Community energy consumption monitoring system based on big data

Abstract

The application discloses a community energy consumption monitoring system based on big data, which comprises a data acquisition module, a user type data processing module, an acquisition type monitoring sub-model module and an energy consumption intelligent monitoring module. The application relates to the technical field of community energy data processing, and particularly relates to a community energy consumption monitoring system based on big data. The application innovatively introduces a community user type difference analysis mechanism, independently learns and differentiates the energy consumption behavior characteristics of different user groups, and improves the energy consumption monitoring precision. The application proposes an energy consumption feature selection algorithm based on a fuzzy lower approximate matrix and an adaptive feature weight learning mechanism, improves the feature selection precision and stability, and enhances the performance of the energy consumption monitoring sub-model. The application adopts a three-stage progressive updating strategy and a multi-vector mean variation strategy to improve the optimization algorithm, improves the optimization efficiency and precision of the energy consumption monitoring sub-model, and realizes intelligent monitoring of community energy consumption.

Description

Technical Field

[0001] This invention relates to the field of community energy data processing technology, specifically to a community energy consumption monitoring system based on big data. Background Technology

[0002] The community energy consumption monitoring system based on big data is a comprehensive energy management system that uses big data technology to collect and analyze various energy consumption data within a community in real time. Through centralized processing and intelligent analysis of energy consumption data, it enables dynamic monitoring of energy consumption within the community, thereby improving energy efficiency, reducing energy costs, and supporting energy conservation management and decision-making in smart communities.

[0003] However, traditional community energy consumption monitoring systems suffer from several technical problems. First, they monitor the overall energy consumption of the community without distinguishing between different types of users, leading to inaccurate monitoring results. Second, existing algorithms for energy consumption feature selection cannot differentiate the internal distribution differences of different result categories or adjust feature weights autonomously based on data characteristics. This results in poor adaptability and weak fitting ability for complex data, leading to inaccurate energy consumption feature extraction, poor performance of the energy consumption monitoring model, and low accuracy of output results. Third, existing sub-models for energy consumption monitoring often have inadequate hyperparameter settings, and individuals are prone to getting trapped in local optima during hyperparameter optimization, resulting in low hyperparameter search accuracy and consequently, low accuracy of the monitoring sub-model's output results. Summary of the Invention

[0004] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides a community energy consumption monitoring system based on big data. Addressing the technical problem of inaccurate monitoring results in traditional community energy consumption monitoring systems that monitor overall community energy consumption without distinguishing between different types of users' energy usage differences, this solution innovatively introduces a community user-type differential analysis mechanism. First, a user-type data processing module is used to classify user type data and select features. Then, energy consumption monitoring sub-models for each user type are constructed and trained. This allows for independent learning and differential identification of energy consumption behavior characteristics of different user groups, improving the accuracy of energy consumption monitoring. It not only supports energy consumption monitoring for multiple energy types and communities of varying sizes, effectively enhancing the system's universality and scalability, but also achieves classified monitoring of community energy consumption, significantly improving the refinement and intelligence of smart community energy management. Furthermore, existing algorithms for energy consumption feature selection suffer from limitations such as inability to distinguish the internal distribution differences of different result categories and inability to autonomously adjust feature weights based on data characteristics. This results in poor adaptability and weak fitting ability for complex data, leading to inaccurate energy consumption feature extraction results, poor performance of energy consumption monitoring models, and low accuracy of output results. To address the technical challenges of low energy consumption, this solution innovatively proposes an energy consumption feature selection algorithm based on constructing a fuzzy approximation matrix using variance-based methods and introducing an adaptive feature weight learning mechanism. This algorithm accurately captures the internal density differences of different energy consumption categories and automatically adjusts weights according to data distribution characteristics, improving the discriminative ability of features in high-density and sparse categories, enhancing feature selection accuracy and stability, removing redundant and low-contribution features, reducing the model's input dimensionality and computational complexity, and simultaneously improving the performance and accuracy of the energy consumption monitoring model's output results. This enables intelligent management of community energy consumption. Furthermore, it addresses existing applicable energy consumption monitoring sub-models... The existing model suffers from insufficient hyperparameter settings and is prone to getting trapped in local optima during hyperparameter optimization, resulting in low hyperparameter search accuracy and consequently, low accuracy of the monitoring sub-model output. This solution innovatively employs a three-stage progressive update strategy and a multi-vector mean mutation strategy in the particle swarm optimization algorithm. This broadens the hyperparameter search range and prevents premature convergence during optimization, ensuring a balance between global search capability and local exploration capability. This improves the precision and accuracy of hyperparameter optimization, enhances the efficiency and accuracy of the energy consumption monitoring sub-model optimization, and ultimately achieves the classified and intelligent monitoring of community energy consumption.

[0005] The technical solution adopted by the present invention is as follows: The community energy consumption monitoring system based on big data provided by the present invention includes a data acquisition module, a user-typed data processing module, a module for acquiring monitoring sub-models of various types, and an intelligent energy consumption monitoring module;

[0006] The data acquisition module specifically obtains raw data on community energy consumption through data collection.

[0007] The user-typed data processing module specifically optimizes and processes raw data and classifies community user data. It then constructs an approximate matrix under fuzzy conditions based on class variance and introduces an adaptive feature weight learning mechanism to build an energy consumption feature selection algorithm. Finally, it obtains the optimal feature set for each type of energy consumption and its corresponding weights.

[0008] The module for obtaining monitoring sub-models of various types is used to construct independent energy consumption monitoring sub-models for different user types. Specifically, it first constructs a unified energy consumption monitoring model, then constructs and trains corresponding energy consumption monitoring sub-models for different user types based on the unified energy consumption monitoring model, then constructs a sub-model hyperparameter optimization algorithm through an improved optimization algorithm to obtain the optimal hyperparameter combination of each energy consumption monitoring sub-model, and finally updates the hyperparameter configuration of each sub-model according to each optimal hyperparameter combination to obtain the optimal energy consumption monitoring sub-model for each type of user.

[0009] The intelligent energy consumption monitoring module specifically inputs real-time energy consumption data of various types of users into the corresponding optimal energy consumption monitoring sub-model to obtain real-time energy monitoring results for each type of user, thereby realizing classified monitoring of community energy consumption.

[0010] Furthermore, the data acquisition module specifically obtains raw community energy consumption data by collecting multi-source data from the community energy management platform; the raw community energy consumption data includes historical energy consumption data and real-time energy consumption data; both the historical energy consumption data and the real-time energy consumption data include community user data, energy consumption record data of each user in the community, and community environmental data; the historical energy consumption data also includes historical energy monitoring results of each user.

[0011] Furthermore, the user-typed data processing module specifically includes the following steps:

[0012] The raw data optimization process specifically involves data cleaning and standardization to obtain optimized community energy consumption data.

[0013] Community user data is categorized, specifically by dividing community energy consumption optimization data into energy consumption data for each user type based on the user type within the community user data.

[0014] The process of constructing an energy consumption feature selection algorithm includes the following steps:

[0015] The initial algorithm parameters are specifically used to construct the feature selection decision table. and initialize the number of iterations. Set the maximum number of iterations. Initialize feature weight association parameters Where U represents the sample sets of user data of various types, This represents the energy consumption characteristic set of various user types. Indicates the decision variables based on energy monitoring results;

[0016] The class variance of a feature is calculated, specifically by calculating the class variance of each energy consumption feature within the sample subset of each type of energy consumption monitoring results; the formula used is as follows:

[0017] ;

[0018] In the formula, This represents the class variance of the k-th energy consumption feature in the sample subset of the t-th class of energy consumption monitoring results. This represents a subset of the energy consumption monitoring results for the t-th category. This represents the number of samples in the subset of energy consumption monitoring results for category t. Let represent the mean of all samples in the subset of energy consumption monitoring results for the k-th energy consumption characteristic in the t-th category. This represents the value of the k-th energy consumption feature in the i-th sample. This represents the k-th energy consumption characteristic. This represents the i-th sample;

[0019] To construct the fuzzy approximation matrix for each energy consumption feature, for each energy consumption feature, iterate through any two samples in the sample set, calculate the fuzzy similarity relationship between the two samples in terms of energy consumption features based on the class variance of the features, and then calculate the fuzzy approximation value of each energy consumption feature based on the fuzzy similarity relationship, finally obtaining the fuzzy approximation matrix for each energy consumption feature; the formula used is as follows:

[0020] ;

[0021] ;

[0022] In the formula, Indicates sample and In terms of energy consumption characteristics Fuzzy similarity relationships, Let j represent the j-th sample. This indicates that the k-th energy consumption characteristic is in The value of , Indicates sample In terms of energy consumption characteristics Below The value, Indicates sample right membership degree This represents the function that takes the minimum value. This represents the function that takes the maximum value.

[0023] Energy consumption characteristic weight optimization includes the following steps:

[0024] The objective function for weight optimization is designed by constructing an objective function that integrates fuzzy approximation error, decision error, and regularization term; the formula used is as follows:

[0025] ;

[0026] In the formula, This represents the value of the objective function for weight optimization. This represents the energy consumption monitoring result indication matrix. Indicates energy consumption characteristics The fuzzy approximation matrix is ​​given by the element in the i-th row and t-th column. , Indicates the Tth iteration Weight, Represents the decision matrix based on energy consumption characteristics. Represents an n-dimensional vector whose elements are all 1s. Denotes the Frobenius norm. This represents the weight of the energy consumption feature in the T-th iteration. and The regularization parameters are used to constrain decision error and weight sparsity, respectively.

[0027] The energy consumption feature weights are calculated by updating the feature weight association parameters using gradient descent, thereby continuously decreasing the objective function value during the iteration process. Based on the updated feature weight association parameters, the updated energy feature weights are then calculated. The formula used is as follows:

[0028] ;

[0029] ;

[0030] In the formula, Indicates the time of the Tth iteration The weighted correlation parameters, Indicates the first During the next iteration The weighted correlation parameters, Represents the learning rate, with a range of values. , This represents the weight optimization objective function value pair. The partial derivatives, No. iteration Weight, Indicates the first During the next iteration The weighted correlation parameters, This represents the l-th energy consumption characteristic;

[0031] The weight iteration optimization terminates when the iteration termination condition is met during the iteration process. Specifically, the iterative update of the energy feature weights is stopped, and the energy consumption feature weights of the current iteration stage are output as the optimal energy consumption feature weights.

[0032] The optimal feature subset is selected by sorting the weights of the optimal energy consumption features in descending order and calculating the cumulative weight percentage of the top s energy consumption features. If the cumulative weight percentage is greater than a weight percentage threshold, then the top s energy consumption features are selected to form the optimal energy consumption feature subset. The formula used is as follows:

[0033] ;

[0034] In the formula, Indicates the cumulative weight percentage. express The optimal weights, where m represents the number of energy consumption features;

[0035] To obtain the optimal feature set for each type, the user data of each type is input into the energy consumption feature selection algorithm to obtain the optimal feature dataset and corresponding weights for each type of energy consumption.

[0036] Furthermore, the acquisition of various types of monitoring sub-model modules specifically includes the following steps:

[0037] The energy consumption monitoring model is constructed by building a dynamic time-series feature extraction layer for energy consumption and an output layer for energy consumption monitoring results.

[0038] The construction of the energy consumption dynamic time series feature extraction layer specifically involves weighting the input energy consumption features with their corresponding weights to form a weighted energy consumption feature sequence, and then inputting it into a bidirectional long short-term memory network to obtain the energy consumption dynamic time series features.

[0039] The construction of the energy consumption monitoring result output layer specifically involves inputting the dynamic time-series features of energy consumption into a fully connected neural network and a Softmax classifier, outputting the probability distribution corresponding to each energy consumption state level, and determining the energy monitoring result based on the category corresponding to the highest probability.

[0040] The construction and training of energy consumption monitoring sub-models are specifically based on the energy consumption monitoring model. Energy consumption monitoring sub-models are constructed for different user types, and the energy consumption monitoring training data corresponding to each type of user is input into the corresponding energy consumption monitoring sub-model. The models are trained separately to obtain the trained energy consumption monitoring sub-models for each type of user.

[0041] The sub-model hyperparameter optimization algorithm is constructed by the following steps:

[0042] The initial population is generated by encoding the hyperparameters of the energy consumption monitoring sub-model into individual position vectors, and randomly initializing and generating the position vectors of N individuals according to the upper and lower boundaries of the hyperparameters to obtain the initial population.

[0043] Calculate the individual fitness value, specifically by calculating the fitness value of individuals in the population; use the performance of the energy consumption monitoring sub-model established based on the individual's location as the individual fitness value;

[0044] The initial update position is determined by two strategies based on the contraction behavior strategy and random conditions; the formula used is as follows:

[0045] ;

[0046] In the formula, Indicates that the i-th individual is in the i-th position. Initial update position in the next iteration This represents the optimal position of an individual globally. Parameters representing individual information interaction. Indicates the first The position of a random individual in the next iteration. Indicates the first The optimal individual position in the next iteration. express Random numbers within a range This indicates the position update control parameter for the i-th individual. Indicates that the i-th individual is in the i-th position. Position in the next iteration Indicates the number of iterations. Indicates the maximum number of search iterations;

[0047] The dynamic diversity adjustment position is specifically achieved by dynamically adjusting the initial update position through a probability-triggered mechanism; the formula used is as follows:

[0048] ;

[0049] In the formula, Indicates that the i-th individual is in the i-th position. The position is dynamically updated in each iteration. , and express Random numbers within a range Pc represents the adjustment magnitude constant, and Pc represents the population change control parameter. Indicates the first The optimal individual position in the initial update of the next iteration;

[0050] The local breakthrough update position is specifically achieved by optimizing the position locally through a multi-vector mean mutation strategy; the formula used is as follows:

[0051] ;

[0052] In the formula, Indicates that the i-th individual is in the i-th position. Position in the next iteration Indicates two firsts The mean of the random individual positions in each iteration. Indicates a first The average position of the random individual and the globally optimal individual in each iteration. Indicates the mutation factor in the early stage of iteration. Indicates the mutation factor in the later stage of iteration;

[0053] The iterative search terminates when, specifically, the fitness values ​​of all updated individuals are recalculated, and the fitness values ​​of all individuals are compared to update the global optimal position of the individual. When any search termination condition is met, the search is terminated and the global individual optimal position is obtained. The global individual optimal position specifically refers to the optimal hyperparameter combination of the energy consumption monitoring sub-model.

[0054] To obtain the optimal hyperparameter combination, the sub-model hyperparameter optimization algorithm is used to search for the hyperparameters of each energy consumption monitoring sub-model and obtain the optimal hyperparameter combination for each energy consumption monitoring sub-model.

[0055] The optimal hyperparameter update of the sub-model is specifically to update the hyperparameter configuration of each energy consumption monitoring sub-model according to the optimal hyperparameter combination of each energy consumption monitoring sub-model, so as to obtain the optimal energy consumption monitoring sub-model for each type of user.

[0056] Furthermore, the intelligent energy consumption monitoring module specifically inputs the real-time energy consumption monitoring data corresponding to each type of user into the corresponding optimal energy consumption monitoring sub-model for each type of user to obtain the real-time energy monitoring results for each type of user, and realizes intelligent monitoring of the community's energy consumption status based on these results.

[0057] The beneficial effects achieved by adopting the above solution are as follows:

[0058] (1) In view of the technical problem that the traditional community energy consumption monitoring system monitors the overall energy consumption of the community without distinguishing the differences in energy use of different types of users, resulting in inaccurate energy consumption monitoring results, this solution innovatively introduces a community user type difference analysis mechanism. First, the user type data is divided and feature is selected through the user type data processing module, and energy consumption monitoring sub-models for each type of user are constructed and trained respectively. This enables independent learning and differential identification of the energy consumption behavior characteristics of different user groups, which improves the accuracy of energy consumption monitoring. It not only supports energy consumption monitoring of multiple energy types and community scenarios of different sizes, but also effectively enhances the universality and scalability of the system, realizes the classified monitoring of community energy consumption, and significantly improves the refinement and intelligence level of smart community energy management.

[0059] (2) In view of the technical problems of existing algorithms for energy consumption feature selection, such as the inability to distinguish the internal distribution differences of different result categories and the inability to adjust feature weights autonomously according to data characteristics, resulting in poor adaptability and weak fitting ability to complex data, which in turn leads to inaccurate energy consumption feature extraction results, poor performance of energy consumption monitoring models and low accuracy of output results, this solution innovatively proposes an energy consumption feature selection algorithm based on constructing an approximate matrix under fuzzy conditions and introducing an adaptive feature weight learning mechanism. It can accurately capture the internal density differences of different energy consumption categories and automatically adjust the weights according to the data distribution characteristics, improve the discrimination ability of features under high density and sparse categories, improve the accuracy and stability of feature selection, remove redundant and low contribution features, reduce the input dimension and computational complexity of the model, and enhance the performance of energy consumption monitoring models and the accuracy of output results, thereby realizing intelligent management of community energy consumption.

[0060] (3) In view of the technical problems that the existing applicable energy consumption monitoring sub-models are not reasonably set, and individuals are prone to getting trapped in local optima during the hyperparameter optimization process, resulting in low hyperparameter search accuracy and thus low accuracy of the monitoring sub-model output results, this solution innovatively adopts a three-stage progressive update strategy and a multi-vector mean mutation strategy in the particle swarm optimization algorithm to broaden the hyperparameter search range and avoid premature convergence in the optimization process. This ensures the balance between global search capability and local exploration capability during the optimization process, improves the precision and accuracy of hyperparameter optimization, and improves the optimization efficiency and accuracy of the energy consumption monitoring sub-model. Finally, it realizes the classification and intelligent monitoring of community energy consumption. Attached Figure Description

[0061] Figure 1 A schematic diagram of the modules of the community energy consumption monitoring system based on big data provided by the present invention;

[0062] Figure 2 A flowchart illustrating the user-typed data processing module;

[0063] Figure 3 A flowchart illustrating the process of obtaining various types of monitoring sub-model modules;

[0064] Figure 4 A flowchart illustrating the process of building an energy consumption feature selection algorithm in the user-typed data processing module;

[0065] Figure 5 A flowchart illustrating the process of constructing hyperparameter optimization algorithms for sub-models in various types of monitoring sub-model modules;

[0066] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0067] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0068] Example 1, see Figure 1 The community energy consumption monitoring system based on big data provided by the present invention includes a data acquisition module, a user-typed data processing module, a module for acquiring various types of monitoring sub-models, and an intelligent energy consumption monitoring module.

[0069] The data acquisition module specifically obtains raw data on community energy consumption through data collection and sends the data to the user-typed data processing module.

[0070] The user-typed data processing module receives data sent by the data acquisition module. Specifically, it optimizes and processes the raw data and classifies community user data. Based on class variance, it constructs an approximate matrix under fuzzy conditions and introduces an adaptive feature weight learning mechanism to build an energy consumption feature selection algorithm. Then, it obtains the optimal feature set for each type, obtains the optimal feature dataset for each type of energy consumption and the corresponding weights, and sends the data to the acquisition module for each type of monitoring sub-model and the intelligent energy consumption monitoring module.

[0071] The module for acquiring monitoring sub-models of various types receives data sent by the user-typed data processing module and is used to construct independent energy consumption monitoring sub-models for different user types. Specifically, a unified energy consumption monitoring model is first constructed. Based on the unified energy consumption monitoring model, corresponding energy consumption monitoring sub-models are constructed and trained for different user types. Then, an improved optimization algorithm is used to construct a sub-model hyperparameter optimization algorithm to obtain the optimal hyperparameter combination for each energy consumption monitoring sub-model. Finally, the hyperparameter configuration of each sub-model is updated according to each optimal hyperparameter combination to obtain the optimal energy consumption monitoring sub-model for each user type, and the data is sent to the intelligent energy consumption monitoring module.

[0072] The intelligent energy consumption monitoring module receives data from the user-type data processing module and the module that acquires monitoring sub-models of various types. It inputs the real-time energy consumption data of each type of user into the corresponding optimal energy consumption monitoring sub-model to obtain the real-time energy monitoring results of each type of user, thereby realizing the classified monitoring of community energy consumption.

[0073] By performing the above operations, this solution addresses the technical problem in traditional community energy consumption monitoring systems that fail to differentiate between different types of users when monitoring overall community energy consumption, leading to inaccurate monitoring results. It innovatively introduces a community user-type differentiation analysis mechanism. First, a user-type data processing module categorizes user data and selects features. Then, it constructs and trains energy consumption monitoring sub-models for each user type. This allows for independent learning and differentiated identification of energy consumption behavior characteristics among different user groups, improving the accuracy of energy consumption monitoring. The solution supports energy consumption monitoring for multiple energy types and communities of varying sizes, effectively enhancing the system's universality and scalability. It achieves categorized monitoring of community energy consumption, significantly improving the refinement and intelligence of smart community energy management.

[0074] Example 2, see Figure 1 This embodiment is based on the above embodiment. Specifically, the data acquisition module obtains raw community energy consumption data by collecting multi-source data from the community energy management platform. The raw community energy consumption data includes historical energy consumption data and real-time energy consumption data. Both the historical energy consumption data and the real-time energy consumption data include community user data, energy consumption record data of each user in the community, and community environmental data.

[0075] The community environmental data includes temperature, humidity, wind speed, weather events, ambient light intensity, season, month, holidays, and day / night time markers;

[0076] The community user data includes user type, number of users, area occupied, and population density;

[0077] The user types include residents, merchants, and public facilities;

[0078] The energy consumption records of each user in the community include electricity consumption records, water consumption records, gas consumption records, and heat consumption records.

[0079] The historical energy consumption data also includes historical energy monitoring results for each user.

[0080] The energy monitoring results include normal level, slight abnormality level, general abnormality level, and severe abnormality level.

[0081] Example 3, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment, and the user-typed data processing module specifically includes the following steps:

[0082] The raw data optimization process specifically involves data cleaning and standardization to obtain optimized community energy consumption data.

[0083] The community energy consumption optimization data includes historical energy consumption optimization data and real-time energy consumption optimization data;

[0084] The data cleaning is used to eliminate errors, missing values ​​and inconsistencies in the data, specifically by filling in missing values ​​and removing outliers from the original data.

[0085] The missing value imputation specifically involves filling in the missing values ​​using the mean imputation method;

[0086] The outlier removal specifically involves using the IQR method to detect and remove extreme values ​​and logical outliers in the original data;

[0087] The data standardization is used to eliminate differences in time scale, data format, unit of measurement and numerical range among multi-source collected data, specifically including time alignment, format standardization, unit standardization and data normalization;

[0088] The time alignment specifically involves mapping raw data with different sampling frequencies, different upload delays, and timestamp discrepancies to a unified time axis;

[0089] The format standardization specifically involves unifying the original data into standard fields and data types;

[0090] The unit standardization specifically refers to unifying the standard units of the original data;

[0091] The data normalization specifically involves using Min-Max normalization to normalize all continuous variables, ensuring that the data is within a uniform range, and using one-hot encoding to uniformly convert categorical fields into numerical forms that can be recognized by the model.

[0092] Community user data classification is used to differentiate optimized community energy consumption data according to user type; specifically, based on the user type in the community user data, the optimized community energy consumption data is divided into energy consumption data for each type of user.

[0093] The energy consumption data for each type of user includes energy consumption data for residents, energy consumption data for merchants, and energy consumption data for public facilities.

[0094] The household energy consumption data includes historical household energy consumption data and household energy consumption data;

[0095] The merchant energy consumption data includes historical merchant energy consumption data and general merchant energy consumption data;

[0096] The energy consumption data for public facilities includes historical energy consumption data for public facilities and general energy consumption data for public facilities.

[0097] The process of constructing an energy consumption feature selection algorithm includes the following steps:

[0098] The initial algorithm parameters are specifically used to construct the feature selection decision table. and initialize the number of iterations. Set the maximum number of iterations. Initialize feature weight association parameters Where U represents the sample sets of user data of various types, This represents the energy consumption characteristic set of various user types. The decision variables representing energy monitoring results are shown in the following formula:

[0099] ;

[0100] ;

[0101] ;

[0102] ;

[0103] ;

[0104] In the formula, This represents the first sample. This represents the second sample. This represents the nth sample. This indicates the first energy consumption characteristic. This indicates the second energy consumption characteristic. This represents the m-th energy consumption characteristic. Indicates a normal level. Indicates a minor abnormality level. Indicates the general level of abnormality. Indicates a severe abnormality level. This represents the subset of samples corresponding to the normal level. This represents the subset of samples corresponding to a minor anomaly level. This represents a subset of samples corresponding to a general anomaly level. This represents the subset of samples corresponding to the severity of the anomaly. This represents the value of the first energy consumption feature weight association parameter. This represents the value of the weighted correlation parameter for the second energy consumption feature. This represents the weight association parameter value of the m-th energy consumption feature, where m represents the number of energy consumption features.

[0105] The class variance of the calculated features is used to capture the distributional differences of energy consumption features under different energy consumption anomaly levels, solving the problem of traditional methods ignoring class density differences when using overall variance, and adapting to the heterogeneity of community energy data. Specifically, it calculates the class variance of each energy consumption feature in the sample subset of each type of energy consumption monitoring results; the formula used is as follows:

[0106] ;

[0107] In the formula, This represents the class variance of the k-th energy consumption feature in the sample subset of the t-th class of energy consumption monitoring results. This represents a subset of the energy consumption monitoring results for the t-th category. This represents the number of samples in the subset of energy consumption monitoring results for category t. Let represent the mean of all samples in the subset of energy consumption monitoring results for the k-th energy consumption characteristic in the t-th category. This represents the value of the k-th energy consumption feature in the i-th sample. This represents the k-th energy consumption characteristic. This represents the i-th sample;

[0108] A feature fuzzy approximation matrix is ​​constructed to quantify the stability of energy consumption features in characterizing the energy consumption anomaly level. Specifically, for each energy consumption feature, any two samples in the sample set are traversed, and the fuzzy similarity relationship between the two samples in the energy consumption feature is calculated based on the class variance of the feature. Then, the fuzzy approximation value of each energy consumption feature is calculated based on the fuzzy similarity relationship, and finally the fuzzy approximation matrix of each energy consumption feature is obtained.

[0109] The specific method for constructing the fuzzy approximation matrix is ​​to sort all the fuzzy approximation values ​​of energy consumption features according to the samples. The dimensional arrangement of the anomaly levels forms the fuzzy approximation matrix corresponding to that feature. The number of rows in the matrix is ​​the total number of samples n, and the number of columns is the total number of energy consumption anomaly levels 4. The formula used is as follows:

[0110] ;

[0111] ;

[0112] ;

[0113] In the formula, Indicates sample and In terms of energy consumption characteristics The fuzzy similarity relationship is defined below, with values ​​ranging from 1 to 2. , Let j represent the j-th sample. This indicates that the k-th energy consumption characteristic is in The value of , Indicates sample In terms of energy consumption characteristics Below The value, Indicates sample right membership degree This represents the function that takes the minimum value. This represents the function that takes the maximum value.

[0114] Energy consumption feature weight optimization is used to ensure that feature weights are accurately adapted to the differences in energy consumption characteristics among various users. This process includes the following steps:

[0115] The objective function for weight optimization is designed by constructing an objective function that integrates fuzzy approximation error, decision error, and regularization term; the formula used is as follows:

[0116] ;

[0117] ;

[0118] ;

[0119] In the formula, This represents the objective function value for weight optimization. The smaller the value, the more accurate the current feature weights are in characterizing the energy consumption monitoring results. This represents the energy consumption monitoring result indication matrix, with dimension 1. , This represents the element in the i-th row and t-th column of the energy consumption monitoring result indication matrix, used to label samples. The actual energy consumption monitoring results Indicates energy consumption characteristics The fuzzy approximation matrix is ​​given by the element in the i-th row and t-th column. , Indicates the Tth iteration Weight, This represents the energy consumption characteristic decision matrix, with dimension 1. If energy consumption characteristics Matching samples Based on the energy consumption monitoring results, the element in the i-th row and k-th column of the energy consumption characteristic decision matrix is... Otherwise , Represents an n-dimensional vector whose elements are all 1s. Denotes the Frobenius norm. This represents the weight of the energy consumption feature in the T-th iteration. and This represents the regularization parameter, and its value range is... Each constraint is applied to determine decision error and weight sparsity.

[0120] The energy consumption characteristics Matching samples The conditions for the energy consumption monitoring results are that the energy consumption characteristics are... The values ​​match the sample Monitoring results Typical distribution ,in, This represents the standard deviation of the k-th energy consumption characteristic within the sample subset of the t-th type of energy consumption monitoring results;

[0121] The energy consumption feature weights are calculated for adaptive optimization. Specifically, the feature weight association parameters are updated using gradient descent, causing the objective function value to decrease continuously during the iteration process. Based on the updated feature weight association parameters, the updated energy feature weights are calculated. The formula used is as follows:

[0122] ;

[0123] ;

[0124] ;

[0125] In the formula, Indicates the time of the Tth iteration The weighted correlation parameters, Indicates the first During the next iteration The weighted correlation parameters, Represents the learning rate, with a range of values. , This represents the weight optimization objective function value pair. The partial derivatives, This represents the diagonal matrix of energy consumption feature weights in the T-th iteration. This represents the outer product matrix of the energy consumption feature weights in the T-th iteration. No. iteration Weight, Indicates the first During the next iteration The weighted correlation parameters, This represents the l-th energy consumption characteristic. and Represents the regularization parameter;

[0126] The weight iteration optimization terminates when the iteration termination condition is met during the iteration process. Specifically, the iterative update of the energy feature weights is stopped, and the energy consumption feature weights of the current iteration stage are output as the optimal energy consumption feature weights.

[0127] The iteration termination conditions include reaching the maximum number of iterations and convergence of the weight optimization objective function value;

[0128] The maximum number of iterations is specifically defined as if the current number of iterations exceeds a preset maximum number of iterations. If the iteration terminates, then the iteration ends.

[0129] The convergence of the objective function value for weight optimization is specifically defined as follows: if the difference between the objective function values ​​of two adjacent iterations is less than a preset convergence threshold, then the iteration is terminated.

[0130] The optimal feature subset is selected by sorting the weights of the optimal energy consumption features in descending order and calculating the cumulative weight percentage of the top s energy consumption features. If the cumulative weight percentage is greater than a weight percentage threshold, then the top s energy consumption features are selected to form the optimal energy consumption feature subset. The formula used is as follows:

[0131] ;

[0132] In the formula, Indicates the cumulative weight percentage. express The optimal weight;

[0133] To obtain the optimal feature set for each type, the user data of each type is input into the energy consumption feature selection algorithm to obtain the optimal feature dataset and corresponding weights for each type of energy consumption.

[0134] By performing the above operations, this solution addresses the technical problems of existing algorithms for energy consumption feature selection, which suffer from the inability to distinguish the internal distribution differences of different result categories and the inability to autonomously adjust feature weights based on data characteristics. This results in poor adaptability and weak fitting ability to complex data, leading to inaccurate energy consumption feature extraction results, poor performance of energy consumption monitoring models, and low accuracy of output results. This innovative approach proposes an energy consumption feature selection algorithm based on constructing a fuzzy approximation matrix using class-variance and introducing an adaptive feature weight learning mechanism. This algorithm can accurately capture the internal density differences of different energy consumption categories and automatically adjust weights based on data distribution characteristics. It improves the discriminative ability of features in high-density and sparse categories, enhances feature selection accuracy and stability, removes redundant and low-contribution features, reduces the input dimension and computational complexity of the model, and simultaneously enhances the performance of the energy consumption monitoring model and the accuracy of output results, thus realizing intelligent management of community energy consumption.

[0135] Example 4, see Figure 1 , Figure 3 and Figure 5 This embodiment is based on the above embodiment, and the acquisition of various types of monitoring sub-model modules specifically includes the following steps:

[0136] The energy consumption monitoring model is constructed by building a dynamic time-series feature extraction layer for energy consumption and an output layer for energy consumption monitoring results.

[0137] The energy consumption dynamic time-series feature extraction layer is constructed to extract the dynamic features of energy consumption changing over time. Specifically, the input energy consumption features are weighted with their corresponding weights to form a weighted energy consumption feature sequence, which is then input into a bidirectional long short-term memory network to obtain the energy consumption dynamic time-series features.

[0138] The energy consumption monitoring result output layer is constructed to identify energy consumption status and determine anomaly level. Specifically, the dynamic time-series features of energy consumption are input into a fully connected neural network and a Softmax classifier, and the probability distribution corresponding to each energy consumption status level is output. The energy monitoring result is determined according to the category corresponding to the highest probability.

[0139] The construction and training of energy consumption monitoring sub-models are specifically based on the energy consumption monitoring model. Energy consumption monitoring sub-models are constructed for different user types, and the energy consumption monitoring training data corresponding to each type of user is input into the corresponding energy consumption monitoring sub-model. The models are trained separately to obtain the trained energy consumption monitoring sub-models for each type of user.

[0140] The energy consumption monitoring training data corresponding to each type of user is specifically historical data of energy consumption data of each type of user after being processed by the energy consumption feature selection algorithm.

[0141] The sub-model training specifically involves using the cross-entropy loss function and iteratively adjusting the model weight matrix and bias vector through the backpropagation algorithm until the error converges.

[0142] The sub-model hyperparameter optimization algorithm is constructed by the following steps:

[0143] The initial population is generated by encoding the hyperparameters of the energy consumption monitoring sub-model into individual position vectors, and randomly initializing and generating the position vectors of N individuals according to the upper and lower boundaries of the hyperparameters to obtain the initial population.

[0144] The model hyperparameters include the number of LSTM layers, the dimension of hidden states, the number of neurons in the hidden layers, the time step, and the Dropout ratio, and upper and lower boundaries are set for the value of each hyperparameter.

[0145] The fitness value of an individual is calculated to evaluate the performance of the model under different combinations of hyperparameters. Specifically, the fitness value of an individual in the population is calculated; the performance of the energy consumption monitoring sub-model established based on the individual's location is used as the fitness value of the individual.

[0146] The initial update position is used to provide a preliminary optimization position with global exploratory capabilities, addressing the problem of narrow global search range and susceptibility to local optima in traditional optimization algorithms. Specifically, based on the shrinking behavior strategy, two strategies are used to update the individual position according to random conditions; the formulas used are as follows:

[0147] ;

[0148] ;

[0149] In the formula, Indicates that the i-th individual is in the i-th position. Initial update position in the next iteration This represents the optimal position of an individual globally. Parameters representing individual information interaction, adjusting the intensity of information transmission, and their range of values. , Indicates the first The position of a random individual in the next iteration. Indicates the first The optimal individual position in the next iteration. express Random numbers within a range This indicates the position update control parameter for the i-th individual. Indicates that the i-th individual is in the i-th position. Position in the next iteration Indicates the number of iterations. Indicates the maximum number of search iterations. This represents the fitness value of the i-th individual. This represents the global optimal fitness value for an individual.

[0150] Dynamic diversity adjustment of the update location is used to address potential issues of population clustering and insufficient diversity at the initial update location. Specifically, it involves dynamically adjusting the initial update location through a probability-triggered mechanism; the formula used is as follows:

[0151] ;

[0152] In the formula, Indicates that the i-th individual is in the i-th position. The position is dynamically updated in each iteration. , and express Random numbers within a range The value represents the adjustment amplitude constant, with a value of 0.1. Pc represents the population change control parameter, with a value range of [value missing]. , Indicates the first The optimal individual position in the initial update of the next iteration;

[0153] The local breakthrough update position is used to obtain the final update position that balances global optimum and local refinement. It is the initial position in the next iteration. Specifically, it is optimized locally by using a multi-vector mean mutation strategy. The formula used is as follows:

[0154] ;

[0155] ;

[0156] ;

[0157] In the formula, Indicates that the i-th individual is in the i-th position. Position in the next iteration Indicates two firsts The mean of the random individual positions in each iteration. Indicates a first The average position of the random individual and the globally optimal individual in each iteration. , and Both indicate that a random individual is in the 1st... Position in the next iteration This represents the mutation factor in the early stage of iteration, with a value range of [value missing]. , This represents the mutation factor in the later stage of the iteration, with a value range of [value missing]. ;

[0158] The iterative search terminates when, specifically, the fitness values ​​of all updated individuals are recalculated, and the fitness values ​​of all individuals are compared to update the global optimal position of the individual. When any search termination condition is met, the search is terminated and the global individual optimal position is obtained. The global individual optimal position specifically refers to the optimal hyperparameter combination of the energy consumption monitoring sub-model.

[0159] The search termination conditions include when the individual's optimal position is higher than the fitness threshold and when the maximum number of search iterations is reached.

[0160] To obtain the optimal hyperparameter combination, the sub-model hyperparameter optimization algorithm is used to search for the hyperparameters of each energy consumption monitoring sub-model and obtain the optimal hyperparameter combination for each energy consumption monitoring sub-model.

[0161] The optimal hyperparameter update of the sub-model is specifically to update the hyperparameter configuration of each energy consumption monitoring sub-model according to the optimal hyperparameter combination of each energy consumption monitoring sub-model, so as to obtain the optimal energy consumption monitoring sub-model for each type of user.

[0162] Furthermore, the energy consumption intelligent monitoring module specifically inputs the real-time energy consumption monitoring data corresponding to each type of user into the corresponding optimal energy consumption monitoring sub-model for each type of user to obtain the real-time energy monitoring results for each type of user, and realizes intelligent monitoring of the community's energy consumption status based on these results.

[0163] The real-time energy monitoring results for each type of user include real-time energy monitoring results for residential users, real-time energy monitoring results for commercial users, and real-time energy monitoring results for public facilities.

[0164] Specifically, the real-time energy consumption monitoring data for each type of user refers to the real-time energy consumption data of each type of user after passing through an energy consumption feature selection algorithm.

[0165] By performing the above operations, this solution addresses the technical problems of insufficient hyperparameter settings in existing applicable energy consumption monitoring sub-models, the tendency for individuals to get trapped in local optima during hyperparameter optimization, low hyperparameter search accuracy, and consequently low accuracy of the monitoring sub-model output. This solution innovatively employs a three-stage progressive update strategy and a multi-vector mean mutation strategy in the particle swarm optimization algorithm. This broadens the hyperparameter search range and avoids premature convergence during the optimization process, ensuring a balance between global search capability and local exploration capability. This improves the precision and accuracy of hyperparameter optimization, enhances the optimization efficiency and accuracy of the energy consumption monitoring sub-model, and ultimately achieves the classified and intelligent monitoring of community energy consumption.

[0166] Example 5, see Figure 1This embodiment is based on the above embodiment. Specifically, the energy consumption intelligent monitoring module inputs the real-time energy consumption monitoring data corresponding to each type of user into the corresponding optimal energy consumption monitoring sub-model for each type of user to obtain the real-time energy monitoring results for each type of user, and realizes intelligent monitoring of the community's energy consumption status based on these results.

[0167] The real-time energy monitoring results for each type of user include real-time energy monitoring results for residential users, real-time energy monitoring results for commercial users, and real-time energy monitoring results for public facilities.

[0168] Specifically, the real-time energy consumption monitoring data for each type of user refers to the real-time energy consumption data of each type of user after passing through an energy consumption feature selection algorithm.

[0169] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0170] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A community energy consumption monitoring system based on big data, characterized in that: It includes a data acquisition module, a user-typed data processing module, a module for acquiring various types of monitoring sub-models, and an intelligent energy consumption monitoring module; The data acquisition module specifically obtains raw data on community energy consumption through data collection. The user-typed data processing module specifically optimizes and processes raw data and classifies community user data. It then constructs an approximate matrix under fuzzy conditions based on class variance and introduces an adaptive feature weight learning mechanism to build an energy consumption feature selection algorithm. Finally, it obtains the optimal feature set for each type of energy consumption and its corresponding weights. Specifically, the following steps are included: The raw data optimization process specifically involves data cleaning and standardization to obtain optimized community energy consumption data. Community user data is categorized, specifically by dividing community energy consumption optimization data into energy consumption data for each user type based on the user type within the community user data. An energy consumption feature selection algorithm is constructed, which includes initial algorithm parameters, calculation of feature class variance, construction of approximate matrix under feature fuzziness, optimization of energy consumption feature weights, and selection of the optimal feature subset. To obtain the optimal feature set for each type, the user data of each type is input into the energy consumption feature selection algorithm to obtain the optimal feature dataset and corresponding weights for each type of energy consumption. The module for obtaining monitoring sub-models of various types specifically involves first constructing a unified energy consumption monitoring model, then constructing and training corresponding energy consumption monitoring sub-models for different user types based on the unified energy consumption monitoring model, then constructing a sub-model hyperparameter optimization algorithm through an improved optimization algorithm to obtain the optimal hyperparameter combination for each energy consumption monitoring sub-model, and finally updating the hyperparameter configuration of each sub-model according to each optimal hyperparameter combination to obtain the optimal energy consumption monitoring sub-model for each type of user. The intelligent energy consumption monitoring module specifically inputs real-time energy consumption data of various types of users into the corresponding optimal energy consumption monitoring sub-model to obtain real-time energy monitoring results for each type of user, thereby realizing classified monitoring of community energy consumption.

2. The community energy consumption monitoring system based on big data according to claim 1, characterized in that: The construction of the energy consumption feature selection algorithm specifically includes the following steps: The initial algorithm parameters are specifically used to construct the feature selection decision table. and initialize the number of iterations. Set the maximum number of iterations. Initialize feature weight association parameters Where U represents the sample sets of user data of various types, This represents the energy consumption characteristic set of various user types. Indicates the decision variables based on energy monitoring results; The class variance of a feature is calculated, specifically by calculating the class variance of each energy consumption feature within the sample subset of each type of energy consumption monitoring results; the formula used is as follows: ； In the formula, This represents the class variance of the k-th energy consumption feature in the sample subset of the t-th class of energy consumption monitoring results. This represents a subset of the energy consumption monitoring results for the t-th category. This represents the number of samples in the subset of energy consumption monitoring results for category t. Let represent the mean of all samples in the subset of energy consumption monitoring results for the k-th energy consumption characteristic in the t-th category. This represents the value of the k-th energy consumption feature in the i-th sample. This represents the k-th energy consumption characteristic. This represents the i-th sample; Construct an approximate matrix under feature fuzziness; Optimization of energy consumption characteristic weights; The optimal feature subset is selected by sorting the weights of the optimal energy consumption features in descending order and calculating the cumulative weight percentage of the top s energy consumption features. If the cumulative weight percentage is greater than a weight percentage threshold, then the top s energy consumption features are selected to form the optimal energy consumption feature subset. The formula used is as follows: ； In the formula, Indicates the cumulative weight percentage. express The optimal weight is given by m, which represents the number of energy consumption features.

3. The community energy consumption monitoring system based on big data according to claim 2, characterized in that: The construction of the feature-based fuzzy approximation matrix specifically involves, for each energy consumption feature, iterating through any two samples in the sample set, calculating the fuzzy similarity relationship between the two samples in terms of energy consumption features based on the class variance of the features, and then calculating the fuzzy approximation value of each energy consumption feature based on the fuzzy similarity relationship, ultimately obtaining the fuzzy approximation matrix for each energy consumption feature; the formula used is as follows: ； ； In the formula, Indicates sample and In terms of energy consumption characteristics Fuzzy similarity relationships are shown below. Let j represent the j-th sample. This indicates that the k-th energy consumption characteristic is in The value of , Indicates sample In terms of energy consumption characteristics Below The value, Indicates sample right membership degree This represents the function that takes the minimum value. This represents the function that takes the maximum value.

4. The community energy consumption monitoring system based on big data according to claim 2, characterized in that: The optimization of energy consumption feature weights specifically includes the following steps: The objective function for weight optimization is designed by constructing an objective function that integrates fuzzy approximation error, decision error, and regularization term; the formula used is as follows: ； In the formula, This represents the value of the objective function for weight optimization. This represents the energy consumption monitoring result indication matrix. Indicates energy consumption characteristics The fuzzy approximation matrix is ​​given by the element in the i-th row and t-th column. , Indicates the Tth iteration Weight, Represents the decision matrix based on energy consumption characteristics. Represents an n-dimensional vector whose elements are all 1s. Describing the Frobenius norm, This represents the weight of the energy consumption feature in the T-th iteration. and The regularization parameters are used to constrain decision error and weight sparsity, respectively. The energy consumption feature weights are calculated by updating the feature weight association parameters through gradient descent, so that the value of the weight optimization objective function decreases continuously during the iteration process, and the updated energy feature weights are calculated based on the updated feature weight association parameters. The formula used is as follows: ； ； In the formula, Indicates the time of the Tth iteration The weighted correlation parameters, Indicates the first During the next iteration The weighted correlation parameters, Represents the learning rate, with a range of values. , This represents the weight optimization objective function value pair. The partial derivatives, No. Second iteration Weight, Indicates the first During the next iteration The weighted correlation parameters, This represents the l-th energy consumption characteristic; The weight iteration optimization terminates when the iteration process meets the iteration termination condition. Specifically, if the iteration process meets the iteration termination condition, the iterative update of the energy feature weights is stopped, and the energy consumption feature weights of the current iteration stage are output as the optimal energy consumption feature weights. The iteration termination condition includes reaching the maximum number of iterations and the convergence of the weight optimization objective function value.

5. The community energy consumption monitoring system based on big data according to claim 1, characterized in that: The module for acquiring various types of monitoring sub-models specifically includes the following steps: The energy consumption monitoring model is constructed by building a dynamic time-series feature extraction layer for energy consumption and an output layer for energy consumption monitoring results. The construction of the energy consumption dynamic time series feature extraction layer specifically involves weighting the input energy consumption features with their corresponding weights to form a weighted energy consumption feature sequence, and then inputting it into a bidirectional long short-term memory network to obtain the energy consumption dynamic time series features. The construction of the energy consumption monitoring result output layer specifically involves inputting the dynamic time-series features of energy consumption into a fully connected neural network and a Softmax classifier, outputting the probability distribution corresponding to each energy consumption state level, and determining the energy monitoring result based on the category corresponding to the highest probability. The construction and training of energy consumption monitoring sub-models are specifically based on the energy consumption monitoring model. Energy consumption monitoring sub-models are constructed for different user types, and the energy consumption monitoring training data corresponding to each type of user is input into the corresponding energy consumption monitoring sub-model. The models are trained separately to obtain the trained energy consumption monitoring sub-models for each type of user. Construct a hyperparameter optimization algorithm for the sub-model; To obtain the optimal hyperparameter combination, the sub-model hyperparameter optimization algorithm is used to search for the hyperparameters of each energy consumption monitoring sub-model and obtain the optimal hyperparameter combination for each energy consumption monitoring sub-model. The optimal hyperparameter update of the sub-model is specifically to update the hyperparameter configuration of each energy consumption monitoring sub-model according to the optimal hyperparameter combination of each energy consumption monitoring sub-model, so as to obtain the optimal energy consumption monitoring sub-model for each type of user.

6. The community energy consumption monitoring system based on big data according to claim 5, characterized in that: The algorithm for constructing the sub-model hyperparameter optimization specifically includes the following steps: The initial population is generated by encoding the hyperparameters of the energy consumption monitoring sub-model into individual position vectors, and randomly initializing and generating the position vectors of N individuals according to the upper and lower boundaries of the hyperparameters to obtain the initial population. Calculate the individual fitness value, specifically by calculating the fitness value of individuals in the population; use the performance of the energy consumption monitoring sub-model established based on the individual's location as the individual fitness value; The initial update position is determined by two strategies based on the contraction behavior strategy and random conditions; the formula used is as follows: ； In the formula, It indicates that the i-th individual is in the first position. Initial update position in the next iteration This represents the optimal position of the global individual. Parameters representing individual information interaction. Indicates the first The position of a random individual in the next iteration. Indicates the first The optimal individual position in the next iteration. express Random numbers within a range This indicates the position update control parameter for the i-th individual. It indicates that the i-th individual is in the first position. Position in the next iteration Indicates the number of iterations. Indicates the maximum number of search iterations; The dynamic diversity adjustment position is specifically achieved by dynamically adjusting the initial update position through a probability-triggered mechanism; the formula used is as follows: ； In the formula, It indicates that the i-th individual is in the first position. The position is dynamically updated in each iteration. , and express Random numbers within a range Pc represents the adjustment magnitude constant, and Pc represents the population change control parameter. Indicates the first The optimal individual position in the initial update of the next iteration; The local breakthrough update position is specifically achieved by optimizing the position locally through a multi-vector mean mutation strategy; the formula used is as follows: ； In the formula, It indicates that the i-th individual is in the first position. Position in the next iteration Indicates two firsts The mean of the random individual positions in each iteration. Indicates a first The average position of the random individual and the globally optimal individual in each iteration. Indicates the mutation factor in the early stage of iteration. Indicates the mutation factor in the later stage of iteration; The iterative search terminates when, specifically, the fitness values ​​of all updated individuals are recalculated, and the fitness values ​​of all individuals are compared to update the global optimal position of the individual. When any search termination condition is met, the search is terminated and the global individual optimal position is obtained. The global individual optimal position specifically refers to the optimal hyperparameter combination of the energy consumption monitoring sub-model.

7. The community energy consumption monitoring system based on big data according to claim 1, characterized in that: Specifically, the intelligent energy consumption monitoring module inputs real-time energy consumption monitoring data corresponding to each type of user into the corresponding optimal energy consumption monitoring sub-model for each type of user to obtain real-time energy monitoring results for each type of user, and realizes intelligent monitoring of the community's energy consumption status based on these results.

8. The community energy consumption monitoring system based on big data according to claim 1, characterized in that: The data acquisition module specifically obtains raw community energy consumption data by collecting data from multiple sources from the community energy management platform. The raw community energy consumption data includes historical energy consumption data and real-time energy consumption data. Both the historical energy consumption data and the real-time energy consumption data include community user data, energy consumption records of each user in the community, and community environmental data. The historical energy consumption data also includes historical energy monitoring results of each user.

Images