Intelligent household power consumption management system, method and equipment

Abstract

The invention provides a smart home power consumption management system, method and equipment, the system comprises a data coordination module, a model cooperation module, a task execution module and an early warning cooperation module, the data coordination module is used for preprocessing a collected first data set of the smart home power consumption management system to obtain a second data set; constructing a household electricity utilization knowledge graph according to the second data set; the model cooperation module is used for performing multi-dimensional analysis based on the household electricity knowledge graph to obtain a plurality of analysis results; the task execution module is used for determining a household electricity consumption prediction curve according to the space-time comprehensive characteristics and the anomaly detection result; determining a personalized electricity consumption prediction result according to the prototype user portrait; and the early warning cooperation module is used for generating an intelligent early warning prompt according to the household electricity consumption prediction curve and the personalized electricity consumption prediction result. According to the embodiment of the invention, the prediction accuracy and early warning reliability of household electricity consumption can be improved.

Description

Technical Field

[0001] This application relates to the field of electricity management technology, and in particular to a smart home electricity management system, method and device. Background Technology

[0002] With the rapid popularization of smart homes, the number and types of household electrical appliances are constantly increasing, and the interaction between devices and user behavior patterns are becoming increasingly complex. Among these, the efficiency of household electricity management not only directly affects residents' quality of life and electricity costs, but also relates to the overall energy efficiency level and load regulation capacity of the power system.

[0003] In existing home electricity management technologies, anomaly detection still largely relies on manually configured fixed rules or thresholds to identify electricity anomalies. Furthermore, the electricity analysis dimensions are limited, making it difficult to adapt to dynamic scenarios such as aging equipment, changing habits, and environmental fluctuations, leading to frequent misjudgments and missed detections. Simultaneously, different households exhibit significant individual differences in their smart home device combinations, daily electricity usage habits, and living environments. However, existing systems often employ uniform analysis and prediction models, lacking the ability to accurately capture and quickly adapt to individual user characteristics. This results in a large discrepancy between the output electricity prediction results and the actual electricity demand of households, failing to meet users' personalized electricity management needs.

[0004] Therefore, improving the accuracy of household electricity consumption forecasts and the reliability of early warnings is an urgent issue that needs to be addressed. Summary of the Invention

[0005] This application provides an intelligent home electricity management system, method, and device, which improves the accuracy of prediction and the reliability of early warning for household electricity consumption.

[0006] In a first aspect, embodiments of this application provide a smart home electricity management system, which includes: a data coordination module, a model collaboration module, a task execution module, and an early warning collaboration module, wherein: The data coordination module is used to preprocess the first dataset collected from the smart home electricity management system to obtain a second dataset; and to construct a household electricity knowledge graph based on the second dataset. The model collaboration module is used to perform multi-dimensional analysis based on the household electricity knowledge graph to obtain multiple analysis results; the multiple analysis results include: spatiotemporal integrated features, anomaly detection results, and prototype user profiles. The task execution module is used to determine the household electricity consumption prediction curve based on the spatiotemporal integrated features and the anomaly detection results; and to determine the personalized electricity consumption prediction results based on the prototype user profile. The early warning coordination module is used to generate intelligent early warning prompts based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results.

[0007] Secondly, embodiments of this application provide a smart home electricity management method, the method comprising: The first dataset collected from the smart home electricity management system is preprocessed to obtain the second dataset; Construct a knowledge graph of household electricity consumption based on the second dataset; Based on the aforementioned household electricity knowledge graph, a multi-dimensional analysis was performed to obtain multiple analysis results; these multiple analysis results include: spatiotemporal comprehensive features, anomaly detection results, and prototype user profiles. The household electricity consumption prediction curve is determined based on the spatiotemporal integrated characteristics and the anomaly detection results. Personalized electricity consumption forecasts are determined based on the prototype user profile. Intelligent early warning prompts are generated based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results.

[0008] Thirdly, embodiments of this application provide an electronic device, including: a processor and a memory, the memory being used to store one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, the programs including instructions for performing the steps in the second aspect of this application.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the second aspect of this application.

[0010] Fifthly, embodiments of this application provide a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the second aspect of embodiments of this application. The computer program product may be a software installation package.

[0011] As can be seen, the smart home electricity management system provided in this application firstly preprocesses the first dataset and constructs a household electricity knowledge graph through the data coordination module, transforming scattered electricity data into a structured and interconnected knowledge system, solving the problems of inconsistent data formats, low correlation, and difficulty in effective reuse in traditional systems. Secondly, the model collaboration module conducts multi-dimensional analysis based on the knowledge graph, simultaneously extracting spatiotemporal comprehensive features, anomaly detection results, and prototype user profiles, breaking through the limitations of single-dimensional analysis in existing technologies, and achieving comprehensive capture of electricity usage scenarios, abnormal behaviors, and personalized user characteristics, thus improving the completeness and accuracy of the analysis results. Then, the task execution module combines spatiotemporal features and anomaly results to generate a prediction curve for overall household electricity usage, while simultaneously outputting personalized electricity usage prediction results based on prototype user profiles. This ensures the accuracy of electricity usage trend prediction at the household level and solves the problem that traditional unified models are difficult to adapt to different combinations of household equipment and differences in electricity usage habits, making the prediction results more in line with actual needs. Finally, the early warning collaboration module generates intelligent early warning prompts based on the overall prediction curve and personalized prediction results, effectively improving the reliability of early warnings for electricity safety risks. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the background art, the accompanying drawings used in the embodiments of this application or the background art will be described below.

[0013] Figure 1 This is a block diagram of the modules of a smart home power management system provided in an embodiment of this application; Figure 2 This is a flowchart of a method for constructing a household electricity knowledge graph according to an embodiment of this application; Figure 3 This is a schematic diagram illustrating the composition of a data coordination module in a smart home power management system provided in an embodiment of this application; Figure 4 This is a schematic diagram of the composition of a model collaboration module of a smart home power management system provided in an embodiment of this application; Figure 5 This is a schematic diagram of the task execution module of a smart home power management system provided in an embodiment of this application; Figure 6 This is a schematic diagram illustrating the composition of an early warning and coordination module in a smart home power management system provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 8 This is a flowchart illustrating a smart home power management method provided in an embodiment of this application. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0015] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0016] It should be understood that the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this document indicates that the preceding and following related objects are in an "or" relationship. In the embodiments of this application, "multiple" refers to two or more.

[0017] In the embodiments of this application, "at least one item" or its similar expression refers to any combination of these items, including any combination of a single item or a plurality of items. "One or more" means one or more, while "multiple" means two or more. For example, "at least one item" of a, b, or c can represent the following seven cases: a, b, c; a and b; a and c; b and c; a, b, and c. Each of a, b, and c can be an element or a set containing one or more elements.

[0018] In this application embodiment, "connection" refers to various connection methods such as direct connection or indirect connection to realize communication between devices. This application embodiment does not limit this in any way.

[0019] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0020] The following is an explanation of the relevant terms used in this application: Smart home refers to a home system that connects various devices in the home (such as lighting, security, appliances, and environmental control) through Internet of Things (IoT) technology to achieve intelligent control, data interaction, and collaboration. It can automatically adjust the operating status of devices based on user habits and environmental changes, and also supports remote or local control by users via mobile apps, voice commands, etc.

[0021] Household electricity knowledge graph: refers to a knowledge network that integrates entities and relationships such as devices, users, environment, and electricity consumption behavior in a structured form in household electricity consumption scenarios, and transforms various types of electricity consumption data (such as device parameters, user habits, time-series power changes, external environmental influences, etc.) into a visualized knowledge structure.

[0022] Intelligent agent: refers to an intelligent module that has the ability to perceive, make decisions and act, and can independently or collaboratively complete specific tasks related to household electricity use.

[0023] Knowledge Extraction Big Language Model: This refers to a dedicated intelligent model based on natural language processing and big language model technology, specifically designed to accurately extract knowledge related to household electricity scenarios (such as entities, attributes, relationships, rules, etc.) from unstructured text data and transform it into a structured, reusable format to support the construction and updating of household electricity knowledge graphs.

[0024] Gated fusion mechanism: refers to a feature fusion technology based on gating units, which is specifically used to dynamically assign weights, remove redundant information and enhance effective information for multi-source heterogeneous features (such as spatial correlation features and temporal dynamic features in spatiotemporal analysis), and finally output comprehensive features with unified dimensions and stronger representation capabilities, so as to improve the accuracy of subsequent prediction and analysis tasks.

[0025] Dual discriminator: refers to a dual discriminator based on a generative adversarial network structure, which can use a time continuity discriminator and a device cooperation consistency discriminator to jointly determine power consumption anomalies.

[0026] With the rapid popularization of smart homes, the number and types of household electrical appliances are constantly increasing, and the interaction between devices and user behavior patterns are becoming increasingly complex. Among these, the efficiency of household electricity management not only directly affects residents' quality of life and electricity costs, but also relates to the overall energy efficiency level and load regulation capacity of the power system.

[0027] In existing home electricity management technologies, anomaly detection still largely relies on manually configured fixed rules or thresholds to identify electricity anomalies. Furthermore, the electricity analysis dimensions are limited, making it difficult to adapt to dynamic scenarios such as aging equipment, changing habits, and environmental fluctuations, leading to frequent misjudgments and missed detections. Simultaneously, different households exhibit significant individual differences in their smart home device combinations, daily electricity usage habits, and living environments. However, existing systems often employ uniform analysis and prediction models, lacking the ability to accurately capture and quickly adapt to individual user characteristics. This results in a large discrepancy between the output electricity prediction results and the actual electricity demand of households, failing to meet users' personalized electricity management needs.

[0028] Therefore, improving the accuracy of household electricity consumption forecasts and the reliability of early warnings is an urgent issue that needs to be addressed.

[0029] To address the aforementioned issues, this application provides an intelligent home electricity management system, method, and device. The intelligent home electricity management system includes: a data coordination module, a model collaboration module, a task execution module, and an early warning collaboration module. The data coordination module preprocesses a first dataset collected from the intelligent home electricity management system to obtain a second dataset; and constructs a household electricity knowledge graph based on the second dataset. The model collaboration module performs multi-dimensional analysis based on the household electricity knowledge graph to obtain multiple analysis results, including: spatiotemporal comprehensive features, anomaly detection results, and prototype user profiles. The task execution module determines a household electricity prediction curve based on the spatiotemporal comprehensive features and the anomaly detection results; and determines personalized electricity prediction results based on the prototype user profiles. The early warning collaboration module generates intelligent early warning prompts based on the household electricity prediction curve and the personalized electricity prediction results. Therefore, by using this system, the accuracy of household electricity prediction and the reliability of early warnings are improved.

[0030] For easier understanding, please refer to Figure 1 , Figure 1 This is a block diagram of the modules of a smart home electricity management system provided in an embodiment of this application. The smart home electricity management system includes: a data coordination module, a model collaboration module, a task execution module, and an early warning collaboration module, wherein: The data coordination module is used to preprocess the first dataset collected from the smart home electricity management system to obtain a second dataset; and to construct a household electricity knowledge graph based on the second dataset.

[0031] Optionally, in preprocessing the first dataset collected from the smart home electricity management system to obtain the second dataset, the data coordination module is specifically used to perform the following steps: A1. Divide the first dataset into multiple first data subsets; each first data subset corresponds to a data type; A2. Obtain the preprocessing scheme corresponding to each of the multiple first data subsets to obtain multiple preprocessing schemes; A3. Preprocess the multiple first data subsets according to the multiple preprocessing schemes to obtain multiple second data subsets; A4. Integrate the multiple second data subsets to obtain the second dataset.

[0032] In this embodiment of the application, the data types of the first dataset include, but are not limited to, time series data, topological data, and external data, and are not specifically limited here.

[0033] In a specific embodiment, the first dataset is divided into multiple first data subsets according to data type, with each first data subset corresponding to a specific data type. Then, a pre-defined preprocessing scheme library is retrieved, and preprocessing schemes are matched against each first data subset based on its data type and data quality characteristics (such as temporal continuity), resulting in matching results. Finally, a specific preprocessing scheme is determined for each first data subset based on the matching results, resulting in multiple preprocessing schemes.

[0034] Next, preprocessing operations are performed on the corresponding first data subset according to each of the multiple preprocessing schemes to obtain multiple second data subsets. Then, based on the construction requirements of the household electricity knowledge graph, the multiple second data subsets are structurally integrated to obtain a second dataset with a unified format and logical connections.

[0035] If the data type of the first data subset is time-series data (such as time-series records of device power and total electricity consumption), firstly, outliers in the data can be identified and removed using the Isolation Forest algorithm. Then, a preset orthogonal wavelet function is used to complete the denoising processing of the time-series signal. Subsequently, the processed time-series data is segmented and features are extracted using a preset first time period (such as 1 hour) as a sliding window and a preset second time period (such as 10 minutes) as a step size. Finally, a standardized second data subset with uniform structure, low noise, and no outliers is output. If the data type of the first data subset is topological data, the hierarchical structure and associated fields of the topological data are parsed to construct a graph structure containing three core nodes: device, user, and environment, as well as the relationships between nodes. Finally, a heterogeneous graph structure with multiple node types is output, which is the second data subset. If the data type of the first data subset is external data (such as weather data, electricity price data, environmental monitoring data), it is first timestamped with the electricity consumption readings of smart meters. Then, through a preset event-driven sampling mechanism, targeted data collection and feature labeling are triggered for significant external events such as price changes and weather warnings. Finally, a second dataset that is time-synchronized with the electricity consumption data and contains external event feature labels is output.

[0036] It is evident that by classifying and splitting data to avoid interference between different types, and then adapting specific preprocessing schemes to each type of data, operations such as outlier removal, format standardization, and graph structure parsing are precisely aligned with data characteristics and quality requirements. This effectively improves the purity and standardization of individual data types. Finally, through structured integration, a second dataset with a unified format and logical connections is formed. This maximizes the preservation of the core value of each type of data and provides high-quality, directly reusable basic data support for the efficient construction of the subsequent household electricity knowledge graph.

[0037] Optional, please refer to Figure 2 , Figure 2 This is a flowchart illustrating a method for constructing a household electricity knowledge graph according to an embodiment of this application. In the step of constructing the household electricity knowledge graph based on the second dataset, the data coordination module is specifically used to execute... Figure 2 The steps shown are as follows: B1. Identify and extract multiple nodes from the second dataset; the multiple nodes include device nodes, user nodes, and environmental event nodes; B2. Obtain the association information between any two nodes among the multiple nodes through a preset semantic analysis method to obtain the target association information set; B3. Construct the household electricity knowledge graph based on the target-related information set and the multiple nodes.

[0038] In this embodiment of the application, the second dataset includes, but is not limited to: time-series sample datasets, heterogeneous graph structures, and datasets labeled with external event features, and is not specifically limited here.

[0039] In a specific embodiment, firstly, based on preset entity recognition rules, entity extraction is performed on the second dataset to identify and extract multiple nodes in the second dataset. These multiple nodes include device nodes (such as smart home device entities like air conditioners, washing machines, and smart meters), user nodes (such as household residents and users with electricity usage rights), and environmental event nodes (such as environmental event entities that affect electricity consumption behavior, such as sudden temperature changes, electricity price adjustments, and weather warnings). At the same time, each node is assigned a unique identifier and attribute features, such as the rated power and service life of device nodes, the electricity consumption preferences and work and rest periods of user nodes, and the occurrence time and impact level of environmental event nodes, which are not specifically limited here.

[0040] Then, a pre-defined semantic analysis model is invoked to perform semantic parsing on the data in the second dataset, mining and obtaining the association information between any two nodes among multiple nodes, thus obtaining the target association information set. The target association information set includes: usage association between device nodes and user nodes, response association between device nodes and environmental event nodes, and decision association between user nodes and environmental event nodes. Each target association information set contains attributes such as association type, association strength, and association time sequence characteristics.

[0041] Finally, using the extracted device nodes, user nodes, and environmental event nodes as graph entities, and the relationships in the target association information set as edges between entities, combined with the attribute characteristics of each node and the edge, a pre-set graph database is used to complete data storage and graph visualization construction, resulting in a household electricity knowledge graph.

[0042] It should be noted that the household electricity knowledge graph includes device nodes, user nodes, and environmental event nodes. Among them, device nodes use rated power and expected lifespan as core attributes to accurately quantify the energy consumption capacity and usage cycle of devices, providing basic data for analyzing the upper limit of device energy consumption and the electricity cost over its life cycle. User nodes focus on daily pattern attributes to capture differences in users' electricity consumption behavior over time (such as turning on kitchen appliances at 7 am on weekdays and using entertainment devices at 10 pm on weekends), which is a key basis for building personalized electricity consumption predictions and identifying abnormal behaviors. Environmental event nodes quantify the degree of external environmental intervention on electricity consumption through energy impact factor attributes (such as high temperatures increasing air conditioner energy consumption by 30% and peak electricity prices reducing user electricity demand by 20%), providing support for analyzing the linkage between the environment and electricity consumption and optimizing electricity consumption strategies.

[0043] It is evident that transforming preprocessed standardized data into a household electricity knowledge graph containing entities and relationships not only achieves the associative integration and knowledge-based expression of multiple types of data, but also clearly presents the inherent connections between equipment operation, user behavior, and environmental impact in electricity usage scenarios.

[0044] For easier understanding, please refer to Figure 3 , Figure 3 This is a schematic diagram illustrating the composition of a data coordination module in a smart home electricity management system provided in this application embodiment. The data coordination module includes a data coordination agent and a knowledge extraction large language model. The data coordination agent is responsible for processing time-series data, topological data, and external data. Through differentiated preprocessing schemes (such as outlier removal, wavelet denoising, graph structure generation, and time alignment), it transforms multi-source heterogeneous data into standardized data. The knowledge extraction large language model is used to process unstructured data. Through semantic analysis, entity recognition, and other technologies, it extracts knowledge features such as user electricity preferences and device function descriptions from textual information. By inputting time-series data (such as time-series records of device power and electricity consumption), topological data (such as device connection relationships), and external data (such as weather, electricity prices, and other external information affecting electricity consumption) into the data coordination agent, a first part of data is obtained. Then, unstructured data (such as textual descriptions of user electricity habits and device manuals) is input into the knowledge extraction large language model for processing, resulting in a second part of data. Finally, the first and second parts of data are integrated to construct a home electricity knowledge graph.

[0045] The model collaboration module is used to perform multi-dimensional analysis based on the household electricity knowledge graph to obtain multiple analysis results; the multiple analysis results include: spatiotemporal integrated features, anomaly detection results, and prototype user profiles.

[0046] Optionally, the model collaboration module includes a spatiotemporal analysis agent, an anomaly detection agent, and a meta-learning agent. The spatiotemporal analysis agent includes a spatial graph convolutional neural network and a temporal graph convolutional neural network. Regarding the multi-dimensional analysis based on the household electricity knowledge graph to obtain multiple analysis results, the model collaboration module is specifically used to perform the following steps: C1. The spatial association features and temporal dynamic features in the household electricity knowledge graph are obtained through the spatial graph convolutional neural network and the temporal graph convolutional neural network, respectively. C2. The spatiotemporal analysis agent fuses the spatial correlation features and the temporal dynamic features based on a preset gating fusion mechanism to obtain the spatiotemporal comprehensive features. C3. The anomaly detection agent judges the electricity consumption data in the household electricity knowledge graph based on a preset dual discriminator to obtain the anomaly detection result. C4. The meta-learning agent analyzes the electricity consumption data in the household electricity knowledge graph according to a preset model-independent meta-learning algorithm to obtain the prototype user profile.

[0047] In this embodiment of the application, the architecture of the dual discriminator is a dual discriminator architecture based on a generative adversarial network structure, including a generator component and a discriminator component. The discriminator component includes a time continuity discriminator and a device cooperation consistency discriminator.

[0048] In a specific embodiment, firstly, the spatial graph convolutional neural network in the spatiotemporal analysis agent is invoked. Taking the device topology and user-device spatial relationships of the household electricity knowledge graph as input, the spatial dependency features of entity nodes in the graph (such as the power linkage features between living room and bedroom air conditioners, and the clustered electricity consumption features of kitchen equipment) are extracted through neighborhood aggregation operations of graph convolution, thus obtaining spatial association features. Simultaneously, the temporal graph convolutional neural network is invoked. Taking time-stamped electricity consumption data in the knowledge graph (such as the power and electricity consumption of each node changing over time) as input, the temporal evolution patterns of electricity consumption behavior are captured through a combination of time-step convolution and gated recurrent units (such as the peak electricity consumption migration features during weekday morning rush hours and the time association features of device start-up and shutdown), thus obtaining temporal dynamic features.

[0049] Then, through the gating fusion mechanism built into the spatiotemporal analysis agent, the spatial correlation features and temporal dynamic features are weighted and fused: the topological correlation weight of spatial features and the temporal change weight of temporal features are dynamically adjusted through the gating unit to eliminate redundant features and strengthen core correlation features, and finally output spatiotemporal comprehensive features.

[0050] Next, through the anomaly detection agent, the electricity consumption data (including time-series power and device collaborative power consumption data) in the household electricity knowledge graph is input into the preset dual discriminator. At the same time, the generator component generates reference data that conforms to normal electricity consumption patterns. Through adversarial training between the discriminator component and the generator component, the accurate identification of abnormal electricity consumption patterns (such as abnormal device overload and abnormal power consumption startup during uncommon periods) is achieved. The anomaly detection results, including anomaly type, abnormal time period, and anomaly confidence level, are output.

[0051] Finally, through the meta-learning agent, a pre-defined model-independent meta-learning algorithm is adopted. The user electricity behavior data in the household electricity knowledge graph is used as the meta-training set. In the meta-training stage, the general features of different users' electricity behavior are learned through the meta-training set. In the meta-testing stage, the personalized electricity characteristics of specific users are quickly adapted. The core representation of user electricity behavior is obtained through gradient descent optimization. Finally, a prototype user profile containing user electricity preferences, behavior deviation thresholds, and device usage patterns is output.

[0052] As can be seen, by using spatial graph convolutional neural networks and temporal graph convolutional neural networks, the spatial correlation features and temporal dynamic features of electricity consumption data in the knowledge graph can be accurately extracted. These features are then combined through a gating fusion mechanism to form complementary spatiotemporal comprehensive features. A dual discriminator is then used to accurately identify electricity consumption anomalies. Based on the model-independent meta-learning algorithm, a prototype user profile that fits user habits can be quickly constructed. This not only breaks through the limitations of single-dimensional analysis and achieves comprehensive capture of electricity consumption scenarios, abnormal behaviors, and user characteristics, but also improves the accuracy of feature extraction, the reliability of anomaly detection, and the adaptability of user profiles.

[0053] For easier understanding, please refer to Figure 4 , Figure 4 This is a schematic diagram illustrating the composition of a model collaboration module in a smart home electricity management system provided in this application embodiment. The model collaboration module includes a spatiotemporal analysis agent, an anomaly detection agent, and a meta-learning agent. The spatiotemporal analysis agent receives data from a household electricity knowledge graph and extracts spatial correlation features (such as topological linkage of device clusters) and temporal dynamic features (such as the temporal variation of peak electricity consumption) from the electricity data through internal spatial graph convolutional neural networks and temporal graph convolutional neural networks, ultimately outputting comprehensive spatiotemporal features. The anomaly detection agent, based on the household electricity knowledge graph, uses a preset dual discriminator to determine anomalies in the electricity data, identifying abnormal situations such as device overload and electricity consumption during uncommon periods, and outputs anomaly detection results (including anomaly type, time period, confidence level, etc.). The meta-learning agent can analyze user electricity habits, preferences, and other characteristics based on user behavior data in the household electricity knowledge graph using a model-independent meta-learning algorithm, constructing a prototype user profile (including user behavior features and behavior deviation thresholds).

[0054] The task execution module is used to determine the household electricity consumption prediction curve based on the spatiotemporal integrated features and the anomaly detection results; and to determine the personalized electricity consumption prediction results based on the prototype user profile.

[0055] Optionally, in determining the household electricity consumption prediction curve based on the spatiotemporal integrated features and the anomaly detection results, the task execution module is specifically used to perform the following steps: D1. Align the spatiotemporal integrated features and the anomaly detection results according to the time dimension and then perform feature fusion to obtain a comprehensive feature sequence; D2. Input the comprehensive feature sequence into the preset time series prediction model and output the preliminary prediction curve; D3. Based on the anomaly type in the anomaly detection results, the preliminary prediction curve is corrected to obtain the household electricity consumption prediction curve.

[0056] In this embodiment of the application, the time series prediction model is a model pre-trained based on historical electricity consumption data in the household electricity knowledge graph, including but not limited to: Long Short-Term Memory Network (LSTM), Time Series Transformer (TST), or Spatiotemporal Fusion Transformer model, without specific limitations.

[0057] In a specific embodiment, firstly, the spatiotemporal comprehensive features (including spatial topological association features and temporal dynamic evolution features of household electricity) and the anomaly detection results (including anomaly type, anomaly period, anomaly confidence level, associated devices, etc.) are aligned according to a unified timestamp dimension (e.g., using 10-minute time granularity, the spatiotemporal features of each time node are matched with the anomaly detection markers of the corresponding period). Then, through feature splicing and attention weighted fusion mechanism, differentiated weights are assigned to the spatiotemporal features of the anomaly period (e.g., the feature weight of the device overload anomaly period is increased by 20%), and irrelevant redundant information is eliminated to generate a comprehensive feature sequence with unified dimensions and continuous temporal sequence.

[0058] Then, the comprehensive feature sequence is input into the preset time series prediction model. Through joint learning of the temporal dependence and spatial correlation features of the comprehensive feature sequence, a preliminary prediction curve reflecting the change of total power consumption / electricity consumption of the household over time in the future preset period is output.

[0059] Next, based on the anomaly types in the anomaly detection results, a preset differential correction strategy is used to dynamically adjust the preliminary prediction curve to obtain the household electricity consumption prediction curve. Specifically, the differential correction strategy is as follows: If the anomaly type is instantaneous equipment overload: based on the rated power of the overloaded equipment and the historical overload duration, the power peak value of the preliminary prediction curve for the corresponding time period is lowered (e.g., by 10% of the overload confidence level × the rated power of the equipment), and this time period is marked as a power fluctuation risk period; if the anomaly type is equipment startup during uncommon times: combined with the equipment usage habits in the prototype user profile, the predicted power baseline for the corresponding time period is corrected (e.g., the air conditioner startup power during uncommon times in the early morning is corrected from a high value to the user's normal standby power); if the anomaly type is multi-device collaborative anomaly: based on the historical collaborative electricity consumption patterns of the device cluster, the power superposition value of the associated devices for the corresponding time period is adjusted to eliminate prediction deviations caused by abnormal collaboration.

[0060] As can be seen, by aligning and fusing spatiotemporal integrated features with anomaly detection results along the time dimension to obtain a comprehensive feature sequence, and then inputting it into a time series prediction model to generate a preliminary prediction curve, the model can perform differentiated corrections based on anomaly types to achieve accurate prediction of household electricity consumption trends.

[0061] Optionally, the prototype user profile includes user behavior characteristics and behavior deviation thresholds. In determining the personalized electricity consumption prediction result based on the prototype user profile, the task execution module is specifically used to perform the following steps: E1. Obtain the electricity consumption data corresponding to the user behavior characteristics in the household electricity consumption knowledge graph to obtain the first electricity consumption data; E2. Determine the basic feature set based on the user behavior characteristics and the first electricity consumption data; E3. Obtain the historical electricity consumption time series data corresponding to the household electricity consumption knowledge graph; E4. Align the basic feature set and the historical electricity consumption time series data according to the time dimension and then perform feature fusion to obtain a time series feature sequence; E5. Input the time series feature sequence into the time series prediction model and output the preliminary prediction result; E6. Correct the preliminary prediction result according to the behavioral deviation threshold to obtain the personalized electricity consumption prediction result.

[0062] In a specific embodiment, firstly, based on user behavior characteristics in the prototype user profile, targeted data retrieval is performed in the household electricity knowledge graph to extract electricity consumption data strongly correlated with user behavior characteristics (such as historical power data matching the characteristic of "kitchen electricity consumption at 7 am"), thus obtaining the first electricity consumption data. Then, the user behavior characteristics are quantified and encoded (such as encoding "washing machine used 3 times a week" as a frequency feature value), and combined with the statistical characteristics of the first electricity consumption data (such as mean, peak value, and variance) for feature integration, constructing a basic feature set containing the user's personalized electricity consumption patterns. The basic feature set covers core dimensions such as time preference characteristics, device usage characteristics, and power selection characteristics.

[0063] Next, historical electricity consumption time-series data associated with the target user is retrieved from the household electricity consumption knowledge graph. This historical electricity consumption time-series data includes power change curves by device and time period, electricity consumption statistics, and electricity consumption adjustment data related to environmental factors (temperature, weather). Then, the basic feature set and the historical electricity consumption time-series data are aligned in a unified time granularity. An attention mechanism is used to assign differentiated weights to the basic features at different time points (e.g., increasing the feature weight for user's habitual electricity consumption periods by 30%). Finally, feature splicing is used to complete the fusion process and generate a time-series feature sequence.

[0064] Finally, the time-series feature sequence is input into the time-series prediction model. Utilizing the user electricity consumption time-series dependency patterns already learned by the model, preliminary prediction results are output, including user device usage probability, time-period power estimation, and device runtime. Then, based on the behavioral deviation threshold in the prototype user profile, differentiated corrections are performed on the preliminary prediction results to obtain personalized electricity consumption prediction results. The specific steps of the differentiated correction are as follows: If the power value of a device in the preliminary prediction results exceeds the power fluctuation range in the behavioral deviation threshold, the predicted power is adjusted downwards by the upper limit of the threshold and marked as a power deviation risk item; if the usage period of a device in the preliminary prediction results exceeds the time-period offset duration in the behavioral deviation threshold, the predicted time period is adjusted based on the user's work-rest patterns, or the device usage probability during that period is reduced; if the environmental response behavior in the preliminary prediction results deviates from the threshold, the device runtime is corrected by referring to the user's historical environmental response patterns.

[0065] As can be seen, by first relying on the knowledge graph of household electricity consumption to accurately extract user behavior feature correlation data and construct a basic feature set, and then merging it with historical electricity consumption time series data to form a feature sequence that combines personalization and time series characteristics, and then using a time series prediction model for preliminary prediction and combining it with behavioral deviation threshold for targeted correction, we can not only fully explore the personalized core characteristics of users' electricity consumption habits, but also improve the accuracy of prediction through historical time series patterns and behavioral deviation correction, effectively breaking through the limitation of traditional unified prediction ignoring individual differences of users, and outputting personalized prediction results that fit users' electricity consumption habits.

[0066] For easier understanding, please refer to Figure 5 , Figure 5 This is a schematic diagram illustrating the composition of a task execution module in a smart home electricity management system provided in this application embodiment. The task execution module includes a predictive agent and a personalized agent. The predictive agent receives spatiotemporal integrated features and anomaly detection results, and through a preset time-series prediction model, first fuses the features and generates a preliminary prediction curve, then corrects it based on the anomaly type, and finally outputs a household electricity consumption prediction curve. The personalized agent, based on a prototype user profile, extracts electricity consumption data associated with user behavior features, constructs a basic feature set, fuses it with historical electricity consumption time-series data, inputs it into the time-series prediction model, and corrects it based on a behavior deviation threshold, outputting a personalized electricity consumption prediction result.

[0067] The early warning coordination module is used to generate intelligent early warning prompts based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results.

[0068] Optionally, the early warning coordination module includes a monitoring agent and a coordinating agent. In generating intelligent early warning prompts based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results, the early warning coordination module specifically performs the following steps: F1. Obtain the historical electricity consumption patterns and user electricity consumption preferences corresponding to the household electricity consumption knowledge graph; F2. Determine a dynamic threshold range set based on the historical electricity consumption patterns and user electricity consumption preferences; the dynamic threshold range set includes multiple dynamic threshold ranges and multiple time points, with each dynamic threshold range corresponding to a time point. F3. Obtain the household electricity consumption prediction curve and the personalized electricity consumption prediction result at multiple time points, respectively, and obtain multiple second electricity consumption data and multiple third electricity consumption data. F4. Through the monitoring intelligent agent, abnormal power consumption data in the multiple second power consumption data and the multiple third power consumption data are determined according to the multiple dynamic threshold ranges, resulting in a abnormal power consumption data; a is an integer greater than 0. F5. Generate a preliminary alarm based on the a abnormal power consumption data; F6. Through the coordinating agent, the a preliminary alarms are verified and filtered to obtain b reference alarms; b is an integer greater than 0 and less than or equal to a. F7. Based on the preset warning interpretation model, generate the intelligent warning prompt according to the b reference alarms.

[0069] In a specific embodiment, firstly, historical electricity consumption patterns (such as power fluctuation patterns at different time points, device co-use habits, and seasonal electricity consumption trends) and user electricity consumption preferences (such as device power selection at specific time points, priority of core device use, and environmentally adaptable electricity consumption tendencies) associated with the target user are retrieved from the household electricity consumption knowledge graph. Quantifiable feature data is then extracted as the basis for threshold setting. Based on the time-point distribution characteristics of historical electricity consumption patterns and the weight parameters of user electricity consumption preferences, a preset statistical modeling method (such as quantile regression and normal distribution fitting) is used to determine the dynamic threshold range set. The dynamic threshold range set contains multiple discrete time points, with each time point corresponding to a specific dynamic threshold range.

[0070] Next, based on the timestamps of multiple time points, data such as total power consumption and total electricity consumption at the corresponding time points are extracted from the household electricity consumption forecast curve to form multiple second electricity consumption data. Data such as sub-device power, device usage probability, and running time at the corresponding time points are extracted from the personalized electricity consumption forecast results to form multiple third electricity consumption data. It is ensured that the second and third electricity consumption data correspond one-to-one with the time points in the dynamic threshold range.

[0071] Then, through the monitoring agent, the second and third power consumption data corresponding to each time point are compared with the dynamic threshold range of that time point. If a power consumption data exceeds the upper limit or falls below the lower limit of the corresponding dynamic threshold range, it is determined to be abnormal power consumption data. All determination results are then summarized to obtain 'a' abnormal power consumption data. Based on 'a' abnormal power consumption data, 'a' preliminary alarms are generated. Each preliminary alarm contains core information: abnormal time point, actual predicted value, dynamic threshold range, abnormal deviation magnitude, and associated device identifier, which are not specifically limited here.

[0072] Next, by coordinating the intelligent agent and combining auxiliary information from the household electricity knowledge graph (such as environmental data corresponding to a specific time point, user schedules, equipment rated parameters, and historical anomaly handling records), the initial alarms (a) are verified and filtered. For example, it can determine whether the anomaly matches the current scenario (e.g., whether exceeding the air conditioner power threshold in high-temperature weather is a reasonable cooling requirement). It can also classify the corresponding risk levels (e.g., high-risk, medium-risk, and low-risk levels) based on the magnitude of the anomaly deviation or the importance of related equipment. Then, the initial alarms that fail verification and have a risk level greater than or equal to the medium-risk level are filtered out, thus obtaining b reference alarms.

[0073] Finally, the pre-set warning interpretation language model is invoked, and the complete information of b reference alarms and related background data in the household electricity knowledge graph (such as user electricity preferences and device attributes) are input. The warning interpretation language model generates intelligent warning prompts through semantic parsing and logical reasoning, including the core cause of the anomaly, potential risks, and targeted optimization suggestions. Finally, it is integrated to form a clear and easy-to-understand intelligent warning prompt, which is then pushed to the user terminal (such as the user's APP interface and smart speaker).

[0074] As can be seen, the system first determines a dynamic threshold range set for matching time points based on historical electricity consumption patterns and user preferences from the household electricity knowledge graph. It then accurately compares the corresponding data of the household electricity prediction curve and personalized prediction results to generate an initial alarm. Next, the system is verified and filtered by a coordinated intelligent agent to eliminate false alarms and low-risk alarms. Finally, the system generates intelligent early warning prompts containing reasons, risks, and suggestions through a large-scale early warning interpretation model. This approach not only breaks through the limitations of traditional fixed threshold alarms, achieving dynamic and accurate early warnings based on user habits and time scenarios, but also improves the reliability of alarms through verification and filtering. Furthermore, it enhances user perception and acceptance through interpretable prompts, comprehensively ensuring electricity safety and optimizing the user's electricity experience.

[0075] For easier understanding, please refer to Figure 6 , Figure 6This is a schematic diagram illustrating the composition of an early warning and coordination module in a smart home electricity management system provided in this application embodiment. The early warning and coordination module includes a monitoring agent and a coordinating agent. The monitoring agent receives household electricity consumption forecast curves and personalized electricity consumption forecast results. Based on a dynamic threshold range set (generated from historical electricity consumption patterns and user preferences within a household electricity consumption knowledge graph), it performs anomaly detection on the data, identifies electricity consumption data exceeding the dynamic threshold range, and generates *a* preliminary alarms (where *a* is a positive integer). Each preliminary alarm includes information such as the abnormal time point, associated data type, and deviation magnitude. The coordinating agent, after receiving *a* preliminary alarms, verifies and filters them using auxiliary information from the household electricity consumption knowledge graph (such as environmental data, user schedules, and device parameters), eliminating false alarms and low-risk alarms, and finally outputs *b* reference alarms (where *b* is a positive integer less than or equal to *a*).

[0076] In one possible implementation, the coordinating agent can verify the validity of the initial alert through multi-dimensional contextual reasoning: Environmental scenario matching: Determine whether the anomaly is consistent with the current environment (e.g., the air conditioner power slightly exceeds the threshold in hot weather, which may be a reasonable cooling requirement). User schedule matching: Combine user schedules (e.g., "The user is working overtime tonight, and the oven is running at high power even though no one is home") to determine if the behavior is abnormal. Device collaboration scenario matching: Analyze the linkage logic between devices (such as "whether it is in line with user habits for the dryer to start at high power at the same time when the washing machine is running).

[0077] Simultaneously, the coordinating agent can prioritize alarms based on the severity of the anomaly (e.g., short-circuit risk > normal power fluctuation), confidence level (probability of determining an anomaly), and potential impact (e.g., whether it will cause line overload), ultimately filtering out high-confidence, high-risk reference alarms (eliminating false alarms and low-risk alarms). The coordinating agent can also collect user response data to historical alarms (e.g., whether users accept the "refrigerator power abnormality alarm in the early morning"), and use this data to optimize the anomaly judgment logic (e.g., adjusting the threshold sensitivity of certain smart home devices), the calculation rules for dynamic threshold ranges (e.g., fine-tuning dynamic thresholds based on seasonal changes), and the rule base for scenario reasoning (e.g., adding reasoning logic such as "high power operation of water heater on rainy days is a reasonable need"), achieving long-term alignment between system decisions and actual user needs.

[0078] In one possible implementation, a large language model for early warning interpretation can be used to verify reference alarms. If the reference alarm is indeed verified to originate from the initial alarm, the large language model will generate a corresponding natural language explanation (such as "The living room light power is 0.5kW at 10 pm, which exceeds the normal threshold of 0.1kW-0.3kW, possibly indicating a device malfunction") and push it to the user. If the reference alarm is not verified to originate from the initial alarm, feedback will be given to the monitoring agent and the coordination agent to trigger the process of regenerating the initial alarm and re-filtering it, ensuring the traceability and reliability of the alarm.

[0079] The following is combined Figure 7 The electronic devices in the embodiments of this application will be described. Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 7 As shown, the electronic device includes one or more processors, a memory, a communication interface, and one or more programs. The processor is connected to the memory and the communication interface via an internal communication bus.

[0080] The processor can be used for: The first dataset collected from the smart home electricity management system is preprocessed to obtain the second dataset; Construct a knowledge graph of household electricity consumption based on the second dataset; Based on the aforementioned household electricity knowledge graph, a multi-dimensional analysis was performed to obtain multiple analysis results; these multiple analysis results include: spatiotemporal comprehensive features, anomaly detection results, and prototype user profiles. The household electricity consumption prediction curve is determined based on the spatiotemporal integrated characteristics and the anomaly detection results. Personalized electricity consumption forecasts are determined based on the prototype user profile. Intelligent early warning prompts are generated based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results.

[0081] The one or more programs are stored in the aforementioned memory and configured to be executed by the aforementioned processor, and the one or more programs include instructions for performing any of the steps in the above embodiments.

[0082] The processor can be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, cells, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc. The communication unit can be a communication interface, transceiver, transceiver circuit, etc., and the storage unit can be a memory.

[0083] The memory can be volatile or non-volatile, or a combination of both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0084] It is understood that the electronic device may include more or fewer structural elements than those shown in the block diagram above, such as a power module, physical buttons, a Wi-Fi module, a speaker, a Bluetooth module, sensors, a display module, etc., without limitation. It is understood that the electronic device may incorporate elements such as... Figure 1 The aforementioned functional modules.

[0085] The following is combined Figure 8 This application describes a smart home power management method based on an embodiment of the present application. Figure 8 This is a flowchart illustrating a smart home power management method provided in an embodiment of this application, specifically including the following steps: S1. Preprocess the first dataset collected from the smart home electricity management system to obtain the second dataset; S2. Construct a knowledge graph of household electricity consumption based on the second dataset; S3. Based on the household electricity knowledge graph, perform multi-dimensional analysis to obtain multiple analysis results; the multiple analysis results include: spatiotemporal comprehensive features, anomaly detection results, and prototype user profiles; S4. Determine the household electricity consumption prediction curve based on the spatiotemporal integrated characteristics and the anomaly detection results; S5. Determine personalized electricity consumption prediction results based on the prototype user profile; S6. Generate intelligent early warning prompts based on the household electricity consumption prediction curve and the personalized electricity consumption prediction results.

[0086] As can be seen, by preprocessing the first dataset and constructing a household electricity consumption knowledge graph, the scattered electricity consumption data is transformed into a structured and interconnected knowledge system, solving the problems of inconsistent data formats, low correlation, and difficulty in effective reuse in traditional systems. Then, multi-dimensional analysis is conducted based on the knowledge graph, simultaneously extracting spatiotemporal comprehensive features, anomaly detection results, and prototype user profiles. This overcomes the limitations of single-dimensional analysis in existing technologies, achieving comprehensive capture of electricity consumption scenarios, abnormal behaviors, and personalized user characteristics, thus improving the completeness and accuracy of the analysis results. Next, a household overall electricity consumption prediction curve is generated by combining spatiotemporal features and anomaly results. Simultaneously, personalized electricity consumption prediction results are output based on prototype user profiles. This ensures the accuracy of household-level electricity consumption trend predictions and solves the problem that traditional unified models are difficult to adapt to different household equipment combinations and differences in electricity consumption habits, making the prediction results more aligned with actual needs. Finally, intelligent early warning prompts are generated based on the overall prediction curve and personalized prediction results, effectively improving the reliability of early warnings for electricity safety risks.

[0087] This application also provides a computer-readable storage medium storing a computer program for electronic data interchange that causes a computer to perform some or all of the steps of the smart home power management method described in the above embodiments, wherein the computer includes an electronic device.

[0088] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of the smart home power management method described in the above embodiments. The computer program product can be a software installation package, and the computer includes an electronic device.

[0089] It should be noted that, for the sake of simplicity, the above embodiments are all described as a series of actions. Those skilled in the art should understand that this application is not limited to the described order of actions, as some steps in the embodiments of this application can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions, steps, modules, or units involved are not necessarily essential to the embodiments of this application.

[0090] In the above embodiments, the descriptions of each embodiment in this application have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0091] Those skilled in the art will understand that implementing all or part of the processes in the above embodiments can be accomplished by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

[0092] The steps of the methods or algorithms described in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in RAM, flash memory, ROM, EPROM, electrically erasable programmable read-only memory (EEPROM), registers, hard disk, portable hard disk, read-only optical disk (CD-ROM), or any other form of storage medium well known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Furthermore, the ASIC can reside in a terminal device or management device. Alternatively, the processor and storage medium can exist as discrete components in the terminal device or management device.

[0093] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in the embodiments of this application can be implemented, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0094] The modules / units included in the various devices and products described in the above embodiments can be software modules / units, hardware modules / units, or a combination of both. For example, for devices and products applied to or integrated into a chip, all modules / units can be implemented using hardware methods such as circuits, or at least some modules / units can be implemented using software programs that run on a processor integrated within the chip, while the remaining (if any) modules / units can be implemented using hardware methods such as circuits. For devices and products applied to or integrated into a chip module, all modules / units can be implemented using hardware methods such as circuits. Different modules / units can be located in the same component (e.g., chip, circuit module, etc.) or different components of the chip module, or at least some modules / units can be implemented using hardware methods such as circuits. The implementation is achieved through a software program that runs on the processor integrated within the chip module. The remaining modules / units (if any) can be implemented using hardware methods such as circuits. For various devices and products applied to or integrated into terminal equipment, each of their modules / units can be implemented using hardware methods such as circuits. Different modules / units can be located in the same component (e.g., chip, circuit module, etc.) or different components within the terminal equipment. Alternatively, at least some modules / units can be implemented through a software program that runs on the processor integrated within the terminal equipment, while the remaining modules / units (if any) can be implemented using hardware methods such as circuits.

[0095] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the embodiments of this application. It should be understood that the above descriptions are merely specific embodiments of the embodiments of this application and are not intended to limit the protection scope of the embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solutions of the embodiments of this application should be included within the protection scope of the embodiments of this application.

Claims

1. A smart home electricity management system, characterized in that, The smart home power management system includes: a data coordination module, a model collaboration module, a task execution module, and an early warning collaboration module, wherein: The data coordination module is used to preprocess the first dataset collected from the smart home electricity management system to obtain a second dataset; and to construct a household electricity knowledge graph based on the second dataset. The model collaboration module is used to perform multi-dimensional analysis based on the household electricity knowledge graph to obtain multiple analysis results; the multiple analysis results include: spatiotemporal integrated features, anomaly detection results, and prototype user profiles. The task execution module is used to determine the household electricity consumption prediction curve based on the spatiotemporal integrated features and the anomaly detection results; and to determine the personalized electricity consumption prediction results based on the prototype user profile. The early warning coordination module is used to generate intelligent early warning prompts based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results.

2. The system as described in claim 1, characterized in that, In the process of preprocessing the first dataset collected from the smart home electricity management system to obtain the second dataset, the data coordination module is specifically used for: The first dataset is divided into multiple first data subsets; each first data subset corresponds to a data type. Obtain the preprocessing scheme corresponding to each of the plurality of first data subsets to obtain a plurality of preprocessing schemes; The multiple first data subsets are preprocessed according to the multiple preprocessing schemes to obtain multiple second data subsets; The multiple second data subsets are integrated to obtain the second dataset.

3. The system as described in claim 2, characterized in that, In the process of constructing a household electricity knowledge graph based on the second dataset, the data coordination module is specifically used for: Identify and extract multiple nodes from the second dataset; the multiple nodes include device nodes, user nodes, and environmental event nodes; The association information between any two nodes among the multiple nodes is obtained by using a preset semantic analysis method, thus obtaining the target association information set; The household electricity knowledge graph is constructed based on the target-related information set and the multiple nodes.

4. The system as described in claim 3, characterized in that, The model collaboration module includes a spatiotemporal analysis agent, an anomaly detection agent, and a meta-learning agent. The spatiotemporal analysis agent includes a spatial graph convolutional neural network and a temporal graph convolutional neural network. Regarding the multi-dimensional analysis based on the household electricity knowledge graph to obtain multiple analysis results, the model collaboration module is specifically used for: The spatial graph convolutional neural network and the temporal graph convolutional neural network are used to obtain the spatial association features and temporal dynamic features in the household electricity knowledge graph, respectively. The spatiotemporal analysis agent fuses the spatial correlation features and the temporal dynamic features based on a preset gating fusion mechanism to obtain the spatiotemporal comprehensive features. The anomaly detection agent uses a preset dual discriminator to judge the electricity consumption data in the household electricity knowledge graph to obtain the anomaly detection result. The meta-learning agent analyzes the electricity consumption data in the household electricity knowledge graph according to a preset model-independent meta-learning algorithm to obtain the prototype user profile.

5. The system according to any one of claims 1-4, characterized in that, In determining the household electricity consumption prediction curve based on the spatiotemporal integrated features and the anomaly detection results, the task execution module is specifically used for: The spatiotemporal integrated features and the anomaly detection results are aligned along the time dimension and then fused to obtain a comprehensive feature sequence. The comprehensive feature sequence is input into a preset time-series prediction model, and a preliminary prediction curve is output. The preliminary prediction curve is corrected based on the anomaly type in the anomaly detection results to obtain the household electricity consumption prediction curve.

6. The system as described in claim 5, characterized in that, The prototype user profile includes user behavior characteristics and behavior deviation thresholds. Regarding the determination of personalized electricity consumption prediction results based on the prototype user profile, the task execution module is specifically used for: Obtain the electricity consumption data corresponding to the user behavior characteristics in the household electricity consumption knowledge graph to obtain the first electricity consumption data; A basic feature set is determined based on the user behavior characteristics and the first electricity consumption data; Obtain the historical electricity consumption time series data corresponding to the household electricity consumption knowledge graph; The basic feature set and the historical electricity consumption time series data are aligned along the time dimension and then feature fusion is performed to obtain a time series feature sequence. The time-series feature sequence is input into the time-series prediction model, and preliminary prediction results are output. The preliminary prediction result is corrected based on the behavioral deviation threshold to obtain the personalized electricity consumption prediction result.

7. The system as described in claim 6, characterized in that, The early warning coordination module includes a monitoring agent and a coordinating agent. Specifically, in generating intelligent early warning prompts based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results, the early warning coordination module is used for: Obtain the historical electricity consumption patterns and user electricity consumption preferences corresponding to the household electricity consumption knowledge graph; A dynamic threshold range set is determined based on the historical electricity consumption patterns and user electricity consumption preferences; the dynamic threshold range set includes multiple dynamic threshold ranges and multiple time points, with each dynamic threshold range corresponding to a time point. The household electricity consumption prediction curve and the personalized electricity consumption prediction result are respectively obtained at multiple time points, along with multiple second and third electricity consumption data. The monitoring agent determines abnormal power consumption data from the multiple second power consumption data and the multiple third power consumption data based on the multiple dynamic threshold ranges, resulting in a abnormal power consumption data; where a is an integer greater than 0. Generate a preliminary alarm based on the a abnormal power consumption data; The coordinating agent verifies and filters the a preliminary alarms to obtain b reference alarms. b is an integer greater than 0 and less than or equal to a; The intelligent early warning prompt is generated based on the b reference alarms using a pre-set early warning interpretation model.

8. A smart home electricity management method, characterized in that, The method includes: The first dataset collected from the smart home electricity management system is preprocessed to obtain the second dataset; Construct a knowledge graph of household electricity consumption based on the second dataset; Based on the aforementioned household electricity knowledge graph, a multi-dimensional analysis was performed to obtain multiple analysis results; these multiple analysis results include: spatiotemporal comprehensive features, anomaly detection results, and prototype user profiles. The household electricity consumption prediction curve is determined based on the spatiotemporal integrated characteristics and the anomaly detection results. Personalized electricity consumption forecasts are determined based on the prototype user profile. Intelligent early warning prompts are generated based on the household electricity consumption forecast curve and the personalized electricity consumption forecast results.

9. An electronic device, characterized in that, include: Processor, memory, communication interface, and one or more programs; The one or more programs are stored in the memory and configured to be executed by the processor, the programs including instructions for performing the steps in the method of claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method as described in claim 8.

Images