Smart home energy control system based on energy consumption prediction

Abstract

The invention discloses a smart home energy control system based on energy consumption prediction. According to the invention, through the adaptive learning optimization module, dynamic precision of energy management and high adaptation of user demands are realized. The system can continuously learn user living habits and environment change rules, automatically recognize behavior modes such as morning lighting starting preference and summer air conditioner temperature setting intervals, and generate an equipment control strategy considering both economical efficiency and comfort in combination with real-time electricity price fluctuation and a power grid load state. Under the scene of sudden power grid demand response or extreme weather, the system can quickly adjust the operation priority of equipment, for example, the system can automatically delay starting of the washing machine in the electricity price peak period or preheat the indoor temperature before cold waves come, so that energy waste is avoided, and the living quality is maintained. According to the active control mode based on prediction, the passive response of a traditional smart home is converted into prospective optimization, and the energy use efficiency is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart home control, and specifically relates to a smart home energy control system based on energy consumption prediction. Background Art

[0002] A smart home energy control system is an intelligent system that integrates monitoring, control, and management, aiming to improve the efficiency of household energy use and achieve energy conservation and emission reduction. Through Internet of Things technology, this system connects various energy-consuming devices in the home, real-time monitors the energy consumption of electricity, gas, water, etc., and through intelligent analysis, provides users with an optimized energy usage plan. Users can remotely control devices such as lights, air conditioners, and water heaters in the home through a mobile phone APP or a voice assistant to achieve automated management. The smart home energy control system not only improves the living comfort but also helps users develop good energy-saving habits, reduces household energy consumption expenses, and promotes a green and environmentally friendly lifestyle.

[0003] However, the existing technologies usually rely on fixed rules or a single prediction model trained offline. The strategy adjustment lags behind the changes in the environment and user behavior, unable to dynamically balance the energy-saving goal and personalized needs, and lacking a flexible coordination mechanism in response to sudden grid demand response. Summary of the Invention

[0004] The purpose of the present invention is to provide a smart home energy control system based on energy consumption prediction to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: A smart home energy control system based on energy consumption prediction, the system includes: a data acquisition module, an energy consumption prediction module, an energy control strategy module, a device control module, an interactive feedback module, and an adaptive learning and optimization module;

[0006] The internal of the adaptive learning and optimization module is provided with a behavior pattern recognition sub-module, a dynamic adjustment strategy sub-module, a self-optimization sub-module, and an energy consumption knowledge graph sub-module;

[0007] The device status output end of the data acquisition module is directly connected to the historical data input end of the energy consumption prediction module, and at the same time, its environmental sensor interface is physically coupled with the dynamic data receiving end of the energy consumption knowledge graph sub-module.

[0008] The prediction result output end of the energy consumption prediction module is connected to the strategy generation algorithm input port of the energy control strategy module, and a two-way data channel is established with the dynamic adjustment strategy sub-module of the adaptive learning and optimization module.

[0009] The control instruction sending end of the energy control strategy module is interconnected with the protocol adaptation layer command entry of the device control module through hardwiring. Meanwhile, its grid response interface shares a communication bus with the alarm information receiving end of the interaction feedback module.

[0010] The execution status feedback end of the device control module transmits real-time signals to the visual data interface of the interaction feedback module and forms a closed-loop verification link with the self-optimization sub-module of the adaptive learning and optimization module.

[0011] The knowledge graph rule base of the adaptive learning and optimization module is logically bound to the optimization algorithm engine of the energy control strategy module through a semantic constraint interface, and its behavior pattern recognition result end is directly embedded in the feature engineering processing layer of the energy consumption prediction module.

[0012] The user preference collection end of the interaction feedback module establishes a one-way instruction channel with the weight adjustment unit of the adaptive learning and optimization module to form a cross-module parameter tuning path.

[0013] In a preferred embodiment, the data acquisition module is composed of an environmental parameter perception unit, a power parameter monitoring network, and a device status acquisition system. The environmental parameter perception unit is deployed in key indoor and outdoor areas and integrates high-precision temperature sensors, humidity detection probes, light intensity measuring instruments, and air quality monitoring nodes to form a full-domain environmental data coverage. The power parameter monitoring network includes an intelligent meter cluster, a multi-channel current transformer array, and a three-phase voltage collector to capture key parameters such as voltage, current, active power, and harmonic distortion rate of the home main circuit and branch circuits in real time. The device status acquisition system connects to the communication interface of smart home appliances through a protocol analysis gateway, continuously obtains the device switch status, operation mode, instantaneous power, and fault codes, and uses edge computing nodes for local data preprocessing to ensure the efficient transmission and storage of raw information.

[0014] In a preferred embodiment, the energy consumption prediction module consists of a historical data warehouse, a multi-model prediction framework, and an external data fusion center. The historical data warehouse uses time series database technology to hierarchically store energy consumption records from the minute level to the annual dimension and has built-in data cleaning tools to process missing values and abnormal fluctuations. The multi-model prediction framework integrates three core algorithms: long short-term memory network, seasonal autoregressive integrated moving average model, and gradient boosting decision tree, and supports parallel training and result cross-validation. The external data fusion center accesses real-time weather data from the meteorological bureau, time-of-use electricity price tables of power companies, and public holiday information platforms through standardized interfaces to construct a dynamic feature matrix containing temperature, precipitation probability, electricity price ladder, and special date markers to provide multi-dimensional input support for the prediction model.

[0015] In a preferred embodiment, the energy control strategy module consists of three parts: a user target configuration platform, a multi-objective optimization engine, and a power grid collaborative controller. The user target configuration platform provides a graphical operation interface, including an energy-saving level slider, a device priority sorting panel, and a comfort tolerance threshold setting bar, allowing users to customize their energy management preferences. The multi-objective optimization engine is built-in with a dynamic programming solver, a genetic algorithm library, and a linear programming toolkit, supporting the search for a global optimal solution among device start-stop constraints, power grid load limits, and user comfort requirements. The power grid collaborative controller is equipped with a demand response protocol parsing module, receives power grid dispatching instructions through an encrypted communication link, and generates a matching load adjustment plan to achieve the coordinated operation of the home energy system and the regional power grid.

[0016] In a preferred embodiment, the device control module consists of three parts: a multi-protocol adaptation layer, an instruction scheduling center, and a device status monitoring network. The multi-protocol adaptation layer is equipped with a Zigbee coordinator, a Wi-Fi communication module, a Bluetooth Mesh node, and an infrared coding library, compatible with the control protocols of mainstream smart home appliance brands. The instruction scheduling center adopts a microservices architecture, decomposes the energy strategy into a sequence of device-executable instructions, and dynamically adjusts the issuing priority based on the urgency of the task and the power grid response requirements. The device status monitoring network deploys a distributed probe cluster, real-time collects device execution feedback signals, detects offline, overload, or instruction conflict problems through an exception code recognition engine, and triggers an adaptive retry mechanism to ensure control reliability.

[0017] In a preferred embodiment, the interactive feedback module consists of a visual interaction terminal, an intelligent message center, and a user feedback integration system. The visual interaction terminal provides a cross-platform adaptable energy management interface, including a real-time energy consumption flow chart, a three-dimensional topology diagram of device operation status, and a historical energy-saving effect comparison dashboard, supporting touch zooming and multi-dimensional data drilling. The intelligent message center integrates a short message service platform, an email push gateway, and a mobile instant notification module, classifies and sends electricity bill warnings, device anomaly reminders, and energy-saving suggestion reports according to event levels. The user feedback integration system sets up a star rating control, a text opinion input box, and a voice message collector, automatically classifies user satisfaction data in combination with an emotion analysis engine, and forms a key input source for the closed-loop optimization link.

[0018] In a preferred embodiment, the interactive feedback module consists of a visual interaction terminal, an intelligent message center, and a user feedback integration system. The visual interaction terminal provides an energy management interface with cross-platform adaptability, including a real-time energy consumption flow chart, a three-dimensional topological map of device operating states, and a historical energy-saving effect comparison dashboard, supporting touch zooming and multi-dimensional data drilling. The intelligent message center integrates a short message service platform, an email push gateway, and a mobile instant notification module, and classifies and sends electricity bill warnings, device anomaly reminders, and energy-saving suggestion reports according to event levels. The user feedback integration system is provided with star rating controls, text opinion input boxes, and voice message collectors, and automatically classifies user satisfaction data in combination with an emotion analysis engine, forming a key input source for the closed-loop optimization link.

[0019] In a preferred embodiment, the dynamic adjustment strategy sub-module consists of four parts: a real-time data interface, an optimization algorithm engine, a strategy priority queue, and an elastic weight adjustment unit. The real-time data interface is responsible for receiving short-term energy consumption results from the prediction module and external dynamic information, including real-time grid electricity prices and sudden weather warnings. The optimization algorithm engine is built-in with mixed integer programming and heuristic solution tools, supporting fast strategy generation under multi-objective constraints. The strategy priority queue arranges device control instructions in time slices, and distinguishes necessary tasks and adjustable tasks in combination with user preference tags. The elastic weight adjustment unit integrates a sliding window statistical module, continuously tracks user manual intervention behaviors, and dynamically corrects the balance parameters between energy-saving goals and comfort goals to ensure that the strategy can flexibly adapt to changes in personalized needs.

[0020] In a preferred embodiment, the self-optimization sub-module includes four core components: a reinforcement learning framework, a strategy version library, an anomaly detector, and a feedback data pool. The reinforcement learning framework is built based on the deep deterministic policy gradient algorithm and is equipped with a custom reward function calculation unit. The strategy version library stores historical control schemes and their execution effect data in a tree structure, supporting version backtracking and incremental merging. The anomaly detector monitors prediction errors and energy consumption deviations in real time through residual analysis, triggering the model retraining process. The feedback data pool integrates user ratings, device fault logs, and external environment mutation records, constructs a multi-dimensional data foundation for closed-loop optimization, and provides a driving source for continuous system optimization.

[0021] In a preferred embodiment, the energy consumption knowledge graph sub-module consists of a knowledge extractor, an entity relationship database, a graph computing engine, and a semantic constraint rule base. The knowledge extractor parses device specifications and policy documents through natural language processing technology to extract structured information such as energy efficiency parameters and electricity usage specifications. The entity relationship database uses an attribute graph model to store the static attributes and dynamic associations of three types of entities: devices, users, and environments, including more than 200 predefined relationship types. The graph computing engine deploys community discovery and path analysis algorithms to identify the collaborative energy consumption patterns of device groups. The semantic constraint rule base solidifies power grid safety regulations and device physical limitations, providing insurmountable boundary conditions for policy generation to ensure that system decisions comply with industry norms and safety standards.

[0022] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present invention are as follows:

[0023] 1. In the present invention, through the adaptive learning optimization module, the dynamic precision of energy management is realized and highly adapted to user needs. The system can continuously learn the user's living habits and environmental change rules, automatically identify behavior patterns such as morning lighting on preferences and summer air conditioner temperature setting ranges, and generate device control strategies that balance economy and comfort in combination with real-time electricity price fluctuations and power grid load conditions. In the event of sudden power grid demand response or extreme weather scenarios, the system can quickly adjust the device operation priority. For example, it can automatically delay the start of the washing machine during peak electricity price periods or pre-heat the indoor temperature before the cold wave arrives, avoiding energy waste and maintaining living quality. This predictive-based active control method transforms the passive response of traditional smart homes into forward-looking optimization, significantly improving energy use efficiency.

[0024] 2. In the present invention, through the modeling of the energy efficiency characteristics of devices and the environmental coupling relationship by the knowledge graph, the system can understand the energy consumption logic chain in complex scenarios, such as identifying the hidden association that increased humidity leads to increased power consumption for air conditioner dehumidification. The self-optimization mechanism enables the control strategy to continuously iterate with user feedback, gradually eliminating initial prediction deviations and adapting to user habit migrations. During long-term operation, the system not only reduces the household electricity cost but also helps the power grid to cut peaks and fill valleys through flexible load adjustment, forming a win-win virtuous ecosystem for users, devices, and the power grid, and promoting the transformation of energy consumption from an extensive mode to a smart and intensive mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is the overall system block diagram of the present invention;

[0026] Figure 2 is the system block diagram of the adaptive learning optimization module in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0027] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0028] Embodiment:

[0029] Referring to Figure 1-2 , a smart home energy control system based on energy consumption prediction, the system includes: a data acquisition module, an energy consumption prediction module, an energy control strategy module, a device control module, an interactive feedback module and an adaptive learning and optimization module;

[0030] Inside the adaptive learning and optimization module, there are a behavior pattern recognition sub-module, a dynamic adjustment strategy sub-module, a self-optimization sub-module and an energy consumption knowledge graph sub-module;

[0031] The device status output end of the data acquisition module is directly connected to the historical data input end of the energy consumption prediction module, and at the same time its environmental sensor interface is physically coupled to the dynamic data receiving end of the energy consumption knowledge graph sub-module.

[0032] The prediction result output end of the energy consumption prediction module is connected to the strategy generation algorithm input port of the energy control strategy module, and a two-way data channel is established with the dynamic adjustment strategy sub-module of the adaptive learning and optimization module.

[0033] The control instruction issuing end of the energy control strategy module is interconnected with the protocol adaptation layer command entry of the device control module through hard wiring, and at the same time its power grid response interface shares a communication bus with the alarm information receiving end of the interactive feedback module.

[0034] The execution status feedback end of the device control module transmits real-time signals to the visual data interface of the interactive feedback module, and forms a closed-loop verification link with the self-optimization sub-module of the adaptive learning and optimization module.

[0035] The knowledge graph rule base of the adaptive learning and optimization module is logically bound to the optimization algorithm engine of the energy control strategy module through a semantic constraint interface, and the behavior pattern recognition result end is directly embedded in the feature engineering processing layer of the energy consumption prediction module.

[0036] The user preference acquisition end of the interactive feedback module and the weight adjustment unit of the adaptive learning and optimization module establish a one-way instruction channel to form a cross-module parameter tuning path.

[0037] The data acquisition module consists of three parts: the environmental parameter perception unit, the power parameter monitoring network, and the device status acquisition system. The environmental parameter perception unit is deployed in key indoor and outdoor areas, integrating high-precision temperature sensors, humidity detection probes, light intensity measuring instruments, and air quality monitoring nodes to form a full-domain environmental data coverage. The power parameter monitoring network includes an intelligent electricity meter cluster, a multi-channel current transformer array, and a three-phase voltage collector to capture key parameters such as voltage, current, active power, and harmonic distortion rate of the total household circuit and branch circuits in real time. The device status acquisition system connects to the communication interfaces of smart home appliances through a protocol analysis gateway, continuously obtains the device switch status, operation mode, instantaneous power, and fault codes, and uses edge computing nodes for local data preprocessing to ensure the efficient transmission and storage of raw information.

[0038] The energy consumption prediction module consists of a historical data warehouse, a multi-model prediction framework, and an external data fusion center. The historical data warehouse uses time-series database technology to hierarchically store energy consumption records from the minute level to the annual dimension, and has built-in data cleaning tools to process missing values and abnormal fluctuations. The multi-model prediction framework integrates three core algorithms: long short-term memory network, seasonal autoregressive integrated moving average model, and gradient boosting decision tree, and supports parallel training and result cross-validation. The external data fusion center accesses real-time weather data from the meteorological bureau, time-of-use electricity price tables from power companies, and public holiday information platforms through standardized interfaces, constructs a dynamic feature matrix containing temperature, precipitation probability, electricity price ladder, and special date markers, and provides multi-dimensional input support for the prediction model.

[0039] The energy control strategy module consists of three parts: the user target configuration platform, the multi-objective optimization engine, and the grid coordination controller. The user target configuration platform provides a graphical operation interface, including an energy-saving level slider, a device priority sorting panel, and a comfort tolerance threshold setting bar, allowing users to customize their energy management preferences. The multi-objective optimization engine has built-in dynamic programming solvers, genetic algorithm libraries, and linear programming toolkits, and supports finding the global optimal solution among device start-stop constraints, grid load limits, and user comfort requirements. The grid coordination controller is equipped with a demand response protocol analysis module, receives grid dispatch instructions through an encrypted communication link, and generates a matching load adjustment plan to achieve the coordinated operation of the home energy system and the regional grid.

[0040] The device control module consists of three parts: a multi - protocol adaptation layer, an instruction scheduling center, and a device status monitoring network. The multi - protocol adaptation layer is equipped with a Zigbee coordinator, a Wi - Fi communication module, a Bluetooth Mesh node, and an infrared coding library, which is compatible with the control protocols of mainstream smart home appliance brands. The instruction scheduling center adopts a microservices architecture, decomposes the energy strategy into a sequence of device - executable instructions, and dynamically adjusts the sending priority based on the task urgency and grid response requirements. The device status monitoring network deploys a distributed probe cluster, which collects device execution feedback signals in real - time, detects offline, overload, or instruction conflict problems through an exception code recognition engine, and triggers an adaptive retry mechanism to ensure control reliability.

[0041] The interaction feedback module consists of a visual interaction terminal, an intelligent message center, and a user feedback integration system. The visual interaction terminal provides an energy management interface with cross - platform adaptation, including a real - time energy consumption flow chart, a three - dimensional topology map of device operation status, and a historical energy - saving effect comparison dashboard, supporting touch - zoom and multi - dimensional data drilling. The intelligent message center integrates a short - message service platform, an email push gateway, and a mobile - terminal instant notification module, and classifies and sends electricity - fee warnings, device anomaly reminders, and energy - saving suggestion reports according to event levels. The user feedback integration system sets up star - rating controls, text - opinion input boxes, and voice - message collectors, and automatically classifies user satisfaction data in combination with an emotion analysis engine, forming a key input source for the closed - loop optimization link.

[0042] The behavior pattern recognition sub - module realizes accurate modeling of user living habits and device usage patterns by integrating multi - source data and deep - learning technologies. This module first obtains device status data (such as air - conditioner start - stop times, lighting - intensity change curves), environmental parameters (temperature and humidity fluctuations), and user interaction feedback (manual adjustment preferences) from the data acquisition module, uses long short - term memory networks (LSTM) and convolutional neural networks (CNN) to extract features from time - series data, and captures periodic behaviors such as "turn on the living - room air - conditioner at 7 pm every day and set it to 25°C". At the same time, the dynamic time warping (DTW) algorithm is combined to align the differences in operation sequences on different dates and eliminate noise interference. Further, through an improved K - means++ clustering algorithm, user behaviors are classified into typical scenarios (such as "weekday home mode" and "vacation energy - saving mode"), and a self - attention mechanism is introduced to assign higher weights to key operations (such as frequently adjusting the temperature). Finally, a multi - dimensional user behavior feature vector is output, including indicators such as device usage frequency, time - period preference, and energy - consumption sensitivity, providing a quantitative basis for subsequent dynamic policy adjustment;

[0043] Among them, the loss function for jointly optimizing time - series prediction and behavior clustering is defined by the formula:

[0044]

[0045] α, β, and γ are hyperparameters that respectively control the balance of prediction accuracy, clustering tightness, and model complexity (default values: 0.6, 0.3, 0.1).

[0046] ŷt and yt represent the predicted energy consumption value and the actual energy consumption value at time t, respectively.

[0047] T represents the total length of the time series.

[0048] Si represents the device operation sequence on the i-th day.

[0049] cj represents the representative behavior pattern of cluster center j.

[0050] C represents the set of cluster centers.

[0051] DTW(·) represents the dynamic time warping distance function, which is used to measure the similarity between two sequences.

[0052] W LSTM represents the weight matrix of the LSTM network, and the regularization term prevents overfitting.

[0053] The dynamic adjustment strategy sub-module consists of four parts: a real-time data interface, an optimization algorithm engine, a policy priority queue, and an elastic weight adjustment unit. The real-time data interface is responsible for receiving the short-term energy consumption results from the prediction module and external dynamic information, including the real-time electricity price of the power grid and sudden weather warnings. The optimization algorithm engine is built-in with a mixed integer programming and heuristic solving tool, supporting the rapid generation of strategies under multi-objective constraints. The policy priority queue arranges device control instructions in time slices, and combines user preference tags to distinguish between essential tasks and adjustable tasks. The elastic weight adjustment unit integrates a sliding window statistical module, continuously tracks the user's manual intervention behavior, dynamically corrects the balance parameters of energy-saving goals and comfort goals, and ensures that the strategy can flexibly adapt to changes in personalized needs.

[0054] The self-optimization sub-module includes four core components: a reinforcement learning framework, a policy version library, an anomaly detector, and a feedback data pool. The reinforcement learning framework is built based on the deep deterministic policy gradient algorithm, equipped with a custom reward function calculation unit. The policy version library stores historical control schemes and their execution effect data in a tree structure, supporting version backtracking and incremental merging. The anomaly detector monitors the prediction error and energy consumption deviation in real time through residual analysis, triggering the model retraining process. The feedback data pool integrates user ratings, device fault logs, and external environment mutation records, constructing a multi-dimensional data foundation for closed-loop optimization and providing a driving source for continuous optimization of the system.

[0055] The energy consumption knowledge graph sub-module consists of a knowledge extractor, an entity relationship database, a graph computing engine, and a semantic constraint rule base. The knowledge extractor parses device manuals and policy documents through natural language processing technology to extract structured information such as energy efficiency parameters and electricity usage specifications. The entity relationship database uses an attribute graph model to store the static attributes and dynamic associations of three types of entities: devices, users, and environments, including more than 200 predefined relationship types. The graph computing engine deploys community discovery and path analysis algorithms to identify the collaborative energy consumption patterns of device groups. The semantic constraint rule base solidifies power grid safety regulations and device physical limitations, providing insurmountable boundary conditions for policy generation to ensure that system decisions comply with industry norms and safety standards.

[0056] It can be known from the above that:

[0057] In the present invention, through the adaptive learning optimization module, the dynamic precision of energy management is realized and highly adapted to user needs. The system can continuously learn the user's living habits and environmental change rules, automatically identify behavior patterns such as morning lighting turning-on preferences and summer air conditioner temperature setting ranges, and generate device control strategies that take into account both economy and comfort in combination with real-time electricity price fluctuations and power grid load status. In the event of an emergency power grid demand response or extreme weather scenario, the system can quickly adjust the device operation priority. For example, it can automatically delay the start of the washing machine during peak electricity price periods, or pre-heat the indoor temperature before the cold wave arrives, avoiding energy waste and maintaining living quality. This predictive-based active control method transforms the passive response of traditional smart homes into forward-looking optimization, significantly improving energy use efficiency.

[0058] In the present invention, through the modeling of the energy efficiency characteristics of devices and the environmental coupling relationship by the knowledge graph, the system can understand the energy consumption logic chain in complex scenarios, such as identifying the hidden association that an increase in humidity leads to an increase in the power consumption of air conditioner dehumidification. The self-optimization mechanism enables the control strategy to continuously iterate with user feedback, gradually eliminating the initial prediction deviation and adapting to the migration of user habits. In the long-term operation, the system not only reduces the household electricity cost, but also helps the power grid to cut peaks and fill valleys through flexible load regulation, forming a win-win virtuous ecological cycle among users, devices, and the power grid, and promoting the transformation of energy consumption from an extensive mode to a smart and intensive mode.

[0059] It should be noted that, in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0060] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart home energy control system based on energy consumption prediction, characterized in that: The system includes: a data acquisition module, an energy consumption prediction module, an energy control strategy module, a device control module, an interactive feedback module, and an adaptive learning optimization module; Inside the adaptive learning optimization module, there are a behavior pattern recognition sub-module, a dynamic adjustment strategy sub-module, a self-optimization sub-module, and an energy consumption knowledge graph sub-module; The device status output terminal of the data acquisition module is directly connected to the historical data input terminal of the energy consumption prediction module, and at the same time, its environmental sensor interface is physically coupled to the dynamic data receiving terminal of the energy consumption knowledge graph sub-module; The prediction result output terminal of the energy consumption prediction module is connected to the strategy generation algorithm input port of the energy control strategy module, and a two-way data channel is established with the dynamic adjustment strategy sub-module of the adaptive learning optimization module; The control instruction issuing terminal of the energy control strategy module is interconnected with the protocol adaptation layer command entry of the device control module through hard wiring. At the same time, its grid response interface shares a communication bus with the alarm information receiving terminal of the interactive feedback module; The execution status feedback terminal of the device control module transmits real-time signals to the visual data interface of the interactive feedback module and forms a closed-loop verification link with the self-optimization sub-module of the adaptive learning optimization module; The knowledge graph rule library of the adaptive learning optimization module is logically bound to the optimization algorithm engine of the energy control strategy module through a semantic constraint interface; The user preference acquisition terminal of the interactive feedback module and the weight adjustment unit of the adaptive learning optimization module establish a one-way instruction channel to form a cross-module parameter tuning path.

2. The intelligent home energy control system based on energy consumption prediction according to claim 1, characterized in that: The data acquisition module consists of three parts: an environmental parameter perception unit, a power parameter monitoring network, and a device status acquisition system; the environmental parameter perception unit is deployed in key indoor and outdoor areas, integrating high-precision temperature sensors, humidity detection probes, light intensity measuring instruments, and air quality monitoring nodes; the power parameter monitoring network includes an intelligent meter cluster, a multi-channel current transformer array, and a three-phase voltage collector; the device status acquisition system is connected to the communication interface of intelligent appliances through a protocol analysis gateway and continuously obtains the device switch status, operating mode, instantaneous power, and fault codes.

3. The smart home energy control system based on energy consumption prediction according to claim 1, characterized in that: The energy consumption prediction module consists of a historical data warehouse, a multi-model prediction framework, and an external data fusion center; the historical data warehouse uses time-series database technology to hierarchically store energy consumption records from the minute level to the annual dimension, and built-in data cleaning tools are used to process missing values and abnormal fluctuations.

4. The smart home energy control system based on energy consumption prediction according to claim 1, characterized in that: The energy control strategy module consists of three parts: a user target configuration platform, a multi-objective optimization engine, and a grid collaborative controller; the user target configuration platform provides a graphical operation interface, including an energy-saving level slider, a device priority sorting panel, and a comfort tolerance threshold setting bar, allowing users to customize their energy management preferences; the multi-objective optimization engine is built-in with a dynamic programming solver, a genetic algorithm library, and a linear programming toolkit, supporting finding the global optimal solution among device start-stop constraints, grid load limits, and user comfort requirements; The grid collaborative controller is equipped with a demand response protocol analysis module.

5. The smart home energy control system based on energy consumption prediction according to claim 1, wherein: The device control module consists of three parts: a multi-protocol adaptation layer, an instruction scheduling center, and a device status monitoring network; the multi-protocol adaptation layer is equipped with a Zigbee coordinator, a Wi-Fi communication module, a Bluetooth Mesh node, and an infrared coding library.

6. The smart home energy control system based on energy consumption prediction according to claim 1, wherein: The interactive feedback module consists of a visual interaction terminal, an intelligent message center, and a user feedback integration system; the visual interaction terminal provides an energy management interface with cross-platform adaptation, including a real-time energy consumption flow chart, a three-dimensional topology map of device operation status, and a historical energy-saving effect comparison dashboard, and supports touch zooming and multi-dimensional data drilling.

7. The smart home energy control system based on energy consumption prediction according to claim 1, characterized in that: The loss function for jointly optimizing the timing prediction and behavior clustering of the behavior pattern recognition sub-module is defined by the formula: α, β, γ are hyperparameters that respectively control the balance of prediction accuracy, clustering tightness, and model complexity (default values: 0.6, 0.3, 0.1) ŷ^t, yt represent the predicted energy consumption value and the actual energy consumption value at time t T represents the total length of the time series Si represents the device operation sequence on the i-th day cj represents the representative behavior pattern of cluster center j C represents the set of cluster centers DTW(·) represents the dynamic time warping distance function, which is used to measure the similarity W between two sequences LSTM represents the weight matrix of the LSTM network, and the regularization term prevents overfitting.

8. The smart home energy control system based on energy consumption prediction according to claim 1, wherein: The dynamic adjustment strategy sub-module consists of four parts: a real-time data interface, an optimization algorithm engine, a policy priority queue, and an elastic weight adjustment unit; the real-time data interface is responsible for receiving short-term energy consumption results and external dynamic information from the prediction module, and the optimization algorithm engine has built-in mixed integer programming and heuristic solving tools; the policy priority queue arranges device control instructions in time slices and distinguishes necessary tasks and adjustable tasks by combining user preference tags; the elastic weight adjustment unit integrates a sliding window statistics module to continuously track user manual intervention behaviors.

9. The smart home energy control system based on energy consumption prediction according to claim 1, wherein: The self-optimization sub-module includes four core components: a reinforcement learning framework, a policy version library, an anomaly detector, and a feedback data pool; the reinforcement learning framework is built based on the deep deterministic policy gradient algorithm and is equipped with a custom reward function calculation unit; the policy version library stores historical control schemes and their execution effect data in a tree structure and supports version backtracking and incremental merging; the anomaly detector monitors prediction errors and energy consumption deviations in real time through residual analysis and triggers the model re-training process; the feedback data pool integrates user ratings, device fault logs, and external environment mutation records to build a multi-dimensional data foundation for closed-loop optimization and provide a driving source for the continuous evolution of the system.

10. The smart home energy control system based on energy consumption prediction according to claim 1, wherein: The energy consumption knowledge graph sub-module consists of a knowledge extractor, an entity relationship database, a graph computing engine, and a semantic constraint rule library.

Images