Method and device for constructing power user profiles

Abstract

This invention discloses a method for constructing electricity user profiles, comprising: S1: acquiring long-term electricity consumption data of electricity users, clustering the long-term electricity consumption data to obtain classification clusters, extracting features of the classification clusters, and obtaining user profiles; S2: selecting users from each classification cluster and acquiring short-term electricity consumption data of the selected users to construct an electricity consumption prediction model; S3: acquiring payment data and arrears data of the selected users to construct an arrears prediction model; S4: acquiring electricity consumption behavior data of the selected users to construct an electricity consumption behavior prediction model; S5: refining the user profiles based on the electricity consumption prediction model, the arrears prediction model, and the electricity consumption behavior prediction model. This invention also provides a construction apparatus. This invention can improve the accuracy and real-time performance of user profiles.

Description

Technical Field

[0001] This invention relates to the field of data mining technology. More specifically, this invention relates to a method and apparatus for constructing power user profiles. Background Technology

[0002] Characterizing and providing precise services based on the electricity consumption characteristics and habits of different users is beneficial for improving the user service experience in many ways and achieving refined user management. However, in existing technologies, electricity user profiling simply uses the k-means algorithm for clustering, which is relatively crude, easily affected by the initial cluster center selection, and difficult to adapt to the continuous changes in user electricity consumption behavior. Therefore, it is necessary to design a technical solution that can improve upon these shortcomings. Summary of the Invention

[0003] One objective of this invention is to provide a method and apparatus for constructing power user profiles, which can improve the accuracy and real-time performance of user profiles.

[0004] To achieve these objectives and other advantages of the present invention, according to one aspect of the present invention, a method for constructing a power user profile is provided, comprising: S1: acquiring long-term electricity consumption data of power users, clustering the long-term electricity consumption data to obtain classification clusters, extracting features of the classification clusters, and obtaining a user profile; S2: selecting users from each classification cluster and acquiring short-term electricity consumption data of the selected users, and constructing an electricity consumption prediction model; S3: acquiring payment data and arrears data of the selected users, and constructing an arrears prediction model; S4: acquiring electricity consumption behavior data of the selected users, and constructing an electricity consumption behavior prediction model; S5: refining the user profile based on the electricity consumption prediction model, the arrears prediction model, and the electricity consumption behavior prediction model.

[0005] Further, S1 includes: cleaning the acquired long-term electricity consumption data to remove outliers; standardizing or normalizing the data; clustering the data using the DBSCAN algorithm to obtain classification clusters; extracting features from each classification cluster and describing them with semantic tags to obtain the user profile.

[0006] Further, S2 includes: selecting a predetermined proportion of electricity users from each of the classification clusters as the selected users; acquiring the historical electricity consumption data and historical meteorological data of the selected users; using the electricity consumption data and meteorological data of the previous few time points as input and the electricity consumption data of the next time point as output, and constructing the electricity consumption prediction model using an LSTM network.

[0007] Further, S3 includes: extracting the average payment cycle, time since the last payment, cumulative number of arrears, maximum arrears amount, arrears amount percentage, on-time payment percentage, and payment amount fluctuation of the selected user from the payment data and the arrears data; and establishing a training set and a test set by combining the label of whether or not arrears exist; and using the training set to train a decision tree model to obtain the arrears prediction model.

[0008] Further, S4 includes: extracting differential features of electricity consumption, sliding window features of electricity consumption, peak features of electricity consumption, periodicity of daily electricity consumption, and periodicity of monthly electricity consumption from the electricity consumption behavior data; combining the electricity consumption behavior labels to establish a training set and a test set; and using the training set to train an MLP model to obtain the electricity consumption behavior prediction model.

[0009] Furthermore, in step S5, the future electricity consumption, probability of arrears, and electricity behavior of the electricity user are predicted using the electricity consumption prediction model, the arrears prediction model, and the electricity behavior prediction model to obtain prediction results; the prediction results are used to supplement the user profile and mark any inconsistencies.

[0010] Furthermore, in step S5, the question-and-answer data of the electricity user is also acquired, the semantics of the question-and-answer data are extracted, and the semantics are used to determine the contradictions.

[0011] According to another aspect of the present invention, an apparatus for constructing an electricity user profile is also provided, comprising a processor and a memory, the memory being used to store program instructions, and the processor being used to invoke the program instructions to execute the electricity user profile construction method.

[0012] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements any of the described methods for constructing power user profiles.

[0013] The present invention has at least the following beneficial effects:

[0014] This invention utilizes long-term electricity consumption data of power users to construct user profiles, and uses short-term electricity consumption data to construct electricity consumption prediction models, payment and arrears data to construct arrears prediction models, and electricity behavior data to construct electricity behavior prediction models. These electricity consumption prediction models, arrears prediction models, and electricity behavior prediction models are then used to refine the user profiles. This invention uses the electricity consumption prediction model, arrears prediction model, and electricity behavior prediction model to supplement and improve the user profiles in real time, thereby enhancing the accuracy and real-time performance of the user profiles.

[0015] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0016] Figure 1 This is a flowchart of one embodiment of this application. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0018] It should be understood that terms such as "having," "comprising," and "including" used in the embodiments of this application do not exclude the presence or addition of one or more other elements or combinations thereof. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of components in a specific posture. If the specific posture changes, the directional indication will also change accordingly. When an element is referred to as "fixed to" or "set on" another element, it can be directly on the other element or may have an intervening element present. When an element is referred to as "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element through an intervening element. Descriptions involving "first," "second," etc., in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features.

[0019] It should be noted that the technical solutions of the various embodiments of this application can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this application.

[0020] like Figure 1 As shown, embodiments of this application provide a method for constructing electricity user profiles, including:

[0021] S1: Obtain long-term electricity consumption data of electricity users, cluster the long-term electricity consumption data to obtain classification clusters, extract the features of the classification clusters, and obtain user profiles;

[0022] For example, user electricity consumption data at different time granularities (such as every 15 minutes, hour, day, month, etc.) and related electricity consumption parameter information such as power consumption and power factor are obtained from the electricity metering system; by connecting with the database of the electricity metering system, detailed electricity consumption data records corresponding to each electricity user are extracted according to the user's unique identifier (such as user number), ensuring that the data covers a sufficiently long time range, such as the past year, two years or even longer, to fully reflect the long-term characteristics and changing trends of users' electricity consumption behavior;

[0023] For example, K-Means clustering or DBSCAN density clustering can be used for clustering. K-Means clustering is a commonly used distance-based clustering algorithm. Its principle is to divide data points into K different clusters such that the sum of the distances from each data point in each cluster to its cluster center is minimized. For example, when clustering long-term electricity consumption data of electricity users, K=3 is determined to be appropriate by the elbow method. Then, the preprocessed electricity consumption data is input into the K-Means algorithm for clustering, and finally, the users are divided into 3 different clusters. DBSCAN density clustering is based on the density of data points, dividing densely connected data points into the same cluster. It can discover clusters of arbitrary shapes and can identify noise points in the data (i.e., isolated points that do not belong to any cluster). When using it, two key parameters need to be specified: neighborhood radius and minimum number of points. These determine the definition of data point density and the conditions for cluster formation. Through multiple experiments and adjustments of these two parameters, better clustering results can be obtained, and it is not necessary to pre-specify the number of clusters.

[0024] For example, the characteristics of the classification clusters include user type, electricity consumption scale, electricity consumption characteristics, electricity consumption time patterns, etc., forming a user profile;

[0025] S2: Select users from each of the classification clusters and obtain short-term electricity consumption data of the selected users to construct an electricity consumption prediction model;

[0026] For example, a certain number of users are randomly selected from each category cluster, such as 30% of each cluster, as selected users. Selecting users from multiple categories clusters can reduce redundant samples, reduce the number of samples used to a certain extent, and cover various types of electricity consumption as much as possible, thereby improving the reliability of the electricity consumption prediction model.

[0027] For example, based on short-term electricity consumption data, a power consumption prediction model can be constructed using regression models, machine learning models, or neural network models. Here, short-term electricity consumption data refers to electricity consumption data for 7 days, 10 days, or 20 days, thereby reflecting changes in users' electricity consumption.

[0028] S3: Obtain the payment data and arrears data of the selected users, and construct an arrears prediction model;

[0029] For example, payment data and arrears data can be obtained from the electricity marketing system. From this data, detailed payment records of users can be obtained, including the time, amount, and payment method of each payment, as well as information related to arrears, such as the time of arrears, the amount of arrears, the duration of arrears, and whether the arrears have been settled. After the above data is preprocessed, an arrears prediction model can be built using a machine learning model or a neural network model.

[0030] S4: Obtain the electricity consumption behavior data of the selected users and construct an electricity consumption behavior prediction model;

[0031] For example, features reflecting user electricity consumption behavior, such as electricity consumption time period, peak electricity consumption and electricity consumption fluctuation, are extracted from short-term electricity consumption data. After preprocessing, an electricity consumption behavior prediction model is constructed using linear regression model, nonlinear regression model, time series analysis model, machine learning model and deep learning model.

[0032] S5: Improve the user profile based on the electricity consumption prediction model, the overdue payment prediction model, and the electricity consumption behavior prediction model;

[0033] For example, the electricity consumption prediction model, the arrears prediction model, and the electricity behavior prediction model are used to predict the future electricity consumption, arrears probability, and electricity behavior of the electricity user, and the prediction results are obtained. The prediction results are used to supplement the user profile to obtain a more accurate user profile that adapts to user changes, and contradictions are marked to remind staff to handle the contradictions.

[0034] For example, a correlation analysis is performed on the user profiles of predicted electricity consumption, outstanding payments, and electricity usage behavior to check for logical inconsistencies or contradictions. For instance, if the electricity consumption forecast indicates a significant increase in future consumption, but the outstanding payment forecast shows low risk and a consistently good payment history, further analysis is needed to determine if the user has new sources of funding or reasonable reasons such as changes in their electricity usage plan. This ensures that the various parts of the profile corroborate each other and reflect the actual situation. Business experts in the power sector are invited to review and optimize the improved user profiles based on their professional knowledge and practical experience. For example, experts might adjust the profiles based on changes in industrial policies in the user's region. Different opinions may arise regarding the predicted results of household electricity consumption and arrears risk. Alternatively, based on common electricity consumption behavior characteristics of similar users within the industry, it may be suggested that certain user profile descriptions be supplemented or modified to better align with actual business scenarios and more accurately reflect users' true electricity consumption characteristics and potential risks. As time progresses and new data is continuously generated, the above steps should be repeated periodically (e.g., monthly, quarterly, depending on business needs) to rerun the prediction models using the latest data, updating and iterating user profiles. This ensures that the profiles reflect users' latest electricity consumption, arrears risk, and trends in electricity consumption behavior in real time, providing continuous and accurate decision-making support for the power company's marketing, operations, and service work.

[0035] As can be seen, this embodiment uses long-term electricity consumption data clustering to obtain an initial user profile, and then uses electricity consumption prediction model, overdue payment prediction model and electricity consumption behavior prediction model that can make short-term predictions to supplement and improve the user profile in real time, thereby improving the accuracy and real-time performance of the user profile.

[0036] In another embodiment, S1 includes: cleaning the acquired long-term electricity consumption data to remove outliers; standardizing or normalizing the data; clustering the data using the DBSCAN algorithm to obtain classification clusters; extracting features from each classification cluster and describing them with semantic tags to obtain the user profile.

[0037] For example, by selecting an appropriate neighborhood radius (eps) and minimum number of samples (min_samples), such as eps between 0.3 and 0.5 and min_samples between 5 and 10, the standardized or normalized long-term electricity consumption data is input into the DBSCAN algorithm for clustering calculation. The algorithm automatically divides the data points into different clusters and marks noise points. It calculates statistical indicators such as the average electricity consumption, median, maximum, minimum, and standard deviation of electricity consumption for users in each cluster, so as to comprehensively understand the overall situation and dispersion of the electricity consumption scale of users in the cluster. Calculate the month-on-month (compared to the previous period) and year-on-year (compared to the same period last year) change rates of electricity consumption within the cluster, and observe whether there are seasonal patterns (long-term growth or decline trends, etc.) to characterize the electricity consumption behavior of users in the cluster at different time dimensions. Statistically analyze the proportion of electricity consumption or the average power consumption of each user within the cluster at different times of the day (e.g., one hour per time period), and plot the daily electricity consumption time distribution curve to observe whether there are obvious peak and off-peak periods and the degree and pattern of concentration of these periods. Analyze the differences in electricity consumption and power consumption between weekdays and weekends for users within the cluster, and calculate the average and ratio of weekday and weekend electricity consumption indicators to understand the electricity consumption variation patterns of users within a week. For example, in some clusters where commercial users reside, weekend electricity consumption is significantly higher than weekdays, reflecting the correlation between their business activities and rest days; while for some clusters where industrial users reside, weekday electricity consumption is higher and fluctuates less, reflecting their stable electricity consumption pattern based on production schedules. Calculate the average power consumption of users within each cluster and compare the average power consumption between different clusters to differentiate... The differences in user electricity consumption scale and equipment usage intensity are considered. Statistical indicators such as the standard deviation of power consumption are used to measure the degree of power fluctuation over a certain period. The magnitude of the fluctuation reflects the stability of user electricity consumption behavior. Power fluctuations of industrial users may be related to factors such as production processes and equipment start-up and shutdown sequences, and the fluctuation amplitude is larger and the regularity is relatively stronger. Residential users' daily power consumption fluctuates relatively frequently but with smaller amplitudes, mainly affected by the switching operation of electrical appliances. By analyzing power fluctuations, the electricity consumption behavior characteristics of different user clusters are further refined. Based on the various features extracted above, a user profile with semantic labels is constructed for each category cluster, which can intuitively reflect the electricity consumption behavior characteristics of the user group in that cluster. For example, the residential electricity consumption stability type means that the cluster is mainly composed of residential users with high stability in electricity consumption behavior; the commercial electricity consumption fluctuation type indicates that the cluster is mainly composed of commercial users with obvious fluctuations in electricity consumption behavior at different time periods; the industrial electricity consumption production type indicates that the cluster is mainly composed of industrial users, and the electricity consumption behavior revolves around production activities and has a certain regularity and a large scale of electricity consumption.

[0038] In another embodiment, S2 includes: selecting a predetermined proportion of electricity users from each of the classification clusters as the selected users; obtaining historical electricity consumption data and historical meteorological data of the selected users; using electricity consumption data and meteorological data from several previous time points as input and electricity consumption data from a later time point as output, and constructing the electricity consumption prediction model using an LSTM network.

[0039] For example, meteorological data includes temperature, precipitation, humidity, wind speed, etc. The collected historical electricity consumption data and historical meteorological data for each selected user are divided chronologically to form continuous time series segments. For instance, using days as the time unit, the electricity consumption and meteorological data for a user over the past month can be arranged sequentially to form a time series of length 30. Then, an appropriate time window size is set, which determines the length of historical data used for prediction each time the model makes a prediction; that is, the data from the first few time points are used as input. Based on the dimensionality of the input features (i.e., the number of features in the electricity consumption and meteorological data within the time window) and the requirements of the prediction task, a suitable LSTM network structure is designed, including an input layer, one or more hidden layers, and an output layer. The parameters of the LSTM network (such as weight matrix and bias vector) are initialized. Preprocessed and partitioned time-series data (based on a set time window, using electricity consumption and meteorological data from multiple time points as input, and the electricity consumption at the next corresponding time point as output) are divided into training, validation, and test sets. Typically, this is done in a certain proportion, such as 70%-80% of the data as the training set for model learning, 10%-20% as the validation set to adjust model parameters and prevent overfitting, and the remaining 10%-20% as the test set to finally evaluate the model's generalization ability on unseen data. Then, the LSTM network is trained using the training set data. During training, the input data is sequentially fed into the network's input layer, and the predicted electricity consumption output is calculated through forward propagation. Finally, the error between the predicted output and the actual electricity consumption (common error functions include mean squared error (MSE) and mean absolute error (MAE)) is calculated. The algorithm uses backpropagation to propagate errors from the output layer to the input layer, updating various parameters in the network (weight matrix and bias vector, etc.). This process is repeated continuously for multiple training epochs, which continuously optimizes the model, gradually reduces the prediction error, and improves the accuracy of electricity consumption prediction.

[0040] In another embodiment, S3 includes: extracting the average payment cycle, time since the last payment, cumulative number of arrears, maximum arrears amount, arrears amount percentage, on-time payment percentage, and payment amount fluctuation of the selected user from the payment data and the arrears data; and establishing a training set and a test set by combining the label of whether or not arrears exist; and using the training set to train a decision tree model to obtain the arrears prediction model.

[0041] For example, regarding the average payment cycle: calculate the number of days between two consecutive payment dates for each selected user, and then average these intervals to obtain the average payment cycle; Time since the last payment: using the current time as a reference, calculate how many days have passed since each user's last payment. This feature can intuitively reflect the user's recent payment activity. The longer the time since the last payment, the greater the risk of arrears may be, especially when the payment cycle is significantly longer than usual, requiring close attention to the possibility of arrears; Cumulative number of arrears: count the cumulative number of times each user has arrears. The cumulative number of arrears directly reflects the frequency of arrears in the user's history. The more times, the more serious the user's past arrears problems, and the greater the possibility of arrears in the future. It is one of the key indicators for measuring the risk of arrears; Maximum arrears amount: find the maximum amount owed by each user in all past arrears. The maximum amount owed in the billing records; the percentage of owed amount: the ratio of accumulated owed amount to accumulated paid amount, which measures the proportion of owed amount in the overall electricity cost; the percentage of on-time payments: by comparing the user's due payment time (determined according to the payment cycle and relevant regulations) with the actual payment time, the percentage of on-time payments is calculated. The percentage of on-time payments clearly shows the user's willingness and habits to pay, and users with a higher percentage usually have a lower risk of default; the fluctuation of payment amount: the standard deviation of the user's payment amount for each payment is calculated to measure the degree of fluctuation in payment amount;

[0042] The extracted features (average payment cycle, time since the last payment, cumulative number of arrears, maximum arrears amount, percentage of arrears, percentage of timely payments, and fluctuations in payment amount) are used as independent variables, and whether or not a user is in arrears (0 indicates no arrears, 1 indicates arrears, labeled according to whether the user actually incurred arrears) is used as the dependent variable to construct a dataset; the dataset is then randomly divided into a training set and a test set according to a certain ratio.

[0043] Before training a decision tree model using the training set, some relevant parameters need to be set. These parameters affect the growth of the decision tree and the final model performance. Common parameters include: maximum depth (max_depth): limits the maximum number of layers in the decision tree to prevent the tree from growing too complex and causing overfitting; minimum sample split (min_samples_split): specifies the minimum number of samples a node must contain before it can be split into child nodes; minimum sample leaf (min_samples_leaf): represents the minimum number of samples each leaf node must contain. Its function is similar to minimum sample split, also to prevent overfitting and ensure that leaf nodes have a certain degree of representativeness.

[0044] The feature data and corresponding whether a user is in arrears in the training set are organized according to the format required by the decision tree model (usually organized into a two-dimensional array or data frame), and then fed into the decision tree model for training. During the training process, the decision tree model will automatically adjust the tree structure and node partitioning rules according to the relationship between each feature in the training data and the arrears result. By calculating information gain (such as measuring the distinguishing ability of features to classify results based on criteria such as information entropy or Gini index, and selecting the feature with the largest information gain for node partitioning), the model will be continuously optimized so that the model can predict whether a user will be in arrears as accurately as possible based on the input features.

[0045] In another embodiment, S4 includes: extracting differential features of electricity consumption, sliding window features of electricity consumption, peak features of electricity consumption, periodicity of daily electricity consumption, and periodicity of monthly electricity consumption from the electricity consumption behavior data; combining the electricity consumption behavior labels to establish a training set and a test set; and using the training set to train an MLP model to obtain the electricity consumption behavior prediction model.

[0046] For example, the differential characteristics of electricity consumption: calculating the difference in electricity consumption between adjacent time points to observe whether the trend of electricity consumption is stable, eliminating trend and seasonal factors in the data, and facilitating subsequent models to better capture the inherent laws of electricity consumption changes; the sliding window characteristics of electricity consumption: extracting statistical features of electricity consumption within a certain time window to capture the fluctuations and patterns of users' electricity consumption behavior in the short term, understanding the stability of users' recent electricity consumption and changes in their electricity consumption habits; and the peak characteristics of power consumption: statistically analyzing the time points and peak values ​​of each user's power consumption in different time periods such as daily and weekly, analyzing the patterns of peak power occurrence (such as whether it is concentrated in certain specific time periods) and the changing trends of peak values, which is helpful for a deeper understanding of users' use of electrical equipment. Habitual and concentrated electricity consumption patterns are of great significance; Daily electricity consumption periodicity: Analyze the distribution patterns of users' electricity consumption at different times of the day to see if there are obvious peak and off-peak periods, and whether these periods repeat within a week; Monthly electricity consumption periodicity: Consider the fluctuations in users' monthly electricity consumption, whether it is affected by factors such as seasons and production plans (for industrial users) and exhibits periodic changes. For example, residential users will have peak monthly electricity consumption in the summer months due to high air conditioning usage, and industrial users will have significantly increased electricity consumption in the months of peak production season. This periodicity is quantified by calculating indicators such as the average monthly electricity consumption, standard deviation, and the month-on-month and year-on-year change rates of electricity consumption between different months, and is transformed into a numerical form that can be recognized and learned by the model;

[0047] The extracted features (differential features of electricity consumption, sliding window features of electricity consumption, peak features of electricity consumption, periodicity of daily electricity consumption, and periodicity of monthly electricity consumption) are used as independent variables, and electricity consumption behavior labels (e.g., electricity consumption behavior can be classified and labeled according to different business needs, such as the earlier or later peak hours, the growth or decline trend of electricity consumption, etc. Here, specific categories or values ​​are used to represent different electricity consumption behaviors, determined according to the actual analysis and prediction objectives, such as peak electricity consumption type, off-peak electricity consumption type, and stable electricity consumption type) are used as dependent variables to construct a dataset. A certain proportion (usually about 70% - 80% of the data is divided into training sets for model learning and training) is used.

[0048] MLP is a simple feedforward neural network consisting of an input layer, one or more hidden layers, and an output layer. Neurons are connected by appropriate weights. The number of nodes in the input layer depends on the number of input features. For example, in the extracted electricity consumption behavior features, there are 5 different dimensions of features, so the input layer has 5 nodes. The number of neurons in the hidden layer can be set based on experience and experimentation. Different numbers of hidden layers and different combinations of the number of neurons in each layer will affect the model's expressive power and learning effect. Its function is to perform nonlinear transformation on the input features and uncover the complex nonlinear mapping relationship between the input features and the output target (electricity consumption behavior label). The number of nodes in the output layer is determined by the dimension of the predicted electricity consumption behavior. For example, if the predicted electricity consumption behavior is peak consumption type, off-peak consumption type, and stable consumption type, then the number of nodes in the output layer is 3.

[0049] During training, data is input from the input layer, processed by neurons in the hidden layer through weighted summation and activation functions (such as the commonly used Sigmoid function and ReLU function, used to introduce nonlinear factors and enhance the model's expressive power), and then passed to the output layer to obtain the prediction result. Then, by comparing the difference between the prediction result and the actual electricity consumption behavior label (common error metrics such as mean squared error (MSE) and mean absolute error (MAE), the error is propagated back from the output layer to the input layer using the backpropagation algorithm. The weights of each connection in the network are adjusted based on the error, and this process is repeated continuously through multiple training epochs, allowing the model to continuously optimize and gradually reduce the prediction error, thereby achieving accurate prediction of electricity consumption behavior.

[0050] In another embodiment, in step S5, the question-and-answer data of the electricity user is also acquired, the semantics of the question-and-answer data are extracted, and the semantics are used to determine the contradictions.

[0051] For example, the Q&A data from electricity users mainly comes from multiple sources. First, there is the customer service system of the power company. When users encounter electricity-related problems, they often seek help from the power company by calling customer service hotlines or consulting online customer service. Customer service personnel record the content of each Q&A interaction, including detailed information such as the questions raised by the user, the specific time of the consultation, and the corresponding customer service response. In addition, interactive communities and frequently asked questions sections set up on some power company official websites or mobile applications also retain user questions and responses from other users or power company staff. These also constitute a rich source of Q&A data.

[0052] To accurately extract semantics from question-and-answer data, it is necessary to use Natural Language Processing (NLP) techniques; common methods include lexical analysis, syntactic analysis, and semantic role labeling.

[0053] Then, semantic roles are used to further extract semantics, or pre-trained language models are employed, to help identify contradictions. The extracted question-and-answer data semantics are then compared and analyzed with other relevant data (such as user profiles, electricity consumption behavior data, electricity consumption prediction results, overdue payment prediction results, and electricity consumption behavior prediction results constructed in previous steps). For example, the user profile might indicate that the user is a stable residential electricity user with relatively fixed daily electrical equipment and habits, and their electricity consumption should fluctuate little under normal circumstances. However, the question-and-answer data semantics might reveal that the user reports a significant increase in electricity bills and suspects a meter malfunction, which forms a potential contradiction requiring further analysis. Through this semantic-based contradiction identification, potentially problematic links can be accurately located, allowing for appropriate solutions to improve the quality and accuracy of electricity services.

[0054] Embodiments of this application also provide an apparatus for constructing a power user profile, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the program instructions to execute the power user profile construction method. Embodiments of this application also provide a target detection apparatus, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the program instructions to execute the target detection method. The apparatus in this embodiment can be a mobile phone, a laptop computer, a tablet computer, etc., which internally provides a memory and a processor to execute the power user profile construction method of the above embodiments.

[0055] The embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements any of the power user profile construction methods described above. In this embodiment, the computer-readable medium may be a portable computer disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, etc., capable of storing a computer program to execute the power user profile construction method of the above embodiments.

[0056] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method of power consumer profiling, characterized by, The method comprises the following steps: S1: obtaining long-term power consumption data of power users, clustering the long-term power consumption data to obtain classification clusters, extracting features of the classification clusters to obtain user portraits; S2: selecting users from each of the classification clusters and obtaining short-term power consumption data of the selected users to construct a power consumption prediction model; S3: obtaining payment data and arrears data of the selected users to construct an arrears prediction model; S4: obtaining power consumption behavior data of the selected users to construct a power consumption behavior prediction model; S5: improving the user portraits according to the power consumption prediction model, the arrears prediction model and the power consumption behavior prediction model; The S1 comprises the following steps: cleaning the obtained long-term power consumption data to remove outliers; standardizing or normalizing the data; clustering by using a DBSCAN algorithm to obtain classification clusters; extracting features of each of the classification clusters and describing them with semantic labels to obtain the user portraits; The S2 comprises the following steps: selecting a predetermined proportion of power users from each of the classification clusters as the selected users; obtaining historical power consumption data and historical weather data of the selected users; using power consumption data and weather data of previous time points as inputs and power consumption data of a later time point as an output to construct the power consumption prediction model by using an LSTM network; The S3 comprises the following steps: extracting the average payment period, the time since the last payment, the cumulative number of arrears, the maximum amount of arrears, the proportion of arrears, the proportion of timely payment and the fluctuation of payment amount of the selected users from the payment data and the arrears data, combining with the label of whether to be in arrears to establish a training set and a test set; training a decision tree model by using the training set to obtain the arrears prediction model; The S4 comprises the following steps: extracting the differential features of power consumption, the sliding window features of power consumption, the peak features of power consumption, the periodicity of daily power consumption and the periodicity of monthly power consumption from the power consumption behavior data, combining with the power consumption behavior label to establish a training set and a test set; training an MLP model by using the training set to obtain the power consumption behavior prediction model; In the S5, the power consumption prediction model, the arrears prediction model and the power consumption behavior prediction model are used to predict the future power consumption, the arrears probability and the power consumption behavior of the power users to obtain prediction results; the prediction results are used to supplement the user portraits and mark the contradictions; in the S5, the question and answer data of the power users are also obtained, the semantics of the question and answer data are extracted, and the semantics are used to judge the contradictions.

2. The power consumer profiling apparatus, characterized by, The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the power user portrait construction method of claim 1.

3. A computer readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the power user portrait construction method of claim 1.

Images