Method, system and equipment for pre-judging potential safety hazard of electricity utilization according to customer feature information, and storage medium

Abstract

The invention discloses a method, system and device for pre-judging potential safety hazards of electricity utilization according to customer feature information and a storage medium, and belongs to the technical field of electric power data, and the method comprises the steps: obtaining the customer feature information and electricity utilization data, carrying out the preprocessing, obtaining a data set, carrying out the aggregation modeling, generating an electricity utilization behavior portrait, and constructing an electricity utilization behavior mode; analyzing the power consumption data and identifying a change trend; comparing the change trend with a preset behavior mode to judge abnormity; if abnormity exists, an anti-misoperation guidance scheme is generated, adaptability judgment is carried out, final guidance is generated and pushed to the client terminal, feedback information is recorded, a behavior portrait and a guidance strategy are updated, and a dynamic behavior analysis and risk identification model is optimized. According to the method, the problem that dynamic behavior modeling and risk identification are lacked based on customer personalized behavior data is solved, and particularly the technical problems of poor adaptability and insufficient feedback updating capability in the aspects of anti-misoperation strategy generation and adaptive optimization are solved.

Description

Technical Field

[0001] This invention relates to the field of power data technology, specifically to a method, system, device, and storage medium for predicting potential electricity safety hazards based on customer characteristic information. Background Technology

[0002] Predicting potential electrical safety hazards is a key research direction in the field of power management and safety. Its core lies in using scientific methods to reduce safety accidents caused by human error or equipment failure, ensuring the stable operation of the power system and the safety of users' lives and property. With the popularization of smart grids and IoT technologies, predicting potential electrical safety hazards based on customer characteristic information has become an important way to improve power supply reliability and user experience. However, existing methods have significant limitations in practical applications.

[0003] Many current solutions rely on single device monitoring or generic safety alerts, lacking in-depth analysis of customers' personalized electricity usage behaviors. This makes it difficult to accurately identify potential operational risks from different customer groups, such as the complex equipment operation procedures of industrial customers or the diverse appliance usage habits of residential customers. This results in insufficient accuracy and specificity in predicting safety hazards, making it difficult to effectively prevent electrical accidents caused by human error. In this field, the core challenges mainly focus on the following technical factors:

[0004] First, the multi-dimensional integration and analysis of customer characteristic information is insufficient, making it difficult to construct a comprehensive profile of customer electricity consumption behavior.

[0005] Secondly, the real-time misoperation risk identification technology based on dynamic electricity consumption data is still immature and it is difficult to respond quickly to potential risks;

[0006] Finally, there is a lack of intelligent and personalized mechanisms for generating anti-misoperation guidance. Existing prompts are mostly static and generic, making it difficult to adapt to the specific needs of different scenarios.

[0007] These technical challenges directly restrict the accuracy and practicality of predicting potential electrical safety hazards, and urgently need to be overcome. Therefore, how to build a dynamic electricity behavior analysis model based on customer characteristic information and integrate multi-dimensional data, and combine it with real-time risk identification technology to generate personalized guidance to prevent misoperation, has become a key issue in improving the effectiveness of predicting potential electrical safety hazards. Summary of the Invention

[0008] To address the aforementioned technical issues, a method for predicting potential electricity safety hazards based on customer characteristic information is proposed. This method includes acquiring customer characteristic information and real-time electricity consumption data, and performing preprocessing to obtain a multi-dimensional data set.

[0009] Aggregate and model the dataset to generate customer electricity consumption behavior profiles and construct electricity consumption behavior patterns;

[0010] Based on electricity consumption behavior profiles and patterns, dynamic behavior analysis is performed on real-time electricity consumption data to identify current trends in electricity consumption behavior.

[0011] The changing trends are compared with preset behavioral patterns to determine whether there are any potential anomalies;

[0012] If an anomaly is found, risk identification and processing will be performed to determine the risk level and corresponding response parameters, and a guidance plan for preventing misoperation will be generated.

[0013] The adaptability of the anti-misoperation guidance plan is assessed, and the final guidance content is generated based on the assessment results;

[0014] The final guidance content is pushed to the client's terminal, feedback information is recorded, behavioral profiles and guidance strategies are updated, and the results of dynamic behavior analysis and risk identification are optimized.

[0015] As a preferred embodiment of the method for predicting potential electricity safety hazards based on customer characteristic information according to the present invention, the step of aggregating and modeling the data set to generate a customer electricity behavior profile and constructing an electricity behavior pattern includes,

[0016] Based on the multidimensional dataset, the structure of customer characteristic information and real-time electricity consumption data is unified and feature extraction is performed. Feature information reflecting differences in user behavior is extracted and sample segmentation is performed to distinguish customer categories with different electricity consumption behavior characteristics.

[0017] For each customer category, generate corresponding feature descriptions to form a distinctive behavioral profile.

[0018] The beneficial effects of this optimized method are as follows: By performing field filtering and structural unification on the original multidimensional dataset, the standardization and consistency of subsequent data processing are significantly improved. Based on this, the introduction of dimensionality reduction transformation and behavioral clustering modeling effectively extracts core features reflecting differences in customer electricity consumption behavior, enhancing the expressive power and adaptability of behavioral profiles to customer group differences. The final constructed customer electricity consumption behavior profile has clear category labels and feature structures, facilitating the accuracy and targeting of subsequent dynamic behavior analysis, abnormal trend identification, and strategy generation, effectively improving the system's behavioral modeling efficiency and profile stability in multi-customer scenarios.

[0019] As a preferred embodiment of the method for predicting potential electricity safety hazards based on customer characteristic information according to the present invention, the step of performing dynamic behavior analysis on real-time electricity consumption data and identifying the changing trend of current electricity consumption behavior includes:

[0020] Real-time electricity consumption data is constructed into a structured behavioral data sequence in chronological order;

[0021] The behavioral data sequence is processed continuously to construct a data structure that reflects the evolution of behavior, and compared with a preset behavioral pattern to identify behavioral segments that deviate from the pattern.

[0022] Comparative data representing the changes in deviant behaviors are extracted to form behavioral trend analysis results.

[0023] The advantages of this preferred method are as follows: by constructing real-time electricity consumption data into a time-ordered dataset and combining it with data cleaning and time series modeling, it can effectively reconstruct customers' electricity consumption behavior patterns over continuous time periods. Furthermore, through trend decomposition and smoothing techniques, it clearly expresses potential periodic changes and behavioral trends within the data. This method, through difference extraction and behavioral deviation analysis, achieves timely perception and structured attribution of short-term abnormal changes, improving the responsiveness and analytical accuracy of dynamic behavior analysis in response to fluctuations in electricity consumption behavior.

[0024] As a preferred embodiment of the method for predicting potential electrical safety hazards based on customer characteristic information according to the present invention, the method for generating anti-misoperation guidance includes:

[0025] Based on the comparison between the results of behavioral trend analysis and the preset behavioral patterns, the risk level and response parameters are determined, and the strategy combination elements that meet the conditions are selected from the preset set of strategy rules.

[0026] The strategy combination elements are structurally organized to construct the strategy expression content that responds to the current risk scenario, and matched with the electricity consumption behavior profile to generate an initial guidance plan to prevent misoperation.

[0027] As a preferred embodiment of the method for predicting potential electrical safety hazards based on customer characteristic information according to the present invention, the step of performing an adaptability judgment on the anti-misoperation guidance scheme and generating final guidance content based on the judgment result includes:

[0028] Acquire real-time electricity consumption data for the current time period, and perform format conversion and consistency processing to form a standardized data set;

[0029] When the standardized dataset meets the suitability assessment requirements, a suitability assessment operation is performed based on the correspondence between the standardized dataset and the initial guidance scheme.

[0030] Based on the suitability assessment results, the content was adjusted to obtain updated guidance content.

[0031] The updated guidance will be output as the final guideline for preventing misoperation.

[0032] The beneficial effects of this optimized method are as follows: By introducing standardized processing and similarity matching mechanisms, the anti-misoperation guidance scheme can achieve structural uniformity and content adaptation in different customer power consumption environments, enhancing the transferability and applicability of strategy generation. Combining decision tree algorithms to complete the applicability evaluation of the guidance scheme and dynamically adjusting the strategy content through optimization modules helps improve the matching degree between the guidance content and the actual customer behavior scenarios. The final output guidance content has clear execution directions in terms of format and structure, which can more effectively support the timely delivery and accurate execution of personalized anti-misoperation prompts, ensuring that the guidance strategy has practical operability and system closed-loop characteristics in complex power consumption scenarios.

[0033] As a preferred embodiment of the method for predicting potential electrical safety hazards based on customer characteristic information described in this invention, the steps include: pushing the final guidance content to the customer terminal, recording feedback information, updating the behavioral profile and guidance strategy, and optimizing the dynamic behavior analysis and risk identification results.

[0034] Obtain the final guidance content generated at the current stage;

[0035] Based on the communication identifier of the corresponding client terminal, the guidance content is sent through a preset push channel;

[0036] After the push notification is sent, record the execution status information of the push action;

[0037] The feedback content returned by the customer terminal and subsequent electricity consumption behavior data are extracted in a structured manner to form a feedback data set;

[0038] Based on the feedback data set, update the customer's corresponding electricity consumption behavior profile and the parameters of the anti-misoperation guidance strategy;

[0039] The updated electricity consumption behavior profile and strategy parameters are integrated and processed, and the processing parameters of the dynamic behavior analysis process and risk identification algorithm are adjusted.

[0040] As a preferred embodiment of the method for predicting potential electrical safety hazards based on customer characteristic information according to the present invention, the steps of pushing the final guidance content to the customer terminal, recording feedback information, updating the behavioral profile and guidance strategy, and optimizing the dynamic behavior analysis and risk identification results further include:

[0041] Receive feedback information from customer terminals and subsequent electricity consumption behavior data;

[0042] The subsequent electricity consumption behavior data is cleaned and restructured, user behavior characteristics are extracted, customer electricity consumption behavior profiles and guidance strategy parameters are updated, and it is determined whether there is a deviation between the updated profile and the current strategy configuration.

[0043] If any deviation is found, the strategy content is revised, updated anti-misoperation guidance content is generated, and sent to the client terminal;

[0044] By combining the latest electricity consumption data from customer terminals, the dynamic behavior analysis model and risk identification parameters are updated.

[0045] Another objective of this invention is to provide a system for predicting potential electrical safety hazards based on customer characteristic information. This invention solves the problem in the prior art of lacking dynamic behavior modeling and risk identification based on personalized customer behavior data, especially the technical difficulties of poor adaptability and insufficient feedback and update capabilities in the generation and adaptive optimization of anti-misoperation strategies.

[0046] As a preferred embodiment of the electricity safety hazard prediction system based on customer characteristic information described in this invention, it is characterized by comprising: a data acquisition and preprocessing module, a profile modeling and behavior pattern construction module, a dynamic behavior analysis module, an anomaly detection and risk identification module, a guidance generation and adaptation judgment module, and a push execution and system optimization module.

[0047] The data acquisition and preprocessing module acquires customer characteristic information and real-time electricity consumption data, and performs preprocessing to obtain a multi-dimensional data set.

[0048] The profile modeling and behavior pattern construction module aggregates and models the data set to generate customer electricity consumption behavior profiles and constructs electricity consumption behavior patterns.

[0049] The dynamic behavior analysis module performs dynamic behavior analysis on real-time data based on profiles and patterns to identify the changing trends of current electricity consumption behavior.

[0050] The anomaly detection and risk identification module compares the changing trend with the preset behavior pattern to determine whether there is an anomaly. If there is, it identifies the risk level and parameters.

[0051] The guidance generation and adaptation judgment module generates a guidance scheme to prevent misoperation, performs an adaptation judgment, and generates the final guidance content.

[0052] The push execution and system optimization module pushes the final guidance content to the client terminal, records feedback information, updates the behavior profile and guidance strategy, and optimizes the dynamic behavior analysis and risk identification results.

[0053] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method for predicting potential electrical safety hazards based on customer characteristic information.

[0054] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a method for predicting potential electrical safety hazards based on customer characteristic information.

[0055] The beneficial effects of this invention are as follows: This invention constructs a profile of electricity consumption behavior by collecting customer electricity consumption data and feature information, enabling dynamic analysis and risk identification of electricity consumption behavior. This invention integrates multi-dimensional data using principal component analysis and clustering algorithms, employs time series analysis for short-term trend prediction, and combines anomaly detection algorithms to identify potential operational risks. For identified risks, this invention generates personalized anti-operation guidance based on a rule base and machine learning model, and optimizes the guidance scheme through real-time stream processing technology. This invention uses a distributed message queue to push real-time prompts to customers and continuously optimizes the model using reinforcement learning and online learning technologies, achieving intelligent prediction and prevention of potential electricity safety hazards, and improving the safety and reliability of power grid operation. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0057] Figure 1 This is a flowchart illustrating an embodiment of the present invention of a method for predicting potential electrical safety hazards based on customer characteristic information. Detailed Implementation

[0058] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0059] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for predicting potential electrical safety hazards based on customer characteristic information, including:

[0060] S1. Obtain customer characteristic information and electricity consumption data, and preprocess them to obtain a multi-dimensional data set.

[0061] Customer characteristic information and electricity consumption data are obtained from smart grid and IoT devices, including device operation procedures, appliance usage habits and historical electricity consumption records. Through data cleaning and standardization, a multi-dimensional data set is obtained.

[0062] The system acquires customer characteristic information, electricity consumption data, equipment operation procedures, appliance usage habits, and historical electricity consumption records from smart grid and IoT devices to generate a raw dataset. If the raw dataset contains missing or outlier values, it is imputed using the mean or outlier records are removed to obtain a cleaned dataset. Based on the cleaned dataset, the Z-score standardization method is used to convert customer characteristic information, electricity consumption data, and historical electricity consumption records into a unified dimension, generating a standardized dataset. The K-means clustering algorithm is used to analyze appliance usage habits and historical electricity consumption records in the standardized dataset to obtain user electricity consumption pattern classifications. If the electricity consumption fluctuation of a certain category in a user electricity consumption pattern classification exceeds a preset threshold, time series analysis is used to extract the equipment operation procedure characteristics of that user category to identify electricity consumption anomalies. Based on these anomalies, a decision tree algorithm is used to perform correlation analysis on customer characteristic information and appliance usage habits to determine the influencing factors of abnormal electricity consumption. Through these influencing factors, a multidimensional dataset is generated, containing user electricity consumption patterns, anomalies, and influencing factors.

[0063] For example, smart grids and IoT devices collect customer electricity consumption information through sensors and communication modules, providing a data foundation for user behavior analysis. Customer characteristic information includes age, number of family members, and house size; electricity consumption data includes real-time power and voltage; device operation processes record the on / off times of appliances; appliance usage habits reflect usage preferences for air conditioners, water heaters, etc.; historical electricity consumption records include monthly or annual electricity consumption. The original dataset may contain missing or outlier values ​​due to equipment failure or transmission interruption.

[0064] For example, a user's daily electricity consumption suddenly increases to 1000 kWh, far exceeding the normal range. When cleaning the data, missing values ​​can be filled using the user's average consumption over the past 7 days, such as 50 kWh, or abnormal records can be deleted to ensure data reliability. After cleaning, the data dimensions are unified, making it suitable for subsequent analysis. Z-score standardization converts data with different dimensions into standard values ​​with a mean of 0 and a standard deviation of 1, eliminating the influence of dimensions.

[0065] For example, a user with an average daily electricity consumption of 50 kWh and a house area of ​​100 square meters can have their values ​​standardized to 0.5 and 0.8 respectively, facilitating cluster analysis. The K-means clustering algorithm categorizes users into energy-saving, regular, and high-consumption types based on their appliance usage habits and historical electricity usage records.

[0066] For example, energy-saving users consume less electricity and use air conditioning less at night; high-consumption users frequently use high-power equipment. After clustering, the electricity consumption of a certain high-consumption user fluctuates by more than 20%, triggering time series analysis. Time series analysis extracts the characteristics of equipment operation processes to identify abnormal electricity consumption points.

[0067] For example, a user's air conditioner ran continuously in the early morning, causing peak electricity consumption. Analyzing its on / off time series confirmed that the anomaly occurred between 2 AM and 4 AM. The decision tree algorithm further correlated customer characteristics and appliance usage habits to determine the cause of the anomaly.

[0068] For example, decision trees show that users with larger houses and more family members are more likely to experience abnormal electricity consumption due to improper air conditioning settings. Multidimensional datasets integrate user electricity consumption patterns, anomalies, and influencing factors.

[0069] For example, the anomaly for high-consumption user A is the operation of the air conditioner in the early morning. Influencing factors include a house area of ​​200 square meters, a family of four, and the habit of using a constant-temperature air conditioner. This data set can be used to optimize power grid dispatch and reduce peak load.

[0070] Preferably, the power supply company can send energy-saving suggestions, such as adjusting the timer for air conditioners to shut off, to reduce abnormal power consumption.

[0071] It should be noted that this method, through data-driven analysis, not only improves electricity efficiency but also saves costs for users and enhances grid stability.

[0072] In one embodiment, if anomalies occur frequently, the operating mode of home appliances can be dynamically adjusted by combining real-time feedback from IoT devices.

[0073] For example, a smart meter can automatically reduce the power of an air conditioner after detecting an anomaly.

[0074] For example, multidimensional datasets can also support personalized electricity pricing strategies, where high-consumption users pay higher fees during peak hours, incentivizing energy-saving behavior.

[0075] Understandably, the aforementioned technologies, through data cleaning, standardization, clustering, sequence analysis, and decision trees, construct a complete link from data collection to anomaly diagnosis, significantly improving the level of refined management of smart grids.

[0076] S2. Perform aggregation modeling on the data set to generate customer electricity consumption behavior profiles and construct electricity consumption behavior patterns.

[0077] Based on the multidimensional dataset, the structure of customer characteristic information and real-time electricity consumption data is unified and feature extraction is performed. Feature information reflecting differences in user behavior is extracted and sample segmentation is performed to distinguish customer categories with different electricity consumption behavior characteristics.

[0078] For each customer category, generate corresponding feature descriptions to form a distinctive behavioral profile.

[0079] A preferred embodiment of this invention, which uses aggregated modeling to generate customer electricity consumption behavior profiles, involves: acquiring customer characteristic information and electricity consumption data from smart grids and IoT devices to generate an original dataset. If the original dataset contains missing values, it is processed using the mean imputation method to obtain a cleaned dataset. Principal component analysis is used to reduce the dimensionality of the cleaned dataset, extracting the main feature vectors to generate a feature dataset. K-means clustering is used to classify the electricity consumption behaviors in the feature dataset, resulting in a set of customer group behavior patterns. If a certain type of electricity consumption behavior in the behavior pattern set does not match a preset threshold, time series analysis is used to extract the time features of that type of behavior to identify abnormal behavior points. Based on the abnormal behavior points, an association rule mining algorithm is used to analyze customer characteristic information and electricity consumption behavior to obtain a set of factors influencing abnormal behavior. A customer electricity consumption behavior profile is generated using this set of factors influencing abnormal behavior, thus determining the customer group's electricity consumption behavior patterns.

[0080] For example, smart grids and IoT devices collect customer electricity consumption information in real time through sensors and communication modules, generating a raw dataset. The raw dataset includes customer characteristic information such as age, occupation, and household income, as well as electricity consumption data such as real-time power and daily electricity consumption.

[0081] For example, a user's household income is 100,000 yuan per year, and the average daily electricity consumption is 30 kWh. The data is collected by smart meters and home appliance sensors.

[0082] It should be noted that the original data may be missing due to equipment failure or network interruption.

[0083] For example, a user's electricity consumption record for a certain day is missing.

[0084] In one possible implementation, mean imputation can be used to fill missing values ​​with the user's average daily electricity consumption over the previous 7 days, such as 25 kWh, ensuring data integrity. After cleaning the dataset and imputing the mean, data consistency is improved, but the dimensionality is high, containing multiple variables such as electricity consumption and house area. Principal component analysis is used for dimensionality reduction, extracting key feature vectors and reducing data complexity.

[0085] Specifically, principal component analysis linearly combines the original variables into a few comprehensive features, such as electricity intensity and lifestyle factors.

[0086] For example, a user with a daily electricity consumption of 30 kWh and a house area of ​​120 square meters, after principal component analysis, generates a main feature vector, represented by an electricity intensity factor of 0.6 and a lifestyle habit factor of 0.4, retaining 90% of the information for subsequent analysis. Based on the dimensionality reduction results, the feature dataset is used to classify electricity consumption behavior using the K-means clustering algorithm, generating a set of customer group behavior patterns.

[0087] For example, clustering categorizes users into low-consumption, balanced, and high-consumption types. Low-consumption users consume less than 20 kWh per day, balanced users consume approximately 30 kWh, and high-consumption users consume more than 50 kWh.

[0088] In one embodiment, a high-consumption user has a daily electricity consumption of 60 kWh, mainly driven by an electric heater at night.

[0089] It should be noted that the clustering results reflect differences in user behavior, providing a basis for personalized management. If a certain type of electricity consumption behavior deviates from a preset threshold, such as a high-consumption user's electricity consumption fluctuating by more than 30%, time series analysis is used to extract time features and locate abnormal behavior points.

[0090] Preferably, time series analysis focuses on the temporal distribution of electricity consumption.

[0091] For example, a user's electricity consumption suddenly increased to 80 kWh at night. Analysis showed that the anomaly occurred between 1 a.m. and 3 a.m., which was caused by the electric heater running for a long time.

[0092] Understandably, time features help pinpoint abnormal time periods. Association rule mining algorithms further analyze customer characteristics and electricity consumption behavior to generate a set of factors influencing abnormal behavior.

[0093] Specifically, the algorithm identifies the correlation between age, income, and abnormal electricity consumption.

[0094] For example, the rules show that users with higher incomes and those over 40 years of age are more likely to have abnormal nighttime electricity consumption due to improper settings of electric heaters.

[0095] In one embodiment, a user with an annual income of 120,000 yuan exhibits abnormal nighttime electricity consumption that is related to their preference for high-temperature settings. The customer's electricity consumption behavior profile is generated by integrating user characteristics and behavioral patterns based on a set of factors influencing abnormal behavior.

[0096] For example, high-consumption user B is 45 years old, has an annual income of 150,000 yuan, and frequently uses electric heaters at night. Their electricity consumption profile shows a preference for high-power devices. This profile supports refined grid management, such as pushing energy-saving suggestions or optimizing electricity pricing strategies.

[0097] In one optional embodiment of this invention, aggregation modeling generates customer electricity consumption behavior profiles by: acquiring historical electricity consumption data and customer characteristic files from a smart grid and IoT platform. The customer characteristic files include apartment size, equipment type list, seasonal electricity consumption preference identifiers, and resident population structure. The raw data is then uniformly encoded to form a raw data set containing timestamps, equipment power, operating status markers, and user category tags.

[0098] During processing, if there are null fields due to device disconnection or data transmission interruption, the nearest neighbor interpolation method is used to reconstruct the data. Specifically, the null values ​​are filled by referring to the multi-day average power values ​​of the same device within the same period, forming a filled data set. To improve the efficiency of subsequent calculations, the high-frequency discrete features (such as device switching frequency) in the filled data set are normalized, and all fields are uniformly mapped to the [0,1] interval to form a standardized data set.

[0099] During data dimensionality reduction, a feature selection method based on information gain scoring is employed to score and rank the dependencies between various customer feature fields and target behavioral variables (such as average power within peak-valley electricity price ranges), thereby identifying high-contribution features and constructing a compact feature subset. For example, the selected dominant features include the equipment proportion coefficient, seasonal preference index, and weekday / holiday load variation.

[0100] Based on the aforementioned feature subset, a density-based spatial clustering method is used to group customer samples, forming a set of customer behavior patterns. This clustering method does not pre-determine the number of categories but dynamically divides customer groups according to density connectivity rules. For example, the clustering results divide target users into three categories: "seasonally sensitive," "load-uneven," and "stable and energy-efficient."

[0101] For example, the average daily load of a type of "seasonally sensitive" user is more than 60% higher than that of a user in winter, and its cooling load is mainly concentrated in the afternoon from 14:00 to 17:00; "load imbalance" users show obvious characteristics of concurrent operation of electrical equipment, with the difference between the load values ​​of the morning peak and the evening peak reaching more than 2 times.

[0102] For each customer behavior group, an aggregated feature representation vector is further constructed. This vector includes structural indicators such as the main cycle of electricity consumption, peak position, and device activity matrix. This structured vector is stored in the behavior profile database and bound to the user's unique identifier to form the final customer electricity consumption behavior profile.

[0103] The behavioral profile can be used in subsequent tasks such as electricity consumption trend monitoring, strategy matching, and abnormal behavior identification, supporting a more targeted personalized strategy generation and push process.

[0104] The beneficial effect of this preferred embodiment is that by constructing a customer behavior modeling process based on principal component analysis and K-means clustering algorithm, it is possible to effectively aggregate and model customer electricity consumption data with high dimensional differences within a unified feature space; this method has good dimensionality compression capability and customer grouping stability when processing multidimensional datasets containing multiple social attributes and device behavior data.

[0105] By introducing a combined processing mechanism of time series analysis and association rule mining, the modeling process is not limited to the current behavior classification, but can also further identify abnormal electricity consumption points and potential causes within the group behavior. This enables a two-dimensional expression of "behavioral category + behavioral risk factor" in the customer profile, enhancing the structural integrity and interpretability of the profile.

[0106] S3. Based on behavioral profiles, perform dynamic behavioral analysis on real-time electricity consumption data to identify the changing trends of current electricity consumption behavior.

[0107] Real-time electricity consumption data is constructed into a structured behavioral data sequence in chronological order;

[0108] The behavioral data sequence is processed continuously to construct a data structure that reflects the evolution of behavior, and compared with a preset behavioral pattern to identify behavioral segments that deviate from the pattern.

[0109] Comparative data representing the changes in deviant behaviors are extracted to form behavioral trend analysis results.

[0110] In a preferred embodiment of the present invention, dynamic behavior analysis is performed on real-time electricity consumption data.

[0111] Real-time electricity consumption data is acquired from smart grid devices to generate a raw dataset. Data cleaning techniques are used to preprocess the raw dataset, removing outliers to obtain a cleaned dataset. Time series analysis is then used to decompose the cleaned dataset, extracting periodic and trend features to generate a feature dataset. Based on the feature dataset, a moving average algorithm is used to smooth electricity consumption behavior, resulting in a smoothed dataset. If the electricity consumption in a certain time period within the smoothed dataset deviates from a preset threshold, difference analysis is used to extract short-term variation characteristics for that period to determine the short-term trend. Based on the short-term trend, correlation analysis is used to mine the relationship between customer behavior patterns and the short-term trend, resulting in a behavior pattern set. Using the behavior pattern set, dynamic analysis results of customer electricity consumption behavior are generated to determine dynamic behavioral characteristics.

[0112] For example, real-time electricity consumption data is acquired from smart grid devices to generate a raw data set.

[0113] For example, a smart meter records household electricity consumption every 15 minutes, including information such as voltage, current, and power factor, generating a raw dataset containing timestamps and electricity usage data. Data cleaning techniques are then used to preprocess the raw dataset, removing outliers to obtain a cleaned dataset.

[0114] Specifically, abnormal values ​​may include negative values ​​caused by meter malfunctions or sudden extremely high values.

[0115] In one possible implementation, outliers are identified using a box plot method, with upper and lower bounds set to 1.5 times the interquartile range, and values ​​outside the range are removed.

[0116] For example, if a household's daily electricity consumption suddenly increases to 1000 kWh, far exceeding the normal range of 10-50 kWh, this value is removed and replaced with the average of nearby time points. This method ensures data reliability and facilitates subsequent analysis. Time series analysis techniques are used to decompose the cleaned dataset, extracting periodic and trend features to generate a feature dataset.

[0117] For example, time series decomposition can employ an additive model to split data into trend, seasonal, and residual components.

[0118] For example, a household's weekly electricity consumption data shows a cyclical pattern of morning and evening peaks, with the trend indicating slightly higher consumption on weekends. After decomposition, the cyclical features can be used to identify daily electricity consumption patterns, while the trend features reflect long-term changes, providing a basis for behavioral analysis. Based on the feature dataset, a moving average algorithm is used to smooth the electricity consumption behavior, resulting in a smoothed dataset.

[0119] Preferably, a 7-day window can be used for the moving average to calculate the average daily electricity consumption to smooth out short-term fluctuations.

[0120] For example, if a household's electricity consumption surges to 60 kWh on a certain day due to temporary activities, a smoothed value of 40 kWh after moving average will better reflect the normal level. This method reduces noise interference and highlights the main trend. If the electricity consumption in a certain period of the smoothed dataset deviates from a preset threshold, short-term variation characteristics of that period are extracted through differential analysis to determine the short-term trend.

[0121] It should be noted that the preset threshold can be based on the historical mean plus or minus two standard deviations.

[0122] For example, if a household's smoothed electricity consumption reaches 70 kWh one evening, exceeding the threshold of 50 kWh, first-order difference calculations reveal that the rate of increase in electricity consumption during that period was 30% higher than the previous day, indicating a short-term abnormal increase. Difference analysis helps to accurately pinpoint the points of change, facilitating the identification of abnormal behavior. Based on short-term trends, correlation analysis techniques are used to uncover the relationship between customer behavior patterns and these short-term trends, resulting in a set of behavior patterns.

[0123] In one embodiment, association analysis can use the Apriori algorithm to uncover the relationship between peak electricity consumption and customer behavior.

[0124] For example, it was found that peak household electricity consumption was frequently associated with evening air conditioning use, with 80% support and 90% confidence. This result reveals the driving factors of behavior and provides a basis for accurate profiling. By using a set of behavioral patterns, dynamic analysis results of customer electricity consumption behavior are generated to determine dynamic behavioral characteristics.

[0125] Understandably, the dynamic analysis results can be visualized to show the correspondence between daily electricity consumption curves and behavioral patterns.

[0126] For example, a household's dynamic characteristics can show the difference in electricity consumption between weekday morning and evening peak hours and weekend distributed throughout the day, reflecting their lifestyle habits. This method helps power grid companies optimize resource allocation and improve service efficiency.

[0127] In one embodiment of the present invention, performing dynamic behavior analysis on real-time electricity consumption data is as follows:

[0128] The system receives real-time electricity consumption data packets uploaded by customer terminals from the smart grid system. Each data packet includes indicators such as the collection timestamp, equipment operating status, power reading, and voltage frequency. The continuously received data packets are arranged in time sequence to generate a real-time electricity consumption data set with a time resolution of 5 minutes.

[0129] In the preprocessing stage, noise suppression is performed using a median filtering method based on a sliding window to address device state conflicts (such as multiple on / off records of the same device within a unit of time) and acquisition drift values ​​(such as continuous fluctuations in power readings exceeding the physical upper limit of change rate) in the dataset, forming a preliminary cleaned dataset.

[0130] Subsequently, the Trend-Enhanced Decomposition (TED) method was used to perform time-series modeling on the cleaned dataset, decomposing the original electricity consumption sequence into three parts: a trend component, a periodic component, and a daily cycle component. The trend component characterizes the overall direction of behavioral changes, the cycle component captures repetitive electricity consumption patterns, and the periodic component retains unstructured abnormal changes. This decomposition structure generates an analytical data structure set.

[0131] A weighted exponential smoothing method is applied to the trend items in the analyzed data structure to enhance the ability to express the dominant trend and suppress spurious changes caused by short-term oscillations, resulting in a smoothed behavioral sequence. A floating threshold range is set based on the behavioral sequence; if a behavioral segment deviates from the trend boundary, it is identified as a possible period of abnormal electricity consumption.

[0132] For abnormal periods, the difference sequence between the abnormal period and the corresponding period in the previous cycle is calculated. Short-term dynamic features, including the maximum amplitude of change, the duration of the change direction, and the fluctuation rate, are extracted, and a set of difference feature vectors is constructed. The difference feature vectors are matched item by item with the equipment behavior templates in the customer profile to extract behavioral factors related to changes in equipment operating status and time point selection.

[0133] Finally, attribution analysis is performed on all differential feature vectors using aggregated decision logic (such as weighted voting mechanism) to form the dynamic behavior analysis results of the customer in the current period. The results include the main cause category of behavior change (such as "sudden change in equipment load" or "misaligned use of the cycle"), time location information, and a set of influencing factor fields.

[0134] The dynamic analysis results will be passed as input to the risk identification module and the strategy generation module to build personalized anti-misoperation strategies.

[0135] It should be further explained that:

[0136] This preferred embodiment constructs a dynamic behavior recognition process that integrates time series modeling, sliding smoothing, and behavioral correlation analysis. This process can stably extract short-term customer electricity consumption behavior trends without increasing computational complexity, and quickly locate potential deviation periods and their dominant factors based on behavioral profiles. This method supports daily-level dynamic monitoring in typical household electricity consumption scenarios, and, in conjunction with a strategy module, achieves closed-loop linkage between behavior and strategy.

[0137] S4. Compare the changing trend with the preset behavior pattern to determine if there are any potential anomalies.

[0138] Based on the comparison between the results of behavioral trend analysis and the preset behavioral patterns, the risk level and response parameters are determined, and the strategy combination elements that meet the conditions are selected from the preset set of strategy rules.

[0139] The strategy combination elements are structurally organized to construct the strategy expression content that responds to the current risk scenario, and matched with the electricity consumption behavior profile to generate an initial guidance plan to prevent misoperation.

[0140] If the trend of change exceeds the preset mode, an anomaly detection algorithm is used to analyze the data to obtain the risk level of misoperation. Based on the risk level, the probability of occurrence is obtained to determine the potential impact range. Based on the probability of occurrence, dynamic features are extracted from electricity consumption behavior to determine the direction of behavior adjustment. Time series decomposition technology is used to process the dynamic features to obtain a set of feature changes. If the set of feature changes shows a continuous deviation, correlation analysis is used to explore the relationship between electricity consumption behavior and the trend of change to determine the adjustment priority. Based on the adjustment priority, smoothing technology is used to process the electricity consumption behavior data to obtain a set of stable behaviors. Based on the set of stable behaviors, it is determined whether the short-term changes have returned to the preset mode to determine the final state.

[0141] For example, anomaly detection algorithms are used in electricity consumption behavior analysis to identify abnormal fluctuations that exceed preset patterns. The core of anomaly detection lies in using statistical or machine learning methods to determine whether data points deviate from the normal behavior range.

[0142] One possible implementation is to use an isolated forest-based algorithm to analyze the cleaned electricity consumption data and identify anomalies.

[0143] For example, if a customer's electricity consumption suddenly increases to 500 kWh at night, while the historical average is 50 kWh, the algorithm marks this as an anomaly and generates a risk level for the misoperation. The risk level is divided into three levels: low, medium, and high, based on the degree of deviation and duration of the anomaly. A sudden increase at night may be rated as high risk.

[0144] Specifically, the risk level is combined with the probability of occurrence to further analyze the potential scope of impact.

[0145] In one embodiment, the probability of occurrence is determined by statistical analysis of historical data. If a high-risk anomaly occurs three times in the past 30 days, the probability is 10%. The scope of impact is determined by customer type; for example, anomalies in industrial customers may affect production scheduling, while those in residential customers may involve equipment malfunctions. Assuming an anomaly probability of 10% for an industrial customer, the impact might involve production line shutdowns, requiring priority attention.

[0146] Understandably, dynamic feature extraction determines the direction of behavior adjustment based on the probability of occurrence.

[0147] For example, a high-probability anomaly might indicate equipment aging or malfunction, and the adjustment should be to optimize the power usage plan. A customer experiences frequent abnormalities at night, with dynamic characteristics showing peak power consumption concentrated between 11:00 PM and 1:00 AM. The adjustment should be to stagger power usage, suggesting shifting some loads to the daytime.

[0148] Preferably, time series decomposition techniques process dynamic features to generate a set of feature changes. The decomposition technique breaks down the data into trend, seasonal, and residual components.

[0149] For example, after breaking down a customer's weekly electricity consumption data, the trend shows a daily increase of 10 kWh, and the seasonality shows peak electricity consumption on weekends. The feature variation set records these patterns for subsequent analysis.

[0150] It should be noted that if the set of feature changes shows a continuous deviation, correlation analysis should be used to explore the relationship between electricity consumption behavior and the changing trend to determine the adjustment priority.

[0151] In one possible implementation, correlation analysis reveals that weekend peak traffic is related to overtime work on the production line, thus having a high priority and requiring adjustments to the production plan. Conversely, occasional anomalies have a low priority and only require monitoring.

[0152] For example, smoothing techniques process electricity consumption data to generate a stable set of behavior patterns. Moving average algorithms are commonly used in this process. Suppose a customer's electricity consumption fluctuates significantly over seven consecutive days; after smoothing, the data tends to stabilize, showing a daily average of approximately 200 kilowatt-hours. This stable set of behavior patterns reflects behavioral regularities, facilitating subsequent analysis.

[0153] In one embodiment, the final state is determined by judging whether short-term changes have recovered to a preset pattern through a set of stable behaviors.

[0154] For example, if a customer's adjusted electricity consumption returns to 180-220 kWh per day, which conforms to the preset pattern, the status is normal. If it still deviates, for example, consistently exceeding 300 kWh, the status is abnormal, requiring further intervention. These steps are interconnected to ensure accurate and efficient analysis.

[0155] S5. If an anomaly is found, perform risk identification processing, determine the risk level and corresponding response parameters, and generate a guideline for preventing misoperation.

[0156] Acquire real-time electricity consumption data for the current time period, and perform format conversion and consistency processing to form a standardized data set;

[0157] When the standardized dataset meets the suitability assessment requirements, a suitability assessment operation is performed based on the correspondence between the standardized dataset and the initial guidance scheme.

[0158] Based on the suitability assessment results, the content was adjusted to obtain updated guidance content.

[0159] The updated guidance will be output as the final guideline for preventing misoperation.

[0160] If the risk level and probability of occurrence exceed a preset threshold, the input data is standardized by the data processing module to obtain a standardized dataset. Using a pre-established rule base, rules related to the risk level are extracted from the standardized dataset to obtain a rule matching set. A machine learning model is used to analyze the rule matching set and the probability of occurrence to generate preliminary anti-misoperation strategies, resulting in a strategy set. Based on the strategy set and the attributes of the electrical appliances, operation suggestions for the appliances are generated, resulting in an operation suggestion set. If the operation suggestion set does not conform to the preset safety standards, the suggestion set is optimized by the data processing module to obtain an optimized suggestion set. Personalized guidance containing safety tips is generated from the optimized suggestion set, resulting in a final guidance set. Based on the final guidance set, it is determined whether the guidance content covers all relevant electrical appliances, thus determining the final output status.

[0161] In one possible implementation, scenarios where the risk level and probability of occurrence exceed a preset threshold typically occur during peak electricity consumption periods or when equipment is aging.

[0162] For example, a factory's power distribution system detects an abnormal current in a motor, with a high risk level assessment and an 80% probability of occurrence. The data processing module first standardizes the input data, such as current, voltage, and running time, removing the influence of dimensions to generate a standardized dataset. This step ensures that subsequent analyses are based on a uniform scale, reducing data bias.

[0163] For example, the current value is converted from the original 200 amps to a standardized 0.8 amps to facilitate rule matching.

[0164] Specifically, the rule base is built based on historical electricity consumption data and equipment characteristics, including rules such as "current fluctuations exceeding 30% for 5 minutes" as high-risk. Matching rules are extracted from the standardized dataset to obtain a rule matching set.

[0165] For example, if a motor current fluctuation of 35% is detected, a high-risk rule is matched, generating a set of rules including "equipment overload" and "potential short circuit". These rules provide a basis for subsequent strategy generation, ensuring the targeted nature of the analysis.

[0166] In one embodiment, the machine learning model uses a decision tree or random forest to analyze the set of rule matching and the probability of occurrence, and generates a preliminary strategy to prevent misoperation.

[0167] For example, for motor overload, the strategy set might include "reduce the load" or "pause operation for 10 minutes." The model learns from historical cases, identifies high-probability risk scenarios, and generates strategies that better match actual needs. This approach improves the accuracy of the strategies and reduces the possibility of misoperation.

[0168] Preferably, operation suggestions are generated by combining the attributes of electrical appliances, taking into account equipment power, operating environment, etc.

[0169] For example, if the motor has a power of 50 kilowatts and operates in a high-temperature environment, the set of operational suggestions might include "reduce the operating frequency to 60%" or "add heat dissipation devices." These suggestions directly affect equipment operation and enhance usability. If a suggestion does not comply with safety standards, such as reducing the frequency potentially causing production stoppages, the data processing module will optimize the suggestion.

[0170] For example, by analyzing historical data, adjustments can be made to "reduce frequency in different time periods" to generate a set of optimization suggestions.

[0171] For example, when generating personalized instructions, the final instruction set includes specific safety tips such as "Check motor temperature every hour" and "Set current threshold alarm." These instructions are intuitive and easy to understand, covering multiple aspects of motor operation.

[0172] It should be noted that when determining whether the guidance covers all relevant electrical appliances, it is necessary to check whether the guidance set includes all key equipment.

[0173] For example, confirm whether the guidelines cover motors, control cabinets, etc., to ensure nothing is omitted. This comprehensiveness ensures the integrity of the guidelines and improves the efficiency of safety management.

[0174] Understandably, the above process, progressing step by step from data standardization to guidance generation, forms a rigorous logical chain. The implementation methods of each step are closely centered on electricity consumption behavior analysis and prevention of misoperation, ensuring the relevance and operability of the solution.

[0175] For example, standardized processing improves data consistency, rule matching ensures accurate risk identification, and optimization suggestions enhance the practicality of the guidance. This multi-faceted, mutually supportive approach makes the final guidance both comprehensive and accurate, applicable to complex electricity usage scenarios.

[0176] S6. Perform an adaptability assessment on the anti-misoperation guidance scheme and generate the final guidance content based on the assessment results.

[0177] Obtain the final guidance content generated at the current stage;

[0178] Based on the communication identifier of the corresponding client terminal, the guidance content is sent through a preset push channel;

[0179] After the push notification is sent, record the execution status information of the push action;

[0180] The feedback content returned by the customer terminal and subsequent electricity consumption behavior data are extracted in a structured manner to form a feedback data set;

[0181] Based on the feedback data set, update the customer's corresponding electricity consumption behavior profile and the parameters of the anti-misoperation guidance strategy;

[0182] The updated electricity consumption behavior profile and strategy parameters are integrated and processed, and the processing parameters of the dynamic behavior analysis process and risk identification algorithm are adjusted.

[0183] Real-time electricity consumption data is acquired from an IoT data acquisition system. This data is preprocessed using stream processing technology to obtain a preprocessed dataset. A data cleaning module then denoises and converts the format of the preprocessed dataset to obtain a cleaned dataset. Data standardization is used to normalize the cleaned dataset to obtain a standardized dataset. If the standardized dataset matches the input requirements of the personalized anti-misoperation guidance, a similarity algorithm is used to calculate the similarity between the two, resulting in a matching score. Based on the matching score, a decision tree algorithm is used to evaluate the applicability of the personalized anti-misoperation guidance, yielding an applicability evaluation result. An optimization module adjusts the applicability evaluation result to generate an optimized guidance scheme. If the optimized guidance scheme covers all relevant electrical appliances, a formatting module generates the final guidance scheme, determining the final output state.

[0184] For example, IoT data acquisition systems use sensors and smart meters to acquire electricity consumption data in real time, providing a foundation for personalized guidance to prevent misoperation. In a residential community scenario, for instance, smart meters collect voltage, current, and power data every second, generating a data stream containing timestamps and device identifiers. Stream processing technology then preprocesses this data.

[0185] Specifically, tools such as Apache Flink are used to perform real-time data sharding and aggregation, filtering out data points with abnormal fluctuations to obtain a preprocessed dataset. Preferably, during preprocessing, the system removes invalid data with voltages below 50V to ensure data reliability. The data cleaning module performs noise reduction and format conversion on the preprocessed dataset.

[0186] In one possible implementation, denoising employs a sliding window technique to detect and remove abrupt changes in current data. For example, if the current of a household appliance jumps from 5A to 50A within a short period, this is identified as noise and removed. Format conversion unifies the heterogeneous data from different devices into JSON format, generating a cleaned dataset.

[0187] Understandably, the cleaning process ensures data consistency, providing high-quality input for subsequent analysis. Data standardization techniques normalize the cleaned dataset to obtain a standardized dataset.

[0188] For example, normalization uses the Min-Max method to map power data from 0-1000W to the 0-1 range. For instance, the power data of an air conditioner of 500W is normalized to 0.5. This standardization facilitates comparison of data between different devices.

[0189] It's important to note that the standardized dataset must match the input requirements of the personalized error prevention guidance. The matching algorithm calculates the similarity between the standardized dataset and the guidance template using cosine similarity, generating a matching score. For example, a similarity of 0.85 between a dataset and the guidance template indicates a high degree of match. A decision tree algorithm is used to evaluate the applicability of the personalized error prevention guidance.

[0190] In one embodiment, the decision tree determines the applicability of a guidance scheme based on features such as matching score, device type, and power consumption pattern. For example, for high-power devices like electric water heaters, the decision tree might prioritize guidance that reduces operating time. The applicability assessment results reflect the feasibility of the guidance scheme. The optimization module adjusts the assessment results to generate an optimized guidance scheme.

[0191] Specifically, optimization might adjust the order of operations in the guidance, such as prioritizing the shutdown of high-risk devices like electric ovens before adjusting low-risk devices like lights. If the optimized guidance covers all relevant electrical appliances, the formatting module generates the final guidance. For example, the formatting module converts the guidance into user-friendly text prompts, such as "It is recommended to turn off the electric water heater after 8:00 PM to reduce the risk of overload." The final output status is determined by checking coverage; if it covers all devices, the output takes effect.

[0192] Preferably, the system records the execution status of each guidance session, providing feedback for subsequent optimization. This approach, driven by real-time data, ensures the accurate application of guidance solutions, improving electricity safety and efficiency.

[0193] S7. Push the final guidance content to the customer terminal, record feedback information, update the behavior profile and guidance strategy, and optimize the dynamic behavior analysis and risk identification results.

[0194] Receive feedback information from customer terminals and subsequent electricity consumption behavior data;

[0195] The subsequent electricity consumption behavior data is cleaned and restructured, user behavior characteristics are extracted, customer electricity consumption behavior profiles and guidance strategy parameters are updated, and it is determined whether there is a deviation between the updated profile and the current strategy configuration.

[0196] If any deviation is found, the strategy content is revised, updated anti-misoperation guidance content is generated, and sent to the client terminal;

[0197] By combining the latest electricity consumption data from customer terminals, the dynamic behavior analysis model and risk identification parameters are updated.

[0198] Real-time electricity consumption data streams are acquired from the smart grid system. The data distribution module allocates these streams to a distributed message queue, resulting in a queue task set. If the number of tasks in the queue task set exceeds a preset threshold, a load balancing algorithm redistributes the tasks, resulting in a balanced task set. The message queue management module prioritizes the balanced task set, generating a priority task sequence. The status monitoring module monitors the execution status of the priority task sequence in real time, determining low-latency and high-reliability states to obtain the task execution status. If the task execution status meets preset conditions, the push generation module generates customer push content based on a real-time anti-misoperation prompt template, resulting in a push content set. Based on the push content set, a real-time prompt is sent to the customer terminal via the smart grid technology communication interface to confirm the push completion status. The log recording module stores the push completion status, generating a push task log, resulting in a log dataset.

[0199] For example, obtaining real-time electricity consumption data streams from smart grid systems is a core component of achieving dynamic electricity consumption management.

[0200] For example, smart grids collect data such as voltage, current and power factor in real time through sensors deployed at substations and user terminals, forming high-frequency data streams.

[0201] For example, smart meters in a residential community collect data once per second, generating a data packet containing a timestamp, electricity consumption, and device identifier. This high-frequency collection ensures the real-time nature of the data stream, providing a foundation for subsequent processing. A data distribution module allocates the data stream to a distributed message queue, aiming to improve the system's scalability.

[0202] In one possible implementation, the data distribution module is based on a Kafka message queue and distributes data streams by region or device type.

[0203] For example, the data stream from 1000 electricity meters in a residential community is divided into 10 queues, with each queue processing data from 100 meters. The distribution module allocates data based on the device ID using hashing to ensure even data distribution. This approach supports the system in processing large-scale data. If the number of tasks in a queue exceeds a preset threshold, a load balancing algorithm is used to reallocate the tasks.

[0204] Specifically, assuming the threshold is 5,000 messages per minute per queue, when a queue reaches 6,000 messages per minute, the load balancing algorithm will transfer the excess tasks to other low-load queues.

[0205] For example, using a consistent hashing algorithm, the excess 1000 tasks are redistributed to queues with lower loads. This dynamic adjustment ensures system stability. A message queue management module prioritizes the balanced task set, generating a priority task sequence.

[0206] Preferably, the power outages are ranked according to their urgency.

[0207] For example, a task that detects an abnormal power consumption of a certain electricity meter is assigned high priority, taking precedence over regular data processing tasks. The message queue management module uses priority tags to rank abnormal tasks at the top of the sequence, ensuring that critical tasks are processed first. The status monitoring module monitors the execution status of the priority task sequence in real time, determining low latency and high reliability.

[0208] In one embodiment, the status monitoring module checks the task execution time and error rate once per second.

[0209] For example, if a task takes more than 100 milliseconds to execute or has an error rate higher than 1%, it is marked as an abnormal state. This real-time monitoring ensures that the system detects problems promptly. If the task execution status meets preset conditions, the push generation module generates push content for the customer based on a real-time anti-misoperation prompt template.

[0210] Understandably, the push notification template includes the exception type, suggested actions, and a timestamp.

[0211] For example, if an abnormal power consumption is detected in a user's refrigerator, a push notification is generated stating, "Refrigerator power consumption is abnormal; it is recommended to check the device status." The push notification content is stored in JSON format for easy subsequent transmission. Real-time notifications are sent to the customer's terminal via a smart grid technology communication interface to confirm the push notification completion status.

[0212] For example, push notifications are sent to the user's mobile app via a 4G network. The app displays a notification and records the reception time. This method ensures that users are promptly notified of any abnormal situations. A log recording module stores the push notification completion status, generating a push task log.

[0213] In one possible implementation, the logging module stores the push time, content, and user ID in a distributed database.

[0214] For example, approximately 10,000 log entries are generated daily, including successful or failed push statuses, facilitating subsequent auditing and analysis. This log storage supports system operation and maintenance and problem tracing.

[0215] Feedback data and electricity consumption behavior data are obtained from customer terminals via a communication interface to obtain a raw data set. Based on this raw data set, a data cleaning module preprocesses the feedback data and electricity consumption behavior data to obtain a standardized data set. Using the standardized data set, a reinforcement learning algorithm is employed to extract features from the electricity consumption behavior data, resulting in a behavioral feature set. Based on this behavioral feature set, a behavioral profile generation module updates the electricity consumption behavior profile, resulting in an updated profile set. If the updated profile set does not match the preset thresholds in the guidance rule base, a rule adjustment module corrects the guidance rules, resulting in an adjusted rule set. Using the adjusted rule set, a parameter optimization module updates the model parameters, resulting in an optimized parameter set. Based on the optimized parameter set, updated anti-misoperation guidance content is pushed to customer terminals via a real-time communication interface, and the push completion status is confirmed.

[0216] Specifically, feedback data and electricity consumption behavior data are obtained from the customer terminal through a communication interface to form a raw data set.

[0217] For example, smart grid systems use 4G / 5G networks to obtain daily electricity consumption, voltage fluctuation records, and user feedback on anti-misoperation prompts from customers' smart meters, such as whether electricity consumption habits have been adjusted according to the prompts. This data may contain noise or missing values, therefore a data cleaning module is needed for preprocessing.

[0218] In one possible implementation, the data cleaning module interpolates missing values. For example, for missing electricity consumption data of a user, it infers and fills in the missing data based on the previous electricity consumption trend. Simultaneously, it removes outliers; for instance, a voltage record far exceeding the normal range is considered invalid data. After cleaning, a standardized data set is obtained, ensuring that the data format is consistent and suitable for subsequent analysis.

[0219] Preferably, reinforcement learning algorithms are used to extract electricity consumption behavior features based on standardized data sets.

[0220] For example, the algorithm analyzes users' electricity consumption during peak hours, the types of electrical devices they use, and their electricity usage habits mentioned in feedback to generate a set of behavioral features.

[0221] For example, a household experiences a surge in electricity consumption between 7 PM and 9 PM. The feature set might indicate that their primary electrical appliances are air conditioners and electric water heaters, and user feedback shows a positive response to suggestions to conserve electricity during peak hours. This suggests that the user is energy-conscious but lacks optimization strategies.

[0222] It should be noted that reinforcement learning algorithms continuously optimize feature extraction through trial and error, gradually focusing on the features most valuable for preventing misoperations.

[0223] In one embodiment, the behavior profile generation module updates the electricity consumption behavior profile based on the behavior feature set.

[0224] For example, the system generates a profile for the user mentioned above, labeling them as a "peak-time electricity-dependent" user who tends to respond to energy-saving prompts but needs more specific guidance. If the profile does not match the preset thresholds in the guidance rule base, such as if the user's energy-saving response rate is lower than expected, the rule adjustment module will revise the guidance rules.

[0225] For example, the system may adjust the original rule's suggestion of "reduce electricity consumption by 20% during peak hours" to "prioritize turning off the electric water heater and delaying its use," generating an adjusted rule set.

[0226] Understandably, the parameter optimization module updates the model parameters based on the adjusted rule set.

[0227] For example, for the aforementioned users, the system optimizes the generation model of push content, increases the personalization weight of push content, such as pushing energy-saving tips related to electric water heaters more frequently, and generates an optimized parameter set. Through a real-time communication interface, the system pushes updated anti-misoperation guidance content to the client terminal.

[0228] For example, a notification can be sent to a user's mobile app stating, "It is recommended to use the electric water heater after 9 PM to save 10% on electricity costs." After the push is completed, the system records the push completion status to ensure that the guidance content is delivered.

[0229] For example, the logging module stores push task logs, forming a log dataset.

[0230] In one embodiment, the log records the user ID, push time, content summary, and delivery status, such as "User A, April 21, 2025, 19:00, energy-saving reminder, delivered successfully." These logs facilitate subsequent analysis of user response effects and optimization of push strategies.

[0231] It should be noted that the above modules work together to form a closed loop, from data collection to push optimization, with each step closely linked to ensure the accuracy and timeliness of the guidance content.

[0232] Real-time electricity consumption behavior data is acquired from customer terminals via a communication interface to obtain a raw behavior dataset. Based on this raw dataset, a data preprocessing module cleans and standardizes the data to obtain a standardized behavior dataset. Using this standardized dataset, online learning technology is employed to extract features from a dynamic behavior analysis model, resulting in a behavior feature set. If the behavior feature set does not match a preset risk threshold, a risk identification algorithm is used to classify potential safety hazards, yielding hazard classification results. Based on these classification results, a parameter adjustment module iteratively updates the dynamic behavior analysis model, resulting in updated model parameters. Using these updated model parameters, a configuration generation module adjusts the configuration of the electricity safety hazard prediction system, resulting in an optimized system configuration. Based on this optimized system configuration, the updated safety hazard prediction content is pushed to customer terminals via a real-time communication interface, and the push completion status is confirmed.

[0233] For example, real-time electricity consumption behavior data can be obtained from the client terminal through a communication interface to obtain the raw behavior dataset.

[0234] For example, in a smart grid system, the communication interface can be an IoT gateway based on the MQTT protocol, which collects electricity consumption data from customer terminals such as smart meters in real time.

[0235] For example, a household user consumes 3.5 kilowatts of electricity during peak hours, including operational data from devices such as air conditioners and water heaters. This data is stored in time-series format, forming the original behavioral dataset.

[0236] It should be noted that data collection must ensure real-time performance and completeness to support subsequent analysis. Based on the original behavioral dataset, a data preprocessing module is used to clean and standardize the data, resulting in a standardized behavioral dataset.

[0237] Specifically, data cleaning includes removing outliers, such as erroneous records where power suddenly jumps to 10 kilowatts, and filling in missing values, such as interpolating to fill in missing data for a particular minute. Standardization normalizes the power data to the 0-1 range, making it easier for the model to process.

[0238] In one possible implementation, the preprocessing module can use a sliding window technique to aggregate electricity consumption data every 5 minutes into a single data point, generating a standardized behavior dataset and reducing noise interference. Using this standardized behavior dataset, online learning techniques are then employed to extract features from the dynamic behavior analysis model, resulting in a behavior feature set.

[0239] Understandably, online learning technologies such as incremental support vector machines are well-suited for processing real-time data streams.

[0240] For example, feature extraction can identify regular patterns in users' peak evening electricity consumption, such as air conditioner operation time concentrated between 19:00 and 22:00, with power fluctuations ranging from 2 to 4 kilowatts. These features constitute a behavioral feature set, reflecting users' electricity consumption habits. If the behavioral feature set does not match a preset risk threshold, potential safety hazards are classified using a risk identification algorithm to obtain hazard classification results.

[0241] Preferably, the risk threshold can be set to a single device power exceeding 5 kilowatts or a short-term power fluctuation exceeding 50%.

[0242] For example, if a user's behavior shows that the water heater's power suddenly increases to 6 kilowatts, the risk identification algorithm will classify it as a "potential overload hazard".

[0243] In one embodiment, the algorithm combines historical data to determine whether an operation is abnormal, classifying the risks into three levels: low, medium, and high. Based on the risk classification results, a parameter adjustment module iteratively updates the dynamic behavior analysis model to obtain the updated model parameters.

[0244] Specifically, parameter adjustments can be made by optimizing model weights using gradient descent to make them more adaptable to new data.

[0245] For example, if multiple overload risks are detected, the model parameters will be made more sensitive to power fluctuations during peak hours.

[0246] It should be noted that iterative updates ensure the model remains accurate as user behavior changes. Using the updated model parameters, the configuration generation module adjusts the configuration of the electricity safety hazard prediction system to obtain an optimized system configuration.

[0247] For example, the configuration generation module can adjust the alarm threshold for overload hazards in the prediction system based on the model parameters, reducing it from 5 kW to 4.5 kW to improve the sensitivity of the early warning.

[0248] In one possible implementation, the system configuration also includes adjusting the push frequency, such as sending an alert to high-risk users once per hour. Based on the optimized system configuration, the updated security risk prediction content is pushed to the client terminal via a real-time communication interface, and the push completion status is determined.

[0249] For example, the push notification might read, "An abnormal power output has been detected in the water heater; it is recommended to check the device status." The communication interface ensures that messages are delivered to the smart meter or the user's mobile app in real time.

[0250] Preferably, after the push is completed, the system records the status as "delivered" to facilitate subsequent tracking.

[0251] Example 2 is the second embodiment of the present invention, which differs from the previous two embodiments in that:

[0252] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0253] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0254] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0255] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0256] Example 3, the third embodiment of the present invention, provides a system for predicting potential electrical safety hazards based on customer characteristic information, including: a data acquisition and preprocessing module, a profile modeling and behavior pattern construction module, a dynamic behavior analysis module, an anomaly detection and risk identification module, a guidance generation and adaptation judgment module, and a push execution and system optimization module.

[0257] The data acquisition and preprocessing module acquires customer characteristic information and real-time electricity consumption data, and performs preprocessing to obtain a multi-dimensional data set.

[0258] The profile modeling and behavior pattern construction module aggregates and models the data set to generate customer electricity consumption behavior profiles and constructs electricity consumption behavior patterns.

[0259] The dynamic behavior analysis module performs dynamic behavior analysis on real-time data based on profiles and patterns to identify the changing trends of current electricity consumption behavior.

[0260] The anomaly detection and risk identification module compares the changing trend with the preset behavior pattern to determine whether there is an anomaly. If there is, it identifies the risk level and parameters.

[0261] The guidance generation and adaptation judgment module generates a guidance scheme to prevent misoperation, performs an adaptation judgment, and generates the final guidance content.

[0262] The push execution and system optimization module pushes the final guidance content to the client terminal, records feedback information, updates the behavior profile and guidance strategy, and optimizes the dynamic behavior analysis and risk identification results.

[0263] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting potential electrical safety hazards based on customer characteristic information, characterized in that: include, Acquire customer characteristic information and real-time electricity consumption data, and preprocess them to obtain a multi-dimensional data set; Aggregate and model the dataset to generate customer electricity consumption behavior profiles and construct electricity consumption behavior patterns; Based on electricity consumption behavior profiles and patterns, dynamic behavior analysis is performed on real-time electricity consumption data to identify current trends in electricity consumption behavior. The changing trends are compared with preset behavioral patterns to determine whether there are any potential anomalies; If an anomaly is found, risk identification and processing will be performed to determine the risk level and corresponding response parameters, and a guidance plan for preventing misoperation will be generated. The adaptability of the anti-misoperation guidance plan is assessed, and the final guidance content is generated based on the assessment results; The final guidance content is pushed to the client's terminal, feedback information is recorded, behavioral profiles and guidance strategies are updated, and the results of dynamic behavior analysis and risk identification are optimized.

2. The method for predicting potential electrical safety hazards based on customer characteristic information as described in claim 1, characterized in that: The process of aggregating and modeling the data set to generate customer electricity consumption behavior profiles and constructing electricity consumption behavior patterns includes, Based on the multidimensional dataset, the structure of customer characteristic information and real-time electricity consumption data is unified and feature extraction is performed. Feature information reflecting differences in user behavior is extracted and sample segmentation is performed to distinguish customer categories with different electricity consumption behavior characteristics. For each customer category, generate corresponding feature descriptions to form a distinctive behavioral profile.

3. The method for predicting potential electrical safety hazards based on customer characteristic information as described in claim 2, characterized in that: The process of performing dynamic behavior analysis on real-time electricity consumption data to identify trends in current electricity consumption behavior includes... Real-time electricity consumption data is constructed into a structured behavioral data sequence in chronological order; The behavioral data sequence is processed continuously to construct a data structure that reflects the evolution of behavior, and compared with a preset behavioral pattern to identify behavioral segments that deviate from the pattern. Comparative data representing the changes in deviant behaviors are extracted to form behavioral trend analysis results.

4. The method for predicting potential electrical safety hazards based on customer characteristic information as described in claim 3, characterized in that: The generated anti-misoperation guidance scheme includes: Based on the comparison between the results of behavioral trend analysis and the preset behavioral patterns, the risk level and response parameters are determined, and the strategy combination elements that meet the conditions are selected from the preset set of strategy rules. The strategy combination elements are structurally organized to construct the strategy expression content that responds to the current risk scenario, and matched with the electricity consumption behavior profile to generate an initial guidance plan to prevent misoperation.

5. The method for predicting potential electrical safety hazards based on customer characteristic information as described in claim 4, characterized in that: The process of assessing the adaptability of the anti-misoperation guidance scheme and generating final guidance content based on the assessment results includes: Acquire real-time electricity consumption data for the current time period, and perform format conversion and consistency processing to form a standardized data set; When the standardized dataset meets the suitability assessment requirements, a suitability assessment operation is performed based on the correspondence between the standardized dataset and the initial guidance scheme. Based on the suitability assessment results, the content was adjusted to obtain updated guidance content. The updated guidance will be output as the final guideline for preventing misoperation.

6. The method for predicting potential electrical safety hazards based on customer characteristic information as described in claim 5, characterized in that: The process includes pushing the final guidance content to the client's terminal, recording feedback information, updating the behavioral profile and guidance strategy, and optimizing the dynamic behavior analysis and risk identification results. Obtain the final guidance content generated at the current stage; Based on the communication identifier of the corresponding client terminal, the guidance content is sent through a preset push channel; After the push notification is sent, record the execution status information of the push action; The feedback content returned by the customer terminal and subsequent electricity consumption behavior data are extracted in a structured manner to form a feedback data set; Based on the feedback data set, update the customer's corresponding electricity consumption behavior profile and the parameters of the anti-misoperation guidance strategy; The updated electricity consumption behavior profile and strategy parameters are integrated and processed, and the processing parameters of the dynamic behavior analysis process and risk identification algorithm are adjusted.

7. The method for predicting potential electrical safety hazards based on customer characteristic information as described in claim 6, characterized in that: The process of pushing the final guidance content to the client's terminal, recording feedback information, updating the behavioral profile and guidance strategy, and optimizing the dynamic behavior analysis and risk identification results also includes... Receive feedback information from customer terminals and subsequent electricity consumption behavior data; The subsequent electricity consumption behavior data is cleaned and restructured, user behavior characteristics are extracted, customer electricity consumption behavior profiles and guidance strategy parameters are updated, and it is determined whether there is a deviation between the updated profile and the current strategy configuration. If any deviation is found, the strategy content is revised, updated anti-misoperation guidance content is generated, and sent to the client terminal; By combining the latest electricity consumption data from customer terminals, the dynamic behavior analysis model and risk identification parameters are updated.

8. A system for predicting potential electrical safety hazards based on customer characteristic information, employing the method for predicting potential electrical safety hazards based on customer characteristic information as described in any one of claims 1 to 7, characterized in that, It includes: a data acquisition and preprocessing module, a profile modeling and behavior pattern construction module, a dynamic behavior analysis module, an anomaly detection and risk identification module, a guidance generation and adaptation judgment module, and a push execution and system optimization module. The data acquisition and preprocessing module acquires customer characteristic information and real-time electricity consumption data, and performs preprocessing to obtain a multi-dimensional data set. The profile modeling and behavior pattern construction module aggregates and models the data set to generate customer electricity consumption behavior profiles and constructs electricity consumption behavior patterns. The dynamic behavior analysis module performs dynamic behavior analysis on real-time data based on profiles and patterns to identify the changing trends of current electricity consumption behavior. The anomaly detection and risk identification module compares the changing trend with the preset behavior pattern to determine whether there is an anomaly. If there is, it identifies the risk level and parameters. The guidance generation and adaptation judgment module generates a guidance scheme to prevent misoperation, performs an adaptation judgment, and generates the final guidance content. The push execution and system optimization module pushes the final guidance content to the client terminal, records feedback information, updates the behavior profile and guidance strategy, and optimizes the dynamic behavior analysis and risk identification results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of any one of claims 1 to 7 of the method for predicting potential electrical safety hazards based on customer characteristic information.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of any one of claims 1 to 7 of the method for predicting potential electrical safety hazards based on customer characteristic information.

Images