricity consumption data by using the data reliability degree in combination with the electricity consumption data of each user and other users at the interpolation time, to obtain a corrected interpolation result and construct an electricity evaluation model. The present application makes the final electricity evaluation model more accurate.

Description

Technical Field

[0001] This invention relates to the field of electricity consumption evaluation model technology, specifically to a method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix. Background Technology

[0002] By comprehensively considering multiple dimensions of data indicators from different users' historical electricity consumption data, such as electricity consumption, peak-valley difference, and power, an objective evaluation of each user's electricity consumption can be conducted. This helps the power supply formula to more accurately understand users' electricity consumption behavior, thereby enabling the formulation of more reasonable electricity consumption strategies. Currently, the weight of each dimension is often determined by the fuzzy entropy of data indicators from multiple dimensions in different users' historical electricity consumption data, in order to improve the accuracy of the TOPSIS algorithm's evaluation results of users' electricity consumption and obtain a more accurate electricity consumption evaluation model.

[0003] However, when acquiring data metrics for each user across various dimensions, historical electricity consumption data may be lost or damaged due to objective factors such as environmental conditions, system or equipment failures, or abnormalities in the data transmission process. This makes it impossible to obtain complete historical electricity consumption data for users. When using partially missing data to acquire data metrics for various dimensions, the weight estimation may have some bias, resulting in low accuracy of the final evaluation results.

[0004] Currently, interpolation algorithms are commonly used to fit missing data. However, when the time series of missing electricity consumption data for some users is long, data interpolation based solely on time series variation features cannot guarantee the accuracy of the interpolation results. Consequently, the accuracy of user electricity consumption evaluation results is low, and the accuracy of the electricity consumption evaluation model is also low. Summary of the Invention

[0005] To address the technical problem that the accuracy of interpolation results cannot be guaranteed, leading to low accuracy in user electricity consumption evaluation results and the electricity consumption evaluation model itself, the present invention aims to provide a method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix. The specific technical solution adopted is as follows:

[0006] Obtain the historical electricity consumption data for each user at each collection time of each day; perform pre-interpolation processing on the historical electricity consumption data for each user to obtain the electricity consumption data for each user at the interpolation time and the collection time of each day;

[0007] The electricity stability of each user is determined by the correlation between all historical electricity consumption in each day and all historical electricity consumption in other adjacent days, as well as the distribution of differences among all historical electricity consumption in each day.

[0008] Based on the distribution of interpolation times for each user's daily interpolation time within the time range of collection times at the same time on other days, as well as the abnormalities in the electricity consumption data at each user's daily interpolation time and the abnormalities in the electricity consumption data at the same time on other days, combined with the electricity consumption stability, the data reliability of each user's daily interpolation time is obtained.

[0009] By combining the reliability of the data with the distribution of each user's electricity consumption data at the interpolation time and the electricity consumption data of other users at the interpolation time, the electricity consumption data of each user at the interpolation time each day is corrected to obtain the corrected interpolation result.

[0010] An electricity consumption evaluation model is constructed based on the corrected interpolation results.

[0011] Preferably, the method of determining the data reliability of each user's daily interpolation time based on the distribution of interpolation times within the time range of collection times at the same time on other days, as well as the anomalies in the electricity consumption data at each user's daily interpolation time and the electricity consumption data at the same time on other days, combined with the electricity consumption stability, specifically includes:

[0012] For any user at any interpolation time on any day, the collection times of other days that are at the same time as the interpolation time are used as the reference times for the interpolation time. Based on the proportion of data containing the interpolation time in the time neighborhood of each reference time, the estimated weight of each reference time is determined.

[0013] For any interpolation time or acquisition time, the outlier degree of the data at the interpolation time or acquisition time is determined based on the difference between the electricity consumption data at the interpolation time or acquisition time and the electricity consumption data in the time neighborhood of the interpolation time or acquisition time.

[0014] Based on the difference distribution of data outlier degree between the interpolation time and each control time, the data correlation coefficient between each control time and the interpolation time is determined;

[0015] The estimated weights of all reference times corresponding to the interpolation time are fused with the corresponding data correlation coefficients, and combined with the power consumption stability of the same user, the data reliability of the user at the interpolation time is determined.

[0016] Preferably, the method for obtaining the degree of data outlier is as follows:

[0017] For any user at any interpolation time, the electricity consumption data of all times within the preset time neighborhood of the interpolation time are used to form the reference dataset for the interpolation time; based on the standardized deviation of the electricity consumption data at the interpolation time relative to the reference dataset, the outlier degree of the data at the interpolation time is determined.

[0018] Preferably, determining the data correlation coefficient between each control time and the interpolation time based on the difference distribution of data outliers between the interpolation time and each control time specifically includes:

[0019] The difference in outlier levels between the data at each control time and interpolation time was negatively correlated to obtain the correlation coefficient between the data at each control time and interpolation time.

[0020] Preferably, determining the estimation weight of each control time based on the proportion of data containing the interpolation time within the time neighborhood of each control time specifically includes:

[0021] For any user at any given time, the negative correlation coefficient of the proportion of interpolated times within the preset time neighborhood of the time at which the time at ... is used as the estimated weight of the time at the time at the time at the time.

[0022] Preferably, the step of using the reliability of the data, combined with the distribution of each user's electricity consumption data at the interpolation time and the electricity consumption data of other users at the interpolation time, to correct the daily electricity consumption data of each user at the interpolation time, and obtaining the corrected interpolation result, specifically includes:

[0023] For any user at any interpolation time on any day, the first correction parameter for that user at the interpolation time is determined based on the data reliability and electricity consumption data corresponding to that user at the interpolation time.

[0024] Based on the distribution of electricity consumption data of this user and other users at the interpolation time and in the time neighborhood of the interpolation time, the second correction parameter of this user at the interpolation time is obtained;

[0025] The corrected data for the user at the interpolation time is obtained by combining the first and second correction parameters; the corrected interpolation result includes the power consumption data of each user at the collection time and the corrected data at the interpolation time.

[0026] Preferably, obtaining the second correction parameter for the user at the interpolation time based on the distribution of electricity consumption data between the user and other users at the interpolation time and within the time neighborhood of the interpolation time specifically includes:

[0027] Obtain the similarity coefficient of the electricity consumption data distribution between the user and the reference dataset of each other at the interpolation time; using the negative correlation coefficient of the data reliability of the user at the interpolation time, fuse the similarity coefficient of the user and the other users at the interpolation time with the product of the electricity consumption data of each other users at the interpolation time to obtain the second correction parameter of the user at the interpolation time.

[0028] Preferably, the step of determining the electricity consumption stability of each user based on the correlation between all historical electricity consumption within a day and all historical electricity consumption on other adjacent days, and the distribution of differences among all historical electricity consumption within a day, specifically includes:

[0029] For any user, all historical electricity consumption for each day is used to form the daily electricity consumption time sequence of the user; a preset number of days adjacent to any given day are recorded as the neighboring days of that given day.

[0030] Based on the degree of dispersion of the first-order difference data of historical electricity consumption in the daily electricity consumption time series of a user, the data stability coefficient of the user is determined for each day, and the degree of dispersion is negatively correlated with the data stability coefficient.

[0031] Based on the fusion results of the correlation coefficient between the user's daily electricity consumption time series and the electricity consumption time series of each neighboring day, the adjacent variation factors of the user's daily life are determined.

[0032] The user's electricity consumption stability is obtained by combining the data stability coefficients for all days and adjacent variation factors.

[0033] Preferably, the step of constructing an electricity consumption evaluation model based on the corrected interpolation results specifically includes:

[0034] Based on the corrected interpolation results for each user, feature data for each user in each dimension is obtained; based on the fuzzy entropy of the feature data for each user in each dimension, the fuzzy weight for each user in each dimension is determined; and the fuzzy weights are used to evaluate the electricity consumption data of each user.

[0035] Preferably, the step of pre-interpolating the historical electricity consumption of each user to obtain the electricity consumption data of each user at the interpolation time and the collection time for each day specifically includes: using the cubic spline interpolation method to pre-interpolate the historical electricity consumption of each user to obtain the electricity consumption data of each user at the interpolation time and the collection time for each day.

[0036] The embodiments of the present invention have at least the following beneficial effects:

[0037] This invention first collects users' historical electricity consumption data and uses pre-interpolation processing to provide a data foundation for subsequent analysis of the reliability of the interpolation results. The interpolation time represents the moment when data is missing, and the collection time represents the moment when data is not missing. Then, by analyzing the distribution of historical electricity consumption actually collected for each user, and fully considering the correlation between each user's daily historical electricity consumption and adjacent days, as well as the daily difference distribution, the stability of each user's electricity consumption is quantified. This involves a holistic analysis of the user's actual data stability from two aspects, fully considering the similarity of data change trends within a local time range each day. Furthermore, considering the stability of the actually collected data, the reliability of the interpolation results should also be considered. Therefore, in conjunction with the degree of electricity consumption stability, two aspects of data distribution characteristics are considered: first, the distribution of the difference moments at the same time on other days compared to the interpolation time, reflecting the situation of missing data at the same time on other days; second, the data anomalies at each interpolation time compared to other days. The more abnormal the data, the lower the reliability of the interpolation results, and the data reliability reflects the reliability of the electricity consumption data at the interpolation time. Finally, based on the level of credibility, i.e. the reliability of the data, and by referring to the interpolation results of other users, a more comprehensive and accurate corrected interpolation result can be obtained. This avoids the inaccuracy of only considering changes in time series data, thus making the final electricity consumption evaluation model more accurate. Attached Figure Description

[0038] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of the steps in constructing a power consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix, provided by the present invention.

[0040] Figure 2 This is a schematic diagram of the curve showing how a user's electricity consumption data changes over time, provided by the present invention.

[0041] Figure 3 This is a schematic diagram of the data distribution of users' historical electricity consumption after interpolation processing, provided by the present invention;

[0042] Figure 4 This is a flowchart of the steps of the method for obtaining the power stability level provided by the present invention;

[0043] Figure 5This is a flowchart of the steps in the method for obtaining data reliability provided by the present invention;

[0044] Figure 6 This is a block diagram of a system for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix, provided by the present invention. Detailed Implementation

[0045] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix, proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0047] The following description, in conjunction with the accompanying drawings, details the specific scheme of the method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix, provided by this invention.

[0048] Please see Figure 1 The diagram illustrates a flowchart of a method for constructing an electricity consumption evaluation model for proxy electricity purchasers based on a fuzzy weight matrix, according to an embodiment of the present invention. The method includes the following steps:

[0049] Step S100: Obtain the historical electricity consumption of each user at each collection time of each day from the historical electricity consumption data; perform pre-interpolation processing on the historical electricity consumption of each user to obtain the electricity consumption data of each user at the interpolation time and the collection time of each day.

[0050] The main objective of this invention is to obtain the historical electricity consumption data distribution of each agent electricity purchaser by analyzing the changes in electricity meters of each user, and to perform interpolation processing on the missing data to obtain accurate multi-dimensional electricity consumption index data. This enables an accurate comprehensive evaluation of the electricity consumption of each agent electricity purchaser, thereby constructing a highly accurate evaluation model.

[0051] Considering that existing interpolation algorithms rely solely on temporal data variation characteristics, the results of interpolation processing for data with significant missing information are difficult to guarantee, this embodiment performs feature analysis on the interpolation results after interpolation processing and then corrects them to obtain more accurate interpolation results.

[0052] Therefore, in order to provide feature analysis for the interpolation results, it is necessary to first collect the user's historical electricity consumption data and perform preliminary pre-interpolation processing on the historical electricity consumption data. Specifically, in this embodiment, for any agent electricity purchase user, the electricity consumption data for each day of the past year at each collection time is collected from the historical electricity consumption data. Under the condition that no data is missing, the time interval between two adjacent collection times is 10 minutes, which can be set by the implementer according to the specific implementation scenario. For example... Figure 2 The figure shown is a schematic diagram illustrating the change in electricity consumption data of some users over time.

[0053] Furthermore, the cubic spline interpolation method pre-interpolates the historical electricity consumption of each user, obtaining the interpolation fitting results for each user. This includes the electricity consumption data at the interpolation time and the electricity consumption data at the collection time for each user each day. For example... Figure 3 The figure shows a schematic diagram of the distribution of a user's historical electricity consumption after interpolation.

[0054] Step S200: Based on the correlation between each user's total historical electricity consumption within a day and the total historical electricity consumption of other adjacent days, and the distribution of differences among all historical electricity consumptions within a day, the electricity consumption stability of each user is obtained.

[0055] It is understandable that each user's historical electricity consumption represents their actual historical electricity usage, while each user's electricity consumption data represents the fitted electricity usage after interpolation. When processing the raw electricity consumption data using interpolation algorithms, the data is often obtained by fitting the trend of actual historical electricity consumption over time. The reliability of the interpolation results largely depends on the orientation of historical electricity consumption changes over time. That is, even if there are many missing data points in the actual historical electricity consumption, if the trend of change over time is relatively stable, the accuracy of prediction based on this missing data is still relatively high. Therefore, when analyzing the distribution stability of each user's historical electricity consumption, the correlation of historical electricity consumption distribution within adjacent time zones each day, as well as the difference distribution of historical electricity consumption each day, are combined to quantify the user's electricity consumption stability. By fully combining the data distribution stability of local time ranges and daily data ranges, the stability evaluation results are more accurate. Figure 4 As shown, the power stability of each user can be achieved through steps S201 to S203.

[0056] Step S201: For any user, construct a daily electricity consumption time sequence of all historical electricity consumption of the user; and record a preset number of days adjacent to any given day as the neighboring days of the given day.

[0057] Throughout the year, users exhibit different electricity consumption habits at different times of day. Within a similar number of days each day, the historical electricity consumption of the same user shows a relatively similar distribution trend. Therefore, when analyzing the data stability of each user's historical electricity consumption, analysis can be based on the similarity of data change trends within a local time frame.

[0058] In this embodiment, for day i, a preset number of days are obtained before and after day i as neighboring days of day i. In this embodiment, the preset number is 14, that is, day i-7 to day i-1, and day i+1 to day i+7 are all neighboring days of day i.

[0059] It should be noted that the neighboring days cannot be obtained using the above method during the first seven days and the last seven days for each user. Specifically, for any day within the first seven days, the 14 days immediately following that day will be considered the neighboring days. For example, the neighboring days for the first day are the second to the 15th day. For any day within the last seven days, the 14 days immediately preceding that day will be considered the neighboring days. For example, the neighboring days for the 364th day are the 349th to the 363rd day.

[0060] The corresponding neighboring day for each day represents the local time range of that day. Meanwhile, the historical electricity consumption of each user at all collection points throughout the day constitutes the user's electricity consumption time-series sequence for that day.

[0061] Step S202: Based on the dispersion of the first-order difference data of historical electricity consumption in the user's daily electricity consumption time series, determine the user's daily data stability coefficient. The dispersion is negatively correlated with the data stability coefficient.

[0062] First-order difference data is obtained by performing first-order differencing on the daily electricity consumption time-series sequence for each user, which can be represented as X. r,i (t+1)-X r,i (t), X r,i (t+1) represents the historical electricity consumption of the r-th user at the (t+1)-th data collection time on day i, X r,i (t) represents the historical electricity consumption of the r-th user at the t-th data collection time on day i. The first-order difference data reflects the degree of data change between adjacent data collection times in the electricity consumption time series.

[0063] Furthermore, the variance of all first-order difference data in the electricity consumption time-series sequence of the r-th user on day i is calculated to characterize the dispersion of the first-order difference data. That is, the larger the variance value, the greater the dispersion and volatility of the first-order difference data in the electricity consumption time-series sequence, and the smaller the corresponding data stability coefficient, indicating that the historical electricity consumption trend on day i of the r-th user is more unstable. This embodiment uses the reciprocal form to represent the negative correlation between the variance of the first-order difference data and the data stability coefficient. To prevent the denominator from being zero, a hyperparameter with a very small positive value, such as 0.1 or 0.01, is set in the denominator. The final data stability coefficient can be expressed as... Among them, S r,i ε0 represents the degree of dispersion of the first-order difference data of the electricity consumption time series sequence of the r-th user on the i-th day, and ε0 is a preset hyperparameter.

[0064] In other embodiments, standard deviation, interquartile range, and coefficient of variation can be used to characterize the dispersion of the data, where the coefficient of variation is the ratio of the standard deviation to the mean. Additionally, other negative correlations can be used to obtain the data stability coefficient, such as the form of negative exponentiation.

[0065] The data stability coefficient for each user per day represents the degree of stability of the data changes in each user's actual daily electricity consumption.

[0066] Step S203: Based on the fusion result of the correlation coefficient between the user's daily electricity consumption time series and the electricity consumption time series of each neighboring day, determine the user's daily adjacent variation factor; combine the user's data stability coefficient and adjacent variation factor for all days to obtain the user's electricity consumption stability.

[0067] For the r-th user on day i, a correlation coefficient can be obtained between the electricity consumption time series of each neighboring day and day i, which is used to characterize the similarity between the electricity consumption time series of the neighboring days and day i. Considering the differences in the missing actual electricity consumption data between each neighboring day and day i, resulting in different data quantities between the electricity consumption time series, this embodiment obtains the correlation coefficient by negatively correlating the DTW distance between each neighboring day and day i. The negative correlation processing can be in the form of reciprocal, negative exponentiation, etc.

[0068] In other embodiments, if the number of data points in the electricity consumption time series of the neighboring days is equal to that of the i-th day, the correlation coefficient can be obtained by calculating cosine similarity, which is used to characterize the data similarity between different days.

[0069] By integrating the correlation coefficients of all neighboring days of the r-th user with those of day i, the adjacent change factor of the r-th user on day i can be obtained. In this embodiment, the mean of the correlation coefficients of all neighboring days of the r-th user with those of day i is used as the adjacent change factor. The adjacent change factor reflects the overall level of similarity of the change trend of the actual electricity consumption of the r-th user within a local time range on day i. The higher the similarity of the change trend, the more stable the change trend of the actual electricity consumption of the r-th user on day i.

[0070] In other embodiments, the median may also be used to characterize the overall level of similarity in the changing trends of the actual electricity consumption of the r-th user within a local time range on day i.

[0071] Finally, considering the two characteristics of the actual daily electricity consumption of the r-th user, a global evaluation of the electricity consumption stability of the r-th user is performed. This evaluation can be achieved by combining summation or product. As a specific example, the formula for calculating the electricity consumption stability of the r-th user can be expressed as:

[0072]

[0073] Among them, W r This represents the power stability of the r-th user, where N represents the total number of days. i ρ represents the total number of neighboring days of day i. r (i,x) represents the correlation coefficient between the electricity consumption time series of day i and the x-th neighboring day, S r,i ε0 represents the degree of dispersion of the first-order difference data of the electricity consumption time series sequence of the r-th user on the i-th day, and ε0 is a preset hyperparameter.

[0074] Let be the data stability coefficient for the r-th user on day i. Let be the adjacent variation factor for the r-th user on day i. A larger value indicates more stable changes in the actual electricity consumption data for the r-th user on day i, suggesting more similar trends in data changes within a local timeframe of day i, and thus a more stable actual electricity consumption on day i. Considering the actual electricity consumption of the r-th user each day, the electricity consumption stability level characterizes the overall electricity consumption stability of the r-th user within the collected historical data. A larger value indicates a more stable trend in the actual electricity consumption data of the r-th user, leading to more reliable data prediction results based on time series analysis.

[0075] Step S300: Based on the distribution of interpolation times for each user's daily interpolation times within the time range of collection times at the same time on other days, as well as the abnormalities in the electricity consumption data of each user's daily interpolation times and the abnormalities in the electricity consumption data of collection times at the same time on other days, and in combination with the electricity consumption stability, the reliability of the data for each interpolation time of each user's day is obtained.

[0076] Step S200 analyzes the stability of data based on each user's actual electricity consumption. However, simply analyzing data stability is insufficient if there is significant data gaps in the user's actual electricity consumption data; otherwise, the accuracy of the data prediction results will be relatively low. In other words, the reliability of the interpolation results does not solely depend on the stability of the actual data. Therefore, it is necessary to further analyze the reliability of the interpolation results based on the distribution of the fitted data after interpolation.

[0077] Considering that interpolation fitting is based on the changing trends of known data points, if the local variation of the known data points used for interpolation is large, the resulting interpolated data may also fluctuate significantly, leading to low reliability of the interpolation results. Therefore, considering the data distribution of the interpolation results in multiple aspects and quantifying the reliability of the data at each interpolation time point can comprehensively reflect the reliability of the interpolation data results at each interpolation time point, providing a data foundation for subsequent adaptive adjustments to the interpolation results.

[0078] Based on this, the method for obtaining the reliability of data at each interpolation time point for each user per day can be as follows: Figure 5 Steps S301 to S303 shown are implemented.

[0079] Step S301: For any user at any interpolation time on any day, the collection times of other days that are at the same time as the interpolation time are taken as the reference times of the interpolation time. Based on the proportion of data containing the interpolation time in the time neighborhood of each reference time, the estimated weight of each reference time is determined.

[0080] Considering that the electricity consumption data at each interpolation point is obtained by fitting the changing trend of time series data, the data at each interpolation point fluctuates greatly compared to other identical points, and the possibility of anomalies is greater. In other words, the actual electricity consumption data fluctuates greatly, which may lead to large fluctuations in the electricity consumption data obtained by fitting at the interpolation point, and thus the reliability of the data is lower.

[0081] Based on this, this embodiment analyzes the distribution of interpolation times for each user's daily interpolation time within a certain time range at the same time period on other days, to reflect the degree of data gaps in the actual electricity consumption collection times. Specifically, taking the k-th interpolation time of the r-th user on the i-th day as an example, the collection times of the r-th user at the same time as the k-th interpolation time on all other days except the i-th day are the corresponding control times for the k-th interpolation time of the r-th user on the i-th day.

[0082] For example, the k-th interpolation time on day i is 8:00 on day i. Then, the collection time corresponding to 8:00 on each of the other days (excluding day i) is recorded as the control time. It should be noted that the collection time refers to the actual time when data is collected, that is, the time when no data is missing. The interpolation time, on the other hand, refers to the time corresponding to the data after interpolation processing, that is, the time when data is missing. Therefore, all control times are times when no data is missing.

[0083] Meanwhile, considering that electricity consumption data closer in time is more valuable for reference, this embodiment still selects a preset number of days adjacent to day i for data analysis. The method for obtaining the preset number of days adjacent to day i can be found in step S201.

[0084] Furthermore, for any user and any reference time, the negative correlation coefficient of the proportion of interpolated times within the preset time neighborhood of the reference time is used as the estimated weight of the reference time. In this embodiment, each reference time is taken as the center, and all times included in the preset time length constitute the time neighborhood of each reference time. It can be understood that the time neighborhood may contain both the acquisition time and the interpolation time. The preset length is 11, which can be set by the implementer according to the specific implementation scenario. It should be noted that for reference times whose time neighborhood cannot be obtained, the method for obtaining the neighborhood days in step S201 can be referred to.

[0085] The more interpolated times are included in the neighborhood of the reference time, the smaller the corresponding estimated weight value, indicating a greater degree of data gaps within the local time range of the reference time, and consequently, a lower reference value for the electricity consumption data at that reference time. The estimated weight of each reference time corresponding to each user's interpolated time reflects the magnitude of the reference value of the reference time in terms of the local proportion of missing data.

[0086] Step S302: For any interpolation time or acquisition time, determine the outlier degree of the data at the interpolation time or acquisition time based on the difference between the electricity consumption data at the interpolation time or acquisition time and the electricity consumption data in the time neighborhood of the interpolation time or acquisition time.

[0087] The estimated weights only consider the reference value of actual electricity consumption data. Therefore, it is necessary to analyze the normality or abnormality of the actual electricity consumption data or the interpolation fit to be evaluated. That is, by analyzing the difference distribution between the electricity consumption data at each moment and the electricity consumption data within a local time range, the outlier degree of the data at each moment can be quantified, which can more accurately reflect the data anomalies at each moment.

[0088] In this embodiment, the method for obtaining the outlier degree of data at the interpolation time is used as an example for explanation. The method for obtaining the outlier degree of data at the acquisition time is the same as that for the interpolation time, and the reference time is the acquisition time. For any user at any interpolation time, the electricity consumption data of all times within a preset time neighborhood of the interpolation time are used to form a reference dataset for the interpolation time. Based on the standardized deviation of the electricity consumption data at the interpolation time relative to the reference dataset, the outlier degree of data at the interpolation time is determined.

[0089] The standardized deviation is calculated using a well-known technique, reflecting the deviation of each data point relative to the whole. The standardized deviation of the r-th user at the k-th interpolation time on day i relative to the reference dataset can be expressed as:

[0090]

[0091] Among them, L r,i (k) represents the standardized deviation of the k-th interpolation time of the r-th user on day i relative to the reference dataset, X' r,i (k) represents the electricity consumption data of the r-th user at the k-th interpolation time on the i-th day, μ r,i (k) represents the mean of all electricity consumption data in the reference dataset corresponding to the k-th interpolation time of the r-th user on the i-th day, σ r,i (k) represents the standard deviation of all electricity consumption data in the reference dataset corresponding to the k-th interpolation time of the i-th user on the i-th day.

[0092] L r,i (k) reflects the degree of deviation of the data from the overall distribution, that is, it characterizes the outlier situation of the electricity consumption data at the interpolation time relative to the overall data within the local time range. Therefore, L r,i (k) represents the outlier rate of the data at the k-th interpolation time on day i for user r. The larger the outlier rate value, the greater the probability of anomalies in the electricity consumption data at the k-th interpolation time on day i for user r, and the greater the deviation from the overall data.

[0093] Step S303: Based on the difference distribution of data outliers between the interpolation time and each control time, determine the data correlation coefficient between each control time and the interpolation time. Merge the estimated weights of all control times corresponding to the interpolation time with the corresponding data correlation coefficients, and combine this with the power consumption stability of the same user to determine the data reliability of that user at the interpolation time.

[0094] The outlier rate at each time point reflects the likelihood of data anomalies and deviations from the overall distribution. It also reflects the changing trend of electricity consumption data at each time point. By analyzing the difference distribution between the outlier rate at the interpolation time and each control time, we can reflect the similarity and difference in the data changing trends between each control time and the interpolation time. The greater the difference, the lower the reference value of the data at the control time, which in turn makes the reliability of the interpolation fitting results at the corresponding interpolation time lower.

[0095] Specifically, for the r-th user, a negative correlation is applied to the difference in outlier levels between the data at each control time and the interpolation time to obtain the data correlation coefficient between each control time and the interpolation time. The difference is represented by the absolute value of the interpolation between the data points, and the negative correlation can be performed using the reciprocal or negative exponent form.

[0096] Furthermore, the estimated weights characterize the reference value of the data at the control time, while the data correlation coefficient characterizes the similarity of the data change trends between the control time and the interpolation time. The greater the reference value of the control time, the greater the corresponding similarity, which in turn indicates that the interpolation fitting result at that interpolation time is more reliable. At the same time, the greater the stability of the user's data, the more reliable the interpolation fitting result. Therefore, the reliability of the interpolation result at the interpolation time can be evaluated by integrating the feature distribution of these three aspects.

[0097] In this embodiment, taking any interpolation time of any user on any day as an example, the reliability of the data at the k-th interpolation time of the r-th user on the i-th day can be expressed by the formula:

[0098]

[0099] Among them, Q r,i (k) represents the data reliability of the r-th user at the k-th interpolation time on day i, Q r N represents the power stability of the r-th user. r,i (k) represents the total number of control times corresponding to the k-th interpolation time on the i-th day for the r-th user. This represents the percentage of the time neighborhood containing the interpolated time corresponding to the k-th interpolated time on day i for the r-th user. To estimate the weights, L r,i (k) represents the outlier degree of the data for the r-th user at the k-th interpolation time on day i, L r,i (k,m) represents the outlier degree of the data at the m-th control time corresponding to the k-th interpolation time on the i-th day for the r-th user, Norm() represents the linear normalization function, and exp() represents the indicator function with the natural constant e as the base.

[0100] exp(-|L r,i (k)-L r,i (k,m)|) represents the data correlation coefficient between the interpolation time and the m-th control time, reflecting the similarity of outliers and the similarity of data trends between the two times. A larger correlation coefficient indicates greater similarity between the data at the interpolation time and the reference data used for interpolation. Conversely, a larger estimated weight indicates greater reference value of the reference data. This embodiment fuses the data using a weighted summation method, resulting in the final fusion result. This reflects the reliability of the interpolation fitting results for the electricity consumption data at the interpolation point. A larger value in the fusion result indicates greater data stability for the user, signifying higher reliability of the corresponding interpolation process and a higher value for data reliability.

[0101] The reliability of data for each user at each interpolation time of the day takes into account the similarity between the distribution of electricity consumption data at the actual collection time and the distribution of data processed by interpolation, and also takes into account the user's data stability, thus comprehensively representing the reliability of the interpolation process at each interpolation time.

[0102] Step S400: Using the reliability of the data and the distribution of each user's electricity consumption data at the interpolation time and the electricity consumption data of other users at the interpolation time, the electricity consumption data of each user at the interpolation time for each day is corrected to obtain the corrected interpolation result.

[0103] Data reliability reflects the credibility of the interpolation results. The higher the data reliability at each user's interpolation time, the better the data fit during interpolation at that time, indicating higher reliability of the interpolated data points and thus a greater degree of retention of the interpolation result at that time. However, analyzing only a single user's electricity consumption is insufficient. Further comparison with data from other users with similar consumption habits can make the electricity consumption estimate for a single user more accurate. In other words, by combining the reliability evaluation of a single user's interpolation results with the similarity to other users, the interpolation results for that user can be corrected, resulting in more accurate interpolation results and avoiding the inaccuracy caused by simply considering data time-series changes.

[0104] Specifically, in this embodiment, the correction of the interpolation result at each user's interpolation time includes two aspects of correction results. First, the correction is made by evaluating the reliability of the interpolation result at each user's interpolation time. That is, for any user at any interpolation time on any day, the first correction parameter for the user at the interpolation time is determined based on the data reliability and electricity consumption data corresponding to the user at the interpolation time.

[0105] In this embodiment, the product of the reliability of the data of the r-th user at the k-th interpolation time on day i and the electricity consumption data at that interpolation time is used as the first correction parameter for the r-th user at the k-th interpolation time on day i. The larger the value of the data reliability, the greater the credibility of the interpolation result at the k-th interpolation time. The greater the reliability, the larger the corresponding correction result, which means a larger value for the first correction parameter.

[0106] Secondly, by analyzing the similarity of data distribution between other users and the r-th user, the interpolation result of the user at the interpolation time is corrected. That is, based on the distribution of electricity consumption data of the user and other users at the interpolation time and in the time neighborhood of the interpolation time, the second correction parameter of the user at the interpolation time is obtained.

[0107] Specifically, the similarity coefficient of the electricity consumption data distribution between the user and the reference dataset of each other at the interpolation time is obtained. Then, using the negative correlation coefficient of the user's data reliability at the interpolation time, the product of the similarity coefficient of the user and the electricity consumption data of each other at the interpolation time is fused to obtain the second correction parameter for the user at the interpolation time. Here, the reference dataset at the interpolation time represents the preset time neighborhood of the interpolation time, reflecting the local time range of the interpolation time.

[0108] For the r-th user, the electricity consumption data within the neighborhood of the k-th interpolation time on day i is used to construct the neighborhood data sequence of the r-th user at the k-th interpolation time on day i. Simultaneously, it is necessary to obtain the neighborhood data sequences of other users besides the r-th user at the same time as the k-th interpolation time on day i. In this embodiment, the DTW distance between the neighborhood data sequence of the r-th user at the k-th interpolation time on day i and the corresponding neighborhood data sequence of the nth other user at the same time is calculated, and a similarity coefficient is obtained by negatively correlating the DTW distance. The negative correlation can be processed using the reciprocal or negative exponent form.

[0109] The similarity coefficient reflects the similarity between the data distribution of the r-th user and the n-th user at the same interpolation time k. The greater the similarity, the greater the trend of data change among different users at the same time in the interpolation result, which in turn indicates a better interpolation result for a single user, and the corresponding corrected data value should be larger.

[0110] As a specific example, the method for obtaining the second correction parameter can be expressed as follows:

[0111] γ r,i (k)=[1-Q r,i (k)]×X″ r,i (k)

[0112]

[0113] Where, γ r,i (k) represents the second correction parameter at the k-th interpolation time on the i-th day for the r-th user, Q r,i (k) represents the data reliability of the r-th user at the k-th interpolation time on day i, X' r,i (k) represents the electricity consumption data of the r-th user at the k-th interpolation time on the i-th day, α r,n (i,k) represents the similarity coefficient between the r-th user and the n-th other user at the k-th interpolation time on day i, and M represents the total number of all users.

[0114] X” r,i (k) reflects the comprehensive evaluation result of the similarity of the interpolation results of the r-th user with the data of all other users at the same time. This embodiment uses the mean for fusion, utilizing 1-Q... r,i (k) is used as a weight for the fusion result X” r,i (k) Perform weighted processing. When the reliability of the interpolation result of the r-th user is low, more attention should be paid to the data similarity between the r-th user and other users. The greater the similarity, the greater the corresponding correction result.

[0115] Finally, considering the interpolation results of each user at the interpolation time by combining the correction results from both aspects, this embodiment combines them by summing, that is, the sum of the first correction parameter and the second correction parameter at the k-th interpolation time of the i-th day of the r-th user is used as the correction data for the k-th interpolation time of the i-th day of the r-th user.

[0116] A higher data reliability value at the interpolation time indicates greater credibility of the user's own interpolation result at that time, leading to greater attention to that result. Conversely, a lower data reliability value indicates lower credibility of the user's own interpolation result, leading to greater focus on the similarity of data trends between that user and other users at the same time point as the interpolation. By combining these two adjustments, the influence of time-series changes on the interpolated data is avoided, resulting in more accurate interpolation results.

[0117] Meanwhile, it is understood that the power consumption data at the interpolation time is data to supplement the missing parts. Therefore, this embodiment only corrects the power consumption data at the interpolation time by adaptively setting the reliability weight, while the power consumption data at other collection times are the data that is not missing and do not need to be interpolated and corrected.

[0118] Step S500: Construct an electricity consumption evaluation model based on the corrected interpolation results.

[0119] The corrected interpolation result means that the complete historical electricity consumption of each user has been obtained. Further, it is necessary to evaluate the electricity consumption of each user by using the complete historical electricity consumption of each user, and then build an evaluation model. There are various ways to evaluate the electricity consumption of users in the existing technology. This embodiment uses a simple evaluation method as an example for illustration.

[0120] Specifically, based on the corrected interpolation results for each user, feature data for each user in each dimension is obtained; based on the fuzzy entropy of the feature data for each user in each dimension, the fuzzy weight for each user in each dimension is determined; the fuzzy weights are used to evaluate the electricity consumption data of each user, and an electricity consumption evaluation model is constructed based on the evaluation results.

[0121] Each user's feature data for each dimension includes average electricity consumption, average peak-to-valley difference, and user data stability. Each dimension corresponds to a specific data type. The fuzzy entropy of the multi-dimensional feature data is calculated for different users. Based on the proportion of the obtained fuzzy entropy, the fuzzy weights corresponding to each dimension are obtained, and the feature data for each user is then positively processed. The processed feature data for each dimension, along with the corresponding fuzzy weights, are input into TOPSIS to obtain the electricity consumption evaluation results for each user.

[0122] In summary, this invention performs pre-interpolation processing on the electricity consumption data of multiple users with missing data. Based on the pre-interpolation results, the stability of electricity consumption changes for each different user is calculated, providing an initial assessment of the stability of the interpolation results for each user. Furthermore, by combining the similarity of the outlier degree of a single interpolation point with the data from its adjacent days, the reliability of each interpolation point is quantified to characterize the reliability of the time series estimation. Furthermore, by incorporating the electricity consumption data of multiple users with similar habits, the results of predictions based solely on time series data are corrected, further improving the accuracy of the interpolation results in predicting real data. Finally, by quantifying the evaluation indicators and the fuzzy weights of multiple evaluation indicators based on the obtained interpolation results, the reliability of subsequent evaluations of different users is improved.

[0123] like Figure 6 The diagram shows a block diagram of a system for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix, including:

[0124] The data preprocessing module is used to obtain the historical electricity consumption of each user at each collection time of each day from the historical electricity consumption data; and to perform pre-interpolation processing on the historical electricity consumption of each user to obtain the electricity consumption data of each user at the interpolation time and the collection time of each day.

[0125] The stability analysis module is used to determine the electricity stability of each user based on the correlation between all historical electricity consumption within a day and all historical electricity consumption on other adjacent days, as well as the distribution of differences among all historical electricity consumption within a day.

[0126] The reliability analysis module is used to determine the data reliability of each user's daily interpolation time based on the distribution of interpolation times within the time range of the same collection time on other days, as well as the abnormalities in the electricity consumption data of each user's daily interpolation time and the abnormalities in the electricity consumption data of the same collection time on other days, combined with the electricity consumption stability, to obtain the data reliability of each user's daily interpolation time.

[0127] The data correction module is used to correct the daily interpolation data of each user by combining the reliability of the data with the distribution of each user's electricity consumption data at the interpolation time and the electricity consumption data of other users at the interpolation time, so as to obtain the corrected interpolation result.

[0128] The model building module is used to build an electricity consumption evaluation model based on the corrected interpolation results.

[0129] In other embodiments, an apparatus is also provided, including a memory and a processor. The memory stores executable program code, and the processor calls and runs the executable program code from the memory, causing the apparatus to execute the aforementioned method for constructing an electricity consumption evaluation model for proxy electricity purchasers based on a fuzzy weight matrix. Specifically, the apparatus may be a chip, component, or module. The chip may include a connected processor and memory; wherein the memory stores instructions, and when the processor calls and executes the instructions, the chip can execute the aforementioned method for constructing an electricity consumption evaluation model for proxy electricity purchasers based on a fuzzy weight matrix.

[0130] In other embodiments, a computer program product is also provided, which, when run on a computer, causes the computer to perform the aforementioned related steps to implement the method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix provided in the above embodiments.

[0131] In other embodiments, a computer-readable storage medium is also provided, which stores computer program code. When the computer program code is run on a computer, the computer executes the above-described method steps to implement the method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix provided in the above embodiments.

[0132] The systems, devices, computer program products, and computer-readable storage media provided are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects of the corresponding methods provided above, and will not be repeated here.

[0133] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix, characterized in that, The method includes the following steps: Obtain the historical electricity consumption data for each user at each collection time of each day; perform pre-interpolation processing on the historical electricity consumption data for each user to obtain the electricity consumption data for each user at the interpolation time and the collection time of each day; The electricity stability of each user is determined by the correlation between all historical electricity consumption in each day and all historical electricity consumption in other adjacent days, as well as the distribution of differences among all historical electricity consumption in each day. Based on the distribution of interpolation times for each user's daily interpolation time within the time range of collection times at the same time on other days, as well as the abnormalities in the electricity consumption data at each user's daily interpolation time and the abnormalities in the electricity consumption data at the same time on other days, combined with the electricity consumption stability, the data reliability of each user's daily interpolation time is obtained. By combining the reliability of the data with the distribution of each user's electricity consumption data at the interpolation time and the electricity consumption data of other users at the interpolation time, the electricity consumption data of each user at the interpolation time each day is corrected to obtain the corrected interpolation result. A power consumption evaluation model is constructed based on the corrected interpolation results; The method for obtaining the reliability of the data specifically includes: For any user at any interpolation time on any day, the collection times of other days that are at the same time as the interpolation time are used as the reference times for the interpolation time. Based on the proportion of data containing the interpolation time in the time neighborhood of each reference time, the estimated weight of each reference time is determined. For any interpolation time or acquisition time, the outlier degree of the data at the interpolation time or acquisition time is determined based on the difference between the electricity consumption data at the interpolation time or acquisition time and the electricity consumption data in the time neighborhood of the interpolation time or acquisition time. Based on the difference distribution of data outlier degree between the interpolation time and each control time, the data correlation coefficient between each control time and the interpolation time is determined; The estimated weights of all reference times corresponding to the interpolation time are fused with the corresponding data correlation coefficients, and combined with the power consumption stability of the same user, the data reliability of the user at the interpolation time is determined.

2. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 1, characterized in that, The method for obtaining the outlier degree of the data is as follows: For any user at any interpolation time, the electricity consumption data of all times within the preset time neighborhood of the interpolation time are used to form the reference dataset for the interpolation time; based on the standardized deviation of the electricity consumption data at the interpolation time relative to the reference dataset, the outlier degree of the data at the interpolation time is determined.

3. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 1, characterized in that, The determination of the data correlation coefficient between each control time and the interpolation time based on the difference distribution of data outliers between the interpolation time and each control time specifically includes: The difference in outlier levels between the data at each control time and interpolation time was negatively correlated to obtain the correlation coefficient between the data at each control time and interpolation time.

4. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 1, characterized in that, The determination of the estimated weight for each control time based on the proportion of data containing the interpolation time within the time neighborhood of each control time specifically includes: For any user at any given time, the negative correlation coefficient of the proportion of interpolated times within the preset time neighborhood of the time at which the time at ... is used as the estimated weight of the 5. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 2, characterized in that, The method involves using the reliability of the data, combined with the distribution of each user's electricity consumption data at the interpolation time and the electricity consumption data of other users at the interpolation time, to correct the daily electricity consumption data of each user at the interpolation time, obtaining the corrected interpolation result. Specifically, this includes: For any user at any interpolation time on any day, the first correction parameter for that user at the interpolation time is determined based on the data reliability and electricity consumption data corresponding to that user at the interpolation time. Based on the distribution of electricity consumption data of this user and other users at the interpolation time and in the time neighborhood of the interpolation time, the second correction parameter of this user at the interpolation time is obtained; The corrected data for the user at the interpolation time is obtained by combining the first and second correction parameters; the corrected interpolation result includes the power consumption data of each user at the collection time and the corrected data at the interpolation time.

6. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 5, characterized in that, The second correction parameter for the user at the interpolation time is obtained based on the distribution of electricity consumption data between the user and other users at the interpolation time and within the time neighborhood of the interpolation time. Specifically, this includes: Obtain the similarity coefficient of the electricity consumption data distribution between the user and the reference dataset of each other at the interpolation time; using the negative correlation coefficient of the data reliability of the user at the interpolation time, fuse the similarity coefficient of the user and the other users at the interpolation time with the product of the electricity consumption data of each other users at the interpolation time to obtain the second correction parameter of the user at the interpolation time.

7. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 1, characterized in that, The method for determining the electricity consumption stability of each user is based on the correlation between all historical electricity consumption within a day and all historical electricity consumption on other adjacent days, as well as the distribution of differences among all historical electricity consumption within a day. Specifically, this includes: For any user, all historical electricity consumption for each day is used to form the daily electricity consumption time sequence of the user; a preset number of days adjacent to any given day are recorded as the neighboring days of that given day. Based on the degree of dispersion of the first-order difference data of historical electricity consumption in the daily electricity consumption time series of a user, the data stability coefficient of the user is determined for each day, and the degree of dispersion is negatively correlated with the data stability coefficient. Based on the fusion results of the correlation coefficient between the user's daily electricity consumption time series and the electricity consumption time series of each neighboring day, the adjacent variation factors of the user's daily life are determined. The user's electricity consumption stability is obtained by combining the data stability coefficients for all days and adjacent variation factors.

8. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 1, characterized in that, The construction of the electricity consumption evaluation model based on the corrected interpolation results specifically includes: Based on the corrected interpolation results for each user, feature data for each user in each dimension is obtained; based on the fuzzy entropy of the feature data for each user in each dimension, the fuzzy weight for each user in each dimension is determined; and the fuzzy weights are used to evaluate the electricity consumption data of each user.

9. The method for constructing an electricity consumption evaluation model for agent-purchased electricity users based on a fuzzy weight matrix according to claim 1, characterized in that, The step of pre-interpolating the historical electricity consumption of each user to obtain the electricity consumption data of each user at the interpolation time and the collection time for each day specifically includes: using the cubic spline interpolation method to pre-interpolate the historical electricity consumption of each user to obtain the electricity consumption data of each user at the interpolation time and the collection time for each day.

Images