Power load-based user clustering method and device, equipment and storage medium

Abstract

Embodiments of the present application disclose a user clustering method and device based on power load, equipment and a storage medium, the method comprising: obtaining power load data set of candidate users for indicating power consumption of the candidate users; performing outlier processing on the power load data set by using annual temperature characteristics and the power load data set of each candidate user, determining target users and target power load data set; performing eigenvalue analysis processing on the target users, the target power load data set and preset power consumption characteristic parameters, determining eigenvalues of the power consumption characteristic parameters corresponding to each target user; and performing clustering processing on each target user according to the eigenvalues, determining power consumption categories corresponding to each target user for reflecting power consumption habits of the target users. Through the above manner, the power consumption categories can be better classified, which helps the power grid company to better understand the load demand and power consumption characteristics of the users, and lays a foundation for individualized demand response scheme.

Description

Technical Field

[0001] This invention relates to the field of power load analysis technology, and in particular to a user clustering method, apparatus, equipment and storage medium based on power load. Background Technology

[0002] The power grid is currently undergoing a critical transformation from a centralized power generation, transmission, distribution, and consumption system to a distributed system. Consequently, the analysis of residential user loads at the end of the power grid has reached a new level. Residential user loads are characterized by high randomness, small individual loads, and complex load types. Accurate prediction of residential loads has always been a bottleneck problem for the power grid, making cluster studies of residential loads even more challenging.

[0003] In domestic and international research on residential user clustering, numerous studies have explored how to extract load characteristics from time-domain load curves (i.e., time-domain analysis) for cluster analysis. However, time-domain analysis is only applicable to situations where the load is relatively stable and has low randomness, which is not the case for residential electricity load. Furthermore, in the foreseeable future, residential electricity consumption behavior will become increasingly difficult to predict due to changes in various socio-economic environments. These factors all affect the accuracy of various time-domain analysis methods. Secondly, general clustering methods neglect other factors that significantly influence load besides time, such as ambient temperature. Especially when clustering residential users, cooling and heating are their primary electricity needs, requiring the temperature dimension to be considered in the clustering process for a more accurate understanding of residential user electricity consumption characteristics.

[0004] Therefore, there is still a lack of an effective means of classifying and analyzing users' electricity consumption characteristics. Summary of the Invention

[0005] The main objective of this invention is to provide a user clustering method, apparatus, device, and storage medium based on power load, which can solve the problem of the lack of an effective means for classifying and analyzing the electricity consumption characteristics of users in the prior art.

[0006] To achieve the above objectives, the first aspect of the present invention provides a user clustering method based on power load, the method comprising:

[0007] Obtain the power load dataset of candidate users, which is used to indicate the power consumption of candidate users;

[0008] By utilizing the annual temperature characteristics and the power load datasets of each candidate user, outlier processing is performed on the power load dataset to identify the target user and the target power load dataset of the target user.

[0009] Based on the target user, target power load dataset, and preset electricity consumption characteristic parameters, feature value analysis is performed to determine the feature value of the electricity consumption characteristic parameter corresponding to each target user;

[0010] Clustering is performed on each target user based on the feature values ​​to determine the power consumption category corresponding to each target user. The power consumption category is used to reflect the power consumption habits of the target users.

[0011] In one feasible implementation, the power load dataset includes the correspondence between candidate users and power load data, the power load data includes at least the grid-connected power generation and offline power consumption of the candidate users, and the annual temperature characteristics include annual temperature data; then, the step of using the annual temperature characteristics and the power load datasets of each candidate user to perform outlier processing on the power load dataset to determine the target user and the target power load dataset of the target user includes:

[0012] The total electricity consumption of each candidate user's house is determined based on the amount of electricity generated by the grid and the amount of electricity consumed offline for each candidate user.

[0013] Identify abnormal users whose total electricity consumption of the house has a negative value;

[0014] Delete the abnormal user and the corresponding power load dataset of the abnormal user to obtain the target user and the first power load dataset of the target user;

[0015] Determine the annual temperature data for the location of each target user in the first power load dataset;

[0016] Using time as the same variable, the annual electricity load data of each target user is correlated with the annual temperature data of the local area to determine the second electricity load dataset corresponding to the target user;

[0017] Determine the temperature load distribution characteristics corresponding to the second power load dataset. The temperature load distribution characteristics are used to reflect the proportion of each power load data point in the second power load dataset of each target user at each temperature.

[0018] The power load data below the first proportion and above the second proportion in the temperature load distribution characteristics of each target user are deleted to determine the target power load dataset for each target user. The first proportion and above the second proportion are used to indicate the proportion threshold of the target user's power load data under extreme temperatures throughout the year. The first proportion is less than the second proportion.

[0019] In one feasible implementation, the electricity consumption characteristic parameters include at least a heating temperature threshold, a cooling temperature threshold, a cooling season temperature gradient change rate, and a heating season temperature gradient change rate. The step of performing characteristic value analysis processing based on the target users, the target electricity load dataset, and the preset electricity consumption characteristic parameters to determine the characteristic value of the electricity consumption characteristic parameter corresponding to each target user includes:

[0020] The temperature data in the target power load dataset is divided using a preset temperature range width to determine the sub-power load dataset corresponding to each temperature range;

[0021] The target sub-power load dataset with fewer than a preset data point threshold is deleted to obtain the updated target power load dataset for each target user.

[0022] Determine the target data points corresponding to the third, fourth, and fifth proportions in the updated target power load dataset. The third, fourth, and fifth proportions increase sequentially, and all three proportions are greater than the first proportion, while the fifth proportion is less than the second proportion.

[0023] The target data points and a preset linear fitting algorithm for temperature segmentation are used to perform fitting processing to determine the first characteristic value of the heating temperature threshold, the second characteristic value of the cooling temperature threshold, the third characteristic value of the temperature gradient change rate during the heating season, and the fourth characteristic value of the temperature gradient change rate during the cooling season for each target user.

[0024] In one feasible implementation, the linear fitting algorithm for the temperature segmentation includes the following mathematical expression:

[0025]

[0026]

[0027] X h -x1≥5 (3)

[0028] X c -X h ≥5 (4)

[0029] x N -X c ≥5 (5)

[0030] 10≤X h ≤20 (6)

[0031] X h X c ∈Z(7)

[0032] In the formula, N represents the target data point (x n y n,i The total number of data points, where x represents temperature, y represents electricity load data, i∈{1,2,3} represents the third, fourth, and fifth percentages of each temperature point respectively, and j∈{1,2,3} represents the heating season, normal season, and cooling season of the temperature segment respectively; X h X is the heating temperature threshold. c This is the cooling temperature threshold, where x1 represents the target data point (x1, y1) when n is 1. 1,i The temperature of x N This represents the target data point (x) when n takes the value N. N y N,i ) temperature, b i，j ω is a constant term. 3，1 ω represents the rate of change of temperature gradient during the heating season. 3，3 Rate of change of temperature gradient during the cooling season.

[0033] In one feasible implementation, the electricity consumption characteristic parameters further include base load and active load, then the method further includes:

[0034] Using the target data points corresponding to the third proportion of each target user and the preset base load algorithm, the fifth characteristic value of the base load of each target user is determined;

[0035] The sixth characteristic value of the activity load of each target user is determined by using the fifth characteristic value of each target user, the target data points under the fifth proportion, and the preset activity load algorithm.

[0036] In one feasible implementation, the basic load algorithm includes the following mathematical expression:

[0037]

[0038] In the formula, h1(x n () represents the data point x corresponding to the third proportion of the target user i=1. n N represents the total number of target data points for the third proportion; BL represents the basic load.

[0039] In one feasible implementation, the activity load algorithm includes the following mathematical expression:

[0040]

[0041] In the formula, h3(x n The fifth percentage of target users, i=3, corresponds to the data point x. n N represents the total number of target data points for the fifth proportion; BL represents the basic load, and AL represents the active load.

[0042] In one feasible implementation, the electricity consumption characteristic parameters include at least a heating temperature threshold, a cooling temperature threshold, a cooling season temperature gradient change rate, a heating season temperature gradient change rate, a base load, and an active load. Then, the step of clustering each target user based on the characteristic values ​​to determine the electricity consumption category corresponding to each target user includes:

[0043] Define the pre-set target clustering requirements;

[0044] According to the data processing method corresponding to the target clustering requirements, the first characteristic value of the heating temperature threshold, the second characteristic value of the cooling temperature threshold, the third characteristic value of the temperature gradient change rate during the heating season, the fourth characteristic value of the temperature gradient change rate during the cooling season, the fifth characteristic value of the base load, and the sixth characteristic value of the active load of each target user are processed to determine the target characteristic value set, which includes the characteristic value set of each target user after data processing.

[0045] The target feature value set is input into a preset K-means clustering model to determine each initial clustering result;

[0046] Determine the silhouette coefficients of each initial clustering result;

[0047] The K value of the initial clustering result corresponding to the highest silhouette coefficient is taken as the optimal K value;

[0048] Cluster the target feature value set according to the optimal K value to determine the K power consumption categories corresponding to each target user output by the K-means clustering model.

[0049] To achieve the above objectives, a second aspect of the present invention provides a user clustering device based on power load, the device comprising:

[0050] Data acquisition module: used to acquire power load datasets of candidate users, which are used to indicate the power consumption of candidate users;

[0051] Data processing module: Used to process outliers in the power load dataset by utilizing the annual temperature characteristics and the power load datasets of each candidate user, and to determine the target user and the target power load dataset of the target user;

[0052] Feature analysis module: used to perform feature value analysis processing based on the target user, target power load dataset and preset power consumption feature parameters, and determine the feature value of the power consumption feature parameter corresponding to each target user;

[0053] User classification module: used to cluster each target user according to the feature value, and determine the power consumption category corresponding to each target user. The power consumption category is used to reflect the power consumption habits of the target user.

[0054] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0055] To achieve the above objectives, a fourth aspect of the present invention provides a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0056] The embodiments of the present invention have the following beneficial effects:

[0057] This invention provides a user clustering method based on electricity load. The method includes: acquiring an electricity load dataset of candidate users, which indicates the electricity consumption of the candidate users; processing outliers in the electricity load dataset using year-round temperature characteristics and the electricity load datasets of each candidate user to determine target users and their target electricity load datasets; performing feature value analysis based on the target users, target electricity load datasets, and preset electricity consumption characteristic parameters to determine the feature values ​​of the electricity consumption characteristic parameters corresponding to each target user; and clustering each target user based on the feature values ​​to determine the corresponding electricity consumption category, which reflects the electricity consumption habits of the target users. By incorporating temperature as a dimension, this user clustering method can better classify users' electricity consumption categories, helping power grid companies better understand users' load demands and electricity consumption characteristics, and laying the foundation for designing personalized demand response solutions for residential users. Attached Figure Description

[0058] To more clearly illustrate the technical solutions 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.

[0059] in:

[0060] Figure 1 This is a flowchart of a user clustering method based on power load in an embodiment of the present invention;

[0061] Figure 2This is another flowchart of a user clustering method based on power load in an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram of the data distribution of various data points in a power load dataset according to an embodiment of the present invention;

[0063] Figure 4 This is a structural block diagram of a user clustering device based on power load in an embodiment of the present invention;

[0064] Figure 5 This is a structural block diagram of a computer device in an embodiment of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Please see Figure 1 , Figure 1 This is a flowchart of a user clustering method based on power load in an embodiment of the present invention, as shown below. Figure 1 The method shown includes the following steps:

[0067] 101. Obtain the power load dataset of candidate users, wherein the power load dataset is used to indicate the power consumption of candidate users;

[0068] It should be noted that the user clustering method based on power load shown in this application is for classifying the electricity consumption habits of different electricity users. Electricity users can be residential users with daily electricity needs. In order to classify the electricity consumption habits of residential users, it is first necessary to obtain power load data that can indicate the electricity consumption of users. Specifically, a power load dataset of candidate users is obtained. The power load dataset is used to indicate the electricity consumption of candidate users. Candidate users can be residential users that need to be supplied by the power grid unit. The power load dataset corresponds one-to-one with each candidate user. Furthermore, the power load dataset includes, but is not limited to, the annual historical power load data of each candidate user. For example, the power load dataset includes the correspondence between candidate users and power load data. The power load data includes at least the grid-connected power generation and offline power consumption of candidate users, such as the hourly power consumption for a whole year.

[0069] 102. Utilize the annual temperature characteristics and the power load datasets of each candidate user to perform outlier processing on the power load dataset, and determine the target user and the target power load dataset of the target user.

[0070] Furthermore, the electricity load data of the aforementioned candidate users can be considered as data samples. After obtaining these data samples, outlier processing is required to remove abnormal data samples and obtain valid data samples. This filtering includes, but is not limited to, filtering out abnormal users and outlier data points within the electricity load dataset. Since residential electricity consumption habits are often related to daily climate and temperature—for example, air conditioning is used for cooling in high temperatures and heating is needed in low temperatures—after obtaining the electricity load dataset of the aforementioned candidate users, not only can outlier filtering be performed, but also a target electricity dataset related to temperature characteristics can be obtained by combining the annual temperature characteristics, including but not limited to annual temperature data. Specifically, outlier processing is performed on the electricity load dataset using the annual temperature characteristics and the electricity load datasets of each candidate user to determine the target users and their target electricity load datasets. This is equivalent to conducting preliminary data processing on residential electricity load data: this step mainly analyzes the anomalies in individual residential user data, eliminating abnormal customers or certain abnormal data points within those customers.

[0071] 103. Based on the target user, target power load dataset, and preset electricity consumption characteristic parameters, perform feature value analysis to determine the feature value of the electricity consumption characteristic parameter corresponding to each target user;

[0072] It should be noted that after obtaining the target electricity load dataset related to the annual temperature characteristics of each target user, the characteristic values ​​of the electricity consumption characteristic parameters of the target users can be determined through the target electricity load dataset to obtain the electricity consumption characteristics of each target user. That is, based on the target users, the target electricity load dataset, and the preset electricity consumption characteristic parameters, characteristic value analysis is performed to determine the characteristic values ​​of the electricity consumption characteristic parameters corresponding to each target user. For example, electricity consumption characteristic parameters include, but are not limited to, heating temperature threshold, cooling temperature threshold, heating season temperature gradient change rate, heating season temperature gradient change rate, base load, and active load, etc., which can reflect electricity consumption habits. This embodiment is only an example and is not specifically limited. That is, load decomposition is carried out to obtain characteristic values: this step aims to extract several parameters from the load distribution to represent the electricity consumption of each house by using the load decomposition method.

[0073] 104. Cluster each target user according to the feature value to determine the power consumption category corresponding to each target user. The power consumption category is used to reflect the power consumption habits of the target user.

[0074] Finally, by using the feature values ​​of the aforementioned electricity consumption characteristic parameters reflecting electricity consumption habits, the electricity consumption habits of each target user can be clearly identified, thereby improving the accuracy of electricity consumption category classification. Specifically, after obtaining the aforementioned feature values, clustering processing can be performed on each target user based on the feature values ​​to determine the corresponding electricity consumption category for each target user. The electricity consumption category is used to reflect the electricity consumption habits of the target user. Clustering methods include, but are not limited to, K-means and other clustering methods; this embodiment is used as an example and not specifically limited. The electricity consumption category reflects the electricity consumption habits of the target user; target users belonging to the same electricity consumption category have the same electricity consumption habits. That is, cluster analysis is performed to identify users with the same characteristics: this step uses a clustering algorithm to divide users into K clusters and analyzes the commonalities among users within the clusters.

[0075] Furthermore, after obtaining the above-mentioned power consumption categories, it is possible to predict and analyze users' future electricity consumption habits, and better allocate and arrange the work on the power grid side. This not only helps to save electricity but also contributes to the stability and security of power supply, enabling better power planning. In other words, effective analysis of users' electricity consumption habits has profound significance for the long-term development of the power grid, laying the foundation for designing personalized demand response solutions for residential users.

[0076] This invention provides a user clustering method based on electricity load. The method includes: acquiring an electricity load dataset of candidate users, which indicates the electricity consumption of the candidate users; processing outliers in the electricity load dataset using year-round temperature characteristics and the electricity load datasets of each candidate user to determine target users and their target electricity load datasets; performing feature value analysis based on the target users, target electricity load datasets, and preset electricity consumption characteristic parameters to determine the feature values ​​of the electricity consumption characteristic parameters corresponding to each target user; and clustering each target user based on the feature values ​​to determine the corresponding electricity consumption category, which reflects the electricity consumption habits of the target users. By incorporating temperature as a dimension, this user clustering method can better classify users' electricity consumption categories, helping power grid companies better understand users' load demands and electricity consumption characteristics, and laying the foundation for designing personalized demand response solutions for residential users.

[0077] Please see Figure 2 , Figure 2 This is another flowchart of a user clustering method based on power load in an embodiment of the present invention, as shown below. Figure 2 The methods shown include:

[0078] 201. Obtain the power load dataset of candidate users, wherein the power load dataset is used to indicate the power consumption of candidate users;

[0079] It should be noted that step 201 and Figure 1 The content of step 101 shown is similar, and will not be repeated here to avoid repetition. Please refer to the previous section for details. Figure 1 The content of step 101 shown.

[0080] It should be noted that, in order to better analyze the electricity consumption habits of residential users, the above-mentioned power load dataset may include the correspondence between candidate users and power load data. The power load data includes, but is not limited to, the grid-connected power generation and offline power consumption of candidate users, and the annual temperature characteristics include annual temperature data. Then, by using the annual temperature characteristics and the power load datasets of each candidate user to perform outlier processing on the power load dataset, the target user and the target power load dataset of the target user may include the following steps 202 to 208, for details.

[0081] 202. Determine the total electricity consumption of each candidate user's house based on the amount of electricity generated on the grid and the amount of electricity consumed off the grid.

[0082] The total electricity consumption of candidate users' homes can be calculated using the grid-connected power generation and grid-connected electricity consumption data. This calculation is performed for each candidate user. It should be noted that if there is no grid-connected power generation, this item can be considered as 0; if it exists, the total electricity consumption is calculated based on the actual data value. For example, for residential users with distributed power sources such as photovoltaic and wind turbine generators, the specific method for calculating the total electricity consumption is to add the absolute values ​​of their grid-connected power generation and grid-connected electricity consumption to obtain the total electricity consumption.

[0083] 203. Identify abnormal users whose total electricity consumption of the aforementioned house has a negative value;

[0084] 204. Delete the abnormal user and the power load dataset corresponding to the abnormal user to obtain the target user and the first power load dataset of the target user;

[0085] Furthermore, by using the positive or negative value of the total electricity consumption of the house, abnormal users that need to be deleted can be identified. The total electricity consumption of the house can reflect the electricity consumption of the user. Therefore, a candidate user with a negative total electricity consumption can be considered to have abnormal electricity consumption. The candidate user with a negative total electricity consumption is regarded as an abnormal user, and the corresponding electricity load dataset of the abnormal user is deleted. This process is used to update all the obtained electricity load datasets. Finally, the target users with valid data and the first electricity load dataset of the target users are obtained.

[0086] 205. Determine the annual temperature data of the location of each target user in the first power load dataset;

[0087] 206. Using time as the same variable, correlate the annual power load data of each target user with the annual temperature data of the location to determine the second power load dataset corresponding to the target user;

[0088] Furthermore, by using the annual temperature data based on year-round temperature characteristics, the annual temperature data for each target user's location can be determined. Using time as a constant variable, the annual electricity load data for each target user is correlated with the annual temperature data for their location to determine the second electricity load dataset corresponding to that target user. This second electricity load dataset includes both annual electricity load data and temperature data. (See reference for details.) Figure 3 , Figure 3 This diagram illustrates the data distribution of various data points in an electricity load dataset. The horizontal axis represents the ambient temperature in degrees Celsius, and the vertical axis represents the hourly electricity consumption in kWh / h. Figure 3 The data points shown in (a) represent the distribution of electricity load data for a target user throughout the year.

[0089] 207. Determine the temperature load distribution characteristics corresponding to the second power load dataset, wherein the temperature load distribution characteristics are used to reflect the proportion of each power load data point in the second power load dataset of each target user at each temperature;

[0090] It should be noted that after determining the second power load dataset for the target users, the distribution of electricity consumption patterns of the target users at different temperatures can be obtained from the second power load dataset. This distribution yields the temperature load distribution characteristics, which reflect the proportion of each power load data point in the second power load dataset for each target user at each temperature. These temperature load distribution characteristics can be expressed as percentages. Therefore, each temperature corresponds to a specific temperature load distribution characteristic. (See further...) Figure 3 , Figure 3 (b) shows the temperature distribution characteristics from -10°C to 40°C, wherein the temperature distribution characteristics include the 10th percentile, 50th percentile, and 90th percentile data at each temperature, the temperature gradient change rate during the heating season, and the temperature gradient change rate during the cooling season.

[0091] 208. Delete the power load data below the first proportion and above the second proportion in the temperature load distribution characteristics of each target user, and determine the target power load dataset for each target user. The first proportion and above the second proportion are used to indicate the proportion threshold of the power load data of the target user under extreme temperatures throughout the year. The first proportion is less than the second proportion.

[0092] Furthermore, by analyzing the hourly power consumption in the second power load dataset and the temperature load distribution of the ambient temperature at which that hourly power consumption occurs, the temperature load distribution characteristics at each temperature can be obtained, along with the percentage of hourly power consumption at that temperature. Based on these percentages, the data points in the second power load dataset are updated to obtain the target power load dataset. Specifically, power load data below the first percentage and above the second percentage in the temperature load distribution characteristics of each target user are deleted to determine the target power load dataset for each target user. For example, the first percentage can be 5 percentage points, and the second percentage can be 95 percentage points. Further, based on the annual temperature distribution range in the new dataset (the second power load dataset), data points below the 5th percentage point and above the 95th percentage point are removed for each target user to obtain the target power load dataset. Taking -10℃ as an example, at this temperature, there are N data points for the target user throughout the year. Each data point is represented as an array of (hourly power consumption, -10℃, time). There are 0.05N arrays for power consumption of 1 kWh / h at -10℃, 0.5N arrays for power consumption of 5 kWh / h at -10℃, and 0.95N arrays for power consumption of 10 kWh / h at -10℃. Therefore, the data points below the 5th percentile are represented as 0.05N arrays for power consumption of 1 kWh / h at -10℃, and the data points above the 95th percentile are represented as 0.05N arrays for power consumption of 10 kWh / h at -10℃. The above deletion operation is performed for each temperature and each user to finally obtain the target power load dataset.

[0093] 209. Based on the target user, target power load dataset, and preset electricity consumption characteristic parameters, perform characteristic value analysis to determine the characteristic value of the electricity consumption characteristic parameter corresponding to each target user;

[0094] It should be noted that step 209 and Figure 1 The content of step 103 shown is similar, and will not be repeated here to avoid repetition. For details, please refer to [link / reference needed]. Figure 1 The content of step 103 shown.

[0095] In one feasible implementation, in order to classify the electricity consumption habits of residential users based on temperature-related electricity consumption characteristics, the aforementioned electricity consumption characteristic parameters include at least the heating temperature threshold and the temperature gradient change rate during the heating season when the outside temperature is low, and the cooling temperature threshold and the temperature gradient change rate during the cooling season when the outside temperature is high. Therefore, step 209 may include steps A1-A4:

[0096] A1. Divide the temperature data in the target power load dataset using a preset temperature range width to determine the sub-power load dataset corresponding to each temperature range;

[0097] A2. Delete the target sub-power load dataset whose number of data points is less than a preset data point threshold, and obtain the updated target power load dataset for each target user;

[0098] To ensure the accuracy of the feature values, the target power load dataset needs to be updated. This dataset is a data sequence arranged from low to high temperature; details can be found in [reference needed]. Figure 3 (a) Subsequently, the temperature data in the target power load dataset can be divided using a preset temperature range width to determine the sub-power load datasets corresponding to each temperature range. Target sub-power load datasets with fewer than a preset data point threshold are then deleted, resulting in an updated target power load dataset for each target user, thus updating the target power load dataset. For example, the temperature range width can be 1 degree Celsius, and the preset data point threshold can be 20. Each 1-degree Celsius temperature range corresponds to one sub-power load dataset. By judging the number of data points in the sub-power load dataset, if the number of data points is less than 20, the sub-power load dataset is deleted. This data point count judgment is performed on all sub-power load datasets to update the target power load dataset.

[0099] A3. Determine the target data points corresponding to the third, fourth, and fifth proportions in the updated target power load dataset. The third, fourth, and fifth proportions increase sequentially, and the third, fourth, and fifth proportions are all greater than the first proportion, while the fifth proportion is less than the second proportion.

[0100] Furthermore, representative data points can be selected from the updated target power load dataset to improve the accuracy of the feature values ​​for each electricity consumption characteristic parameter of each target user. Specifically, representative data points for each temperature can be selected based on the proportion of electricity load data at each temperature. Continuing with the example of hourly electricity consumption as the electricity load data, the load proportion in the updated target power load dataset can be referenced... Figure 3 (b) Determine the target data points corresponding to the third, fourth, and fifth percentages in the updated target power load dataset based on the percentage distribution, wherein the third, fourth, and fifth percentages increase sequentially, and all three percentages are greater than the first percentage, while the fifth percentage is less than the second percentage; for example, continue with... Figure 3(b) Taking this as an example, the first percentage is 5 percentage points, the second percentage is 95 percentage points, the third percentage is 10 percentage points, the fourth percentage is 50 percentage points, and the fifth percentage is 90 percentage points. The target data points for 10 percentage points, 50 percentage points, and 90 percentage points correspond to... Figure 3 (b) shows the 10th percentile, 50th percentile, and 90th percentile data.

[0101] A4. Using the target data points and a preset linear fitting algorithm for temperature segmentation, perform fitting processing to determine the first characteristic value of the heating temperature threshold, the second characteristic value of the cooling temperature threshold, the third characteristic value of the heating season temperature gradient change rate, and the fourth characteristic value of the cooling season temperature gradient change rate for each target user.

[0102] Furthermore, by fitting the target data points corresponding to the third, fourth, and fifth proportions mentioned above with a preset linear fitting algorithm for temperature segments, the first characteristic value of the heating temperature threshold, the second characteristic value of the cooling temperature threshold, the third characteristic value of the heating season temperature gradient change rate, and the fourth characteristic value of the cooling season temperature gradient change rate are determined for each target user. Here, both the heating and cooling temperature thresholds are temperature critical values, with the heating temperature threshold being the temperature critical value for the heating season and the cooling temperature threshold being the temperature critical value for the heating season. The heating season temperature gradient change rate and the cooling season temperature gradient change rate are used to reflect the temperature gradient change rate during the heating and cooling seasons, respectively.

[0103] An exemplary linear fitting algorithm for temperature segments includes the following mathematical expression:

[0104]

[0105]

[0106] X h -x1≥5 (3)

[0107] X c -X h ≥5 (4)

[0108] x N -X c ≥5 (5)

[0109] 10≤X h ≤20 (6)

[0110] X h X c ∈Z (7)

[0111] In the formula, N represents the target data point (xn y n,i The total number of data points, where x represents temperature, y represents electricity load data, i∈{1,2,3} represents the third, fourth, and fifth percentages of each temperature point respectively, and j∈{1,2,3} represents the heating season, normal season, and cooling season of the temperature segment respectively; X h X is the heating temperature threshold. c This is the cooling temperature threshold, where x1 represents the target data point (x1, y1) when n is 1. 1,i The temperature of x N This represents the target data point (x) when n takes the value N. N y N,i ) temperature, b i，j ω is a constant term. 3，1 ω represents the rate of change of temperature gradient during the heating season. 3，3 Rate of change of temperature gradient during the cooling season.

[0112] Furthermore, the electricity consumption characteristic parameters also include base load and active load. Therefore, it is also necessary to determine the characteristic values ​​of base load and active load. Thus, the method further includes steps B1-B2:

[0113] B1. Using the target data points corresponding to the third proportion of each target user and the preset basic load algorithm, determine the fifth characteristic value of the basic load of each target user;

[0114] Continuing with the example of the third percentile, from Figure 3 (b) It can be known that the hourly power consumption of the target data point at the 10th percentile is the minimum at that temperature. Therefore, the base load can be determined based on the data point corresponding to the third percentile. This base load is used to reflect the minimum standard power demand at each temperature.

[0115] For example, the base load algorithm includes the following mathematical expression:

[0116]

[0117] In the formula, h1(x n () represents the data point x corresponding to the third proportion of the target user i=1. n N represents the total number of target data points for the third proportion; BL represents the basic load. Where i=1 can represent the third proportion, i=2 can represent the fourth proportion, and i=3 can represent the fifth proportion. h1(x n It can be obtained through a linear fitting algorithm based on temperature segments.

[0118] B2. Using the fifth characteristic value of each target user, the target data points under the fifth proportion, and the preset activity load algorithm, determine the sixth characteristic value of the activity load of each target user.

[0119] Continuing with the example of the third percentage being at the 10th percentile and the fifth percentage being at the 90th percentile, from Figure 3 (b) It can be known that the hourly power consumption of the target data point at the 90th percentile is the highest at that temperature. Therefore, the highest power consumption can be determined based on the data point corresponding to the fifth percentage. The active load is then the highest power consumption minus the base load, which reflects the maximum change in power consumption at various temperatures. Therefore, the sixth characteristic value of the active load of each target user is determined using the fifth characteristic value of each target user, the target data point at the fifth percentage, and the preset active load algorithm.

[0120] An example, the activity load algorithm includes the following mathematical expression:

[0121]

[0122] In the formula, h3(x n The fifth percentage of target users, i=3, corresponds to the data point x. n N represents the total number of target data points for the fifth proportion; BL represents the basic load, and AL represents the active load.

[0123] 2010. Cluster each target user according to the feature value to determine the power consumption category corresponding to each target user. The power consumption category is used to reflect the power consumption habits of the target user.

[0124] It should be noted that the content shown in step 2010 is the same as... Figure 1 Step 104 shown is similar in content and will not be repeated here to avoid repetition. Please refer to [link to relevant documentation] for details. Figure 1 The content of step 104 shown.

[0125] In one feasible implementation, the electricity consumption characteristic parameters include at least the heating temperature threshold, the cooling temperature threshold, the cooling season temperature gradient rate of change, the heating season temperature gradient rate of change, the base load, and the active load, wherein the heating season temperature gradient rate of change (HSG) ω 3，1 Units are kWh / h / ℃; Cooling season gradient (CSG) ω 3，3 The unit is kWh / h / ℃; heating temperature threshold X h The unit is ℃; cooling temperature threshold X c The unit is ℃; the base load is BL (Base load), the unit is kWh / h; the activity load is AL (Activity load), the unit is kWh / h. Taking K-means as an example, step 2010 can include steps C1-C6:

[0126] C1. Determine the pre-set target clustering requirements;

[0127] Among them, the target clustering requirements are related to how to process the electricity consumption characteristic parameters. For example, the target clustering requirements are used to reflect the difference between basic load and active load in electricity consumption habits, and the target clustering requirements are used to reflect the strong correlation of temperature in electricity consumption habits, etc. The electricity consumption characteristic parameters can be processed in a targeted manner according to the specific target clustering requirements, so that different clustering results can be obtained.

[0128] C2. Based on the data processing method corresponding to the target clustering requirements, perform data processing on the first characteristic value of the heating temperature threshold, the second characteristic value of the cooling temperature threshold, the third characteristic value of the temperature gradient change rate during the heating season, the fourth characteristic value of the temperature gradient change rate during the cooling season, the fifth characteristic value of the base load, and the sixth characteristic value of the active load for each target user to determine the target characteristic value set. The target characteristic value set includes the characteristic value set of each target user after data processing.

[0129] It should be noted that different target clustering requirements have different requirements for the dataset. Therefore, the dataset needs to be processed according to the corresponding data processing methods to obtain a dataset that meets the requirements. Specifically, according to the data processing methods corresponding to the target clustering requirements, the first feature value of the heating temperature threshold, the second feature value of the cooling temperature threshold, the third feature value of the heating season temperature gradient change rate, the fourth feature value of the cooling season temperature gradient change rate, the fifth feature value of the base load, and the sixth feature value of the active load for each target user are processed to determine the target feature value set. The data processing methods include, but are not limited to, scaling or normalization.

[0130] C3. Input the target feature value set into the preset K-means clustering model to determine each initial clustering result;

[0131] The processed dataset can then be used as input to a clustering model to classify target users. This clustering model is a K-means clustering model. The target feature set is then input into the preset K-means clustering model to determine the initial clustering results. Since the data centroid K is uncertain, multiple initial clustering results will be obtained, each corresponding to a data centroid K.

[0132] C4. Determine the silhouette coefficients of each initial clustering result;

[0133] Furthermore, to obtain the best clustering results, it is necessary to evaluate the quality of the initial clustering results. Specifically, this is done by determining the silhouette coefficient of each initial clustering result, which reflects the quality of the clustering. Generally, a larger silhouette coefficient indicates more obvious differences between clusters and a better clustering effect.

[0134] C5. Take the K value of the initial clustering result corresponding to the highest silhouette coefficient as the optimal K value;

[0135] C6. Cluster the target feature value set according to the optimal K value to determine the K power consumption categories corresponding to each target user output by the K-means clustering model.

[0136] It should be noted that a higher silhouette coefficient indicates a better and more reliable clustering result. Therefore, the K value of the initial clustering result corresponding to the highest silhouette coefficient is taken as the optimal K value. The target feature value set is clustered according to the optimal K value to determine the K power consumption categories corresponding to each target user output by the K-means clustering model.

[0137] The purpose of this application is to obtain temperature-related parameters through load decomposition and to conduct residential user clustering. The load decomposition method used in this invention determines the composition of various household appliances through non-intrusive metering, aiming to address the problem that time-domain analysis methods cannot account for the high randomness of residential electricity load, and to solve the pain point of conventional clustering analysis methods that struggle to incorporate the temperature dimension. The final results can reveal similar specific types of user information (house size, presence of electric vehicle charging, etc.) or similar temperature-load sensitivity relationships within the same cluster. The method proposed in this invention helps power grid companies better understand the load demand of residential users and lays the foundation for designing personalized demand response solutions for residents.

[0138] The beneficial effects of this application are as follows:

[0139] 1. The load analysis of users is carried out by comprehensively considering factors such as time, space and external environment. At the same time, the most critical influencing factor of residential user load - the outside temperature - is grasped. The factors are considered in a relatively comprehensive way, and a relatively complete residential load analysis system is constructed.

[0140] 2. The clustering algorithm efficiently uncovers the relationship between electricity load and objective unknown factors such as building area, high-power appliances, and central air conditioning settings, providing auxiliary guidance for power grid companies to carry out load management on the residential user side.

[0141] Please see Figure 4 , Figure 4 This is a structural block diagram of a user clustering device based on power load in an embodiment of the present invention, as shown below. Figure 4 The apparatus shown includes:

[0142] Data acquisition module 401: used to acquire the power load dataset of candidate users, the power load dataset being used to indicate the power consumption of candidate users;

[0143] Data processing module 402: Used to process outliers in the power load dataset by utilizing the annual temperature characteristics and the power load datasets of each candidate user, and to determine the target user and the target power load dataset of the target user;

[0144] Feature analysis module 403: used to perform feature value analysis processing based on the target user, target power load dataset and preset power consumption feature parameters, and determine the feature value of the power consumption feature parameter corresponding to each target user;

[0145] User classification module 404: used to perform clustering processing on each target user according to the feature value, and determine the power consumption category corresponding to each target user, wherein the power consumption category is used to reflect the power consumption habits of the target user.

[0146] It should be noted that, Figure 4 The functions of each module in the device shown are as follows: Figure 1 The steps in the described method are similar, and will not be repeated here to avoid repetition. For details, please refer to [link / reference needed]. Figure 1 The content of each step in the method.

[0147] This invention provides a user clustering device based on electricity load. The device includes: a data acquisition module for acquiring electricity load datasets of candidate users, which indicate the electricity consumption of candidate users; a data processing module for processing outliers in the electricity load datasets using year-round temperature characteristics and the electricity load datasets of each candidate user, to determine target users and their target electricity load datasets; a feature analysis module for performing feature value analysis based on target users, target electricity load datasets, and preset electricity consumption characteristic parameters, to determine the feature values ​​of the electricity consumption characteristic parameters corresponding to each target user; and a user classification module for clustering each target user based on the feature values, to determine the electricity consumption category corresponding to each target user, which reflects the electricity consumption habits of the target user. Through this method, the user clustering method can incorporate the temperature dimension, enabling better classification of user electricity consumption categories. This helps power grid companies better understand users' load demands and electricity consumption characteristics, and lays the foundation for designing personalized demand response solutions for residential users.

[0148] Figure 5 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 5As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. The internal memory may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. Those skilled in the art will understand that… Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0149] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform actions such as... Figure 1 or Figure 2 The steps of the method shown.

[0150] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following actions: Figure 1 or Figure 2 The steps of the method shown.

[0151] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0152] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0153] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for clustering users based on power load, characterized in that, The method comprises: obtaining a power load data set of a candidate user, the power load data set being used to indicate the power consumption of the candidate user; performing outlier processing on the power load data set by using annual temperature characteristics and the power load data set of each candidate user, determining a target user and a target power load data set of the target user; performing eigenvalue analysis processing on the target user, the target power load data set and a preset power consumption characteristic parameter, determining an eigenvalue of the power consumption characteristic parameter corresponding to each target user; performing clustering processing on each target user according to the eigenvalue, determining a power consumption category corresponding to each target user, the power consumption category being used to reflect the power consumption habit of the target user; wherein the power load data set comprises a corresponding relationship between the candidate user and the power load data, the power load data at least comprising the online power generation amount and the offline power consumption amount of the candidate user, and the annual temperature characteristics comprise annual temperature data; the outlier processing on the power load data set by using the annual temperature characteristics and the power load data set of each candidate user, the determination of the target user and the target power load data set of the target user, comprises: determining the total house power consumption of each candidate user according to the online power generation amount and the offline power consumption amount of the candidate user; determining an abnormal user with a negative total house power consumption; deleting the abnormal user and the power load data set corresponding to the abnormal user, obtaining a target user and a first power load data set of the target user; determining the annual temperature data of the location of each target user in the first power load data set; correlating the annual power load data of each target user with the annual temperature data of the location of the target user with the same variable of time, determining a second power load data set corresponding to the target user; determining a temperature load distribution characteristic corresponding to the second power load data set, the temperature load distribution characteristic being used to reflect the proportion of each power load data in the second power load data set of each target user under each temperature; deleting the power load data below a first proportion and above a second proportion in the temperature load distribution characteristic of each target user, determining a target power load data set of each target user, the first proportion and the second proportion being used to indicate the proportion critical value of the power load data of the target user under the extreme temperature in the year, the first proportion being less than the second proportion; wherein the power consumption characteristic parameter at least comprises a heating temperature threshold, a cooling temperature threshold, a cooling season temperature gradient change rate and a heating season temperature gradient change rate; the eigenvalue analysis processing on the target user, the target power load data set and the preset power consumption characteristic parameter, the determination of the eigenvalue of the power consumption characteristic parameter corresponding to each target user, comprises: dividing the temperature data in the target power load data set by using a preset temperature interval width, determining a sub-power load data set corresponding to each temperature interval; The target sub-power load data set with a number of data points less than a preset data point threshold in the sub-power load data set is deleted to obtain an updated target power load data set of each target user; Target data points corresponding to third, fourth and fifth proportions in the updated target power load data set are determined, the third, fourth and fifth proportions are sequentially increased, the third, fourth and fifth proportions are all greater than the first proportion, and the fifth proportion is less than the second proportion; The target data points and a preset temperature segmented linear fitting algorithm are used for fitting processing to determine a first characteristic value of a heating temperature threshold, a second characteristic value of a cooling temperature threshold, a third characteristic value of a heating season temperature gradient change rate and a fourth characteristic value of a cooling season temperature gradient change rate corresponding to each target user.

2. The method of claim 1, wherein, The temperature segmented linear fitting algorithm includes the following mathematical expression: （1） s.t. (2) （3） （4） （5） （6） （7） In the formula, N is the total number of data points of target data points (T) x n , y n,i , x represents temperature, y is power load data, respectively represent the third, fourth and fifth proportions of each temperature point, respectively represent the heating season, normal season and cooling season of the temperature segmentation; is the heating temperature threshold, is the cooling temperature threshold, x 1 represents the temperature of the target data point (T n 1, x 1, y 1,i ) when the value is 1, x N represents the temperature of the target data point (T n 1, N 1, x N , y N,i ) when the value is 1, b i，j is a constant term, is the heating season temperature gradient change rate, is the cooling season temperature gradient change rate.

3. The method of claim 2, wherein, The power consumption characteristic parameters further include a basic load and an activity load, and the method further includes: A fifth characteristic value of the basic load of each target user is determined by using the target data point corresponding to the third proportion of each target user and a preset basic load algorithm; A sixth characteristic value of the activity load of each target user is determined by using the fifth characteristic value of each target user, the target data point under the fifth proportion and a preset activity load algorithm.

4. The method of claim 3, wherein, The basic load algorithm includes the following mathematical expression: In the formula, h 1 x n ) represents h i = h When 1, the data point corresponding to the third proportion of the target user x n The calculation result of the linear fitting algorithm of temperature segmentation; N is the total number of target data points of the third proportion; BL is the basic load.

5. The method of claim 4, wherein, The activity load algorithm includes the following mathematical expression: In the formula, h 3 x n ) represents h i = h When t = 3, the data point corresponding to the fifth proportion of the target user x n The calculation result of the temperature segmentation linear fitting algorithm; N a total number of target data points for the fifth proportion; BL a base load, AL an active load.

6. The method of claim 1, wherein, The power consumption characteristic parameters at least include a heating temperature threshold, a cooling temperature threshold, a cooling season temperature gradient change rate, a heating season temperature gradient change rate, a basic load and an activity load, and the clustering processing of each target user according to the characteristic values to determine the power consumption category corresponding to each target user includes: A target clustering requirement is determined in advance; Data processing of the first characteristic value of the heating temperature threshold, the second characteristic value of the cooling temperature threshold, the third characteristic value of the heating season temperature gradient change rate, the fourth characteristic value of the cooling season temperature gradient change rate, the fifth characteristic value of the basic load and the sixth characteristic value of the activity load of each target user is performed according to a data processing mode corresponding to the target clustering requirement to determine a target characteristic value set, and the target characteristic value set includes the characteristic value set of each target user after data processing; The target characteristic value set is input into a preset K-means clustering model to determine each initial clustering result; The silhouette coefficient of each initial clustering result is determined; The K value of the initial clustering result corresponding to the highest silhouette coefficient is taken as the best K value; The target characteristic value set is clustered according to the best K value to determine K power consumption categories corresponding to each target user output by the K-means clustering model.

7. A user clustering device based on power load, characterized in that, The device includes: A data acquisition module is configured to acquire a power load data set of a candidate user, and the power load data set is used to indicate the power consumption of the candidate user; The data processing module is used for performing outlier processing on the power load data set by using the annual temperature characteristics and the power load data set of each candidate user, determining the target user and the target power load data set of the target user; wherein the power load data set comprises a corresponding relationship between the candidate user and the power load data, the power load data at least comprises the online power generation amount and the offline power consumption amount of the candidate user, and the annual temperature characteristics comprise annual temperature data; the utilization of the annual temperature characteristics and the power load data set of each candidate user to perform outlier processing on the power load data set, and the determination of the target user and the target power load data set of the target user, comprises: determining the total power consumption of each candidate user according to the online power generation amount and the offline power consumption amount of each candidate user; determining the abnormal user of the total power consumption of the house; deleting the abnormal user and the power load data set corresponding to the abnormal user to obtain the target user and the first power load data set of the target user; determining the annual temperature data of the location of each target user in the first power load data set; correlating the annual power load data of each target user with the annual temperature data of the location by taking time as the same variable, and determining the second power load data set corresponding to the target user; determining the temperature load distribution characteristics corresponding to the second power load data set, the temperature load distribution characteristics are used for reflecting the proportion of each power load data in the second power load data set of each target user at each temperature; deleting the power load data below the first proportion and above the second proportion in the temperature load distribution characteristics of each target user, and determining the target power load data set of each target user, the first proportion and the second proportion are used to indicate the proportion critical value of the power load data of the target user at the extreme temperature in the year, and the first proportion is less than the second proportion; The characteristic analysis module is configured to perform characteristic value analysis on the target users, the target power load data set, and preset power consumption characteristic parameters to determine characteristic values of the power consumption characteristic parameters corresponding to each target user; wherein the power consumption characteristic parameters at least include a heating temperature threshold, a cooling temperature threshold, a cooling season temperature gradient change rate, and a heating season temperature gradient change rate; and the characteristic value analysis includes: dividing temperature data in the target power load data set by using a preset temperature interval width to determine sub-power load data sets corresponding to each temperature interval; deleting target sub-power load data sets in which the number of data points is less than a preset data point threshold to obtain updated target power load data sets of each target user; determining target data points corresponding to third, fourth, and fifth proportions in the updated target power load data sets, wherein the third, fourth, and fifth proportions increase in sequence, the third, fourth, and fifth proportions are all greater than the first proportion, and the fifth proportion is less than the second proportion; and performing fitting processing on the target data points and a preset temperature segmentation linear fitting algorithm to determine first, second, third, and fourth characteristic values of the heating temperature threshold, the cooling temperature threshold, the heating season temperature gradient change rate, and the cooling season temperature gradient change rate corresponding to each target user. The user classification module is configured to perform clustering processing on the target users according to the characteristic values to determine power consumption categories corresponding to each target user, wherein the power consumption categories reflect power consumption habits of the target users.

8. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to cause the processor to perform the steps of the method of any one of claims 1 to 6. 9.A computer device, comprising a memory and a processor, and characterized in that, The memory stores a computer program, and the computer program is executed by the processor to cause the processor to perform the steps of the method of any one of claims 1 to 6.

Images