A method for energy consumption optimization and fault early warning of heterogeneous home appliances

Abstract

This invention discloses a method for energy consumption optimization and fault early warning of heterogeneous home appliances, specifically relating to the field of home appliance detection technology. The method includes connecting heterogeneous home appliances to a home network and collecting multi-source heterogeneous data from these appliances in real time; processing the collected multi-source heterogeneous data to form a time-series dataset; receiving the time-series dataset and performing parallel intelligent analysis; constructing and running an energy consumption optimization model to analyze the energy consumption data of the home appliances; constructing and running a fault early warning model to analyze the operating status data of the home appliances; and executing response operations based on the results of the parallel intelligent analysis to send alarm notifications to the user terminal. This invention solves problems such as insufficient data processing capabilities, low analysis efficiency, poor adaptability, and limited early warning accuracy by collecting multi-source heterogeneous data, processing the multi-source heterogeneous data, performing parallel intelligent analysis, and executing response operations based on the results of the parallel intelligent analysis.

Description

Technical Field

[0001] This invention relates to the field of home appliance testing technology, and more specifically, to a method for energy consumption optimization and fault early warning of heterogeneous home appliances. Background Technology

[0002] With the rapid development of IoT technology and the smart home industry, home appliances in the home environment are exhibiting heterogeneous characteristics, encompassing various types such as air conditioners, refrigerators, and washing machines. These devices also differ significantly in brand, communication protocols (such as WiFi, Bluetooth, Modbus, and MQTT), and data formats. While the widespread adoption of these heterogeneous appliances greatly enriches users' lives, it also brings prominent problems such as chaotic energy management and delayed fault warnings. Therefore, effectively managing these devices and achieving energy optimization and fault warnings has become an important research direction in the smart home field.

[0003] Existing methods for energy consumption optimization and fault early warning of heterogeneous home appliances can meet basic usage needs, but they still have some shortcomings in practical use: First, they lack data processing capabilities and a unified mechanism for cleaning, aligning, and standardizing multi-source heterogeneous data, resulting in low data quality and inconsistent model inputs; second, they have low analysis efficiency, mostly using serial processing methods, which cannot achieve parallel real-time analysis of energy consumption optimization and fault early warning; third, they have poor adaptability, making it difficult to support device access with multiple communication protocols, and they do not fully consider multiple objective constraints such as user comfort and device operating efficiency; fourth, their early warning accuracy is limited, fault detection methods are simple, and they fail to combine time series characteristics and probability models for risk quantification, resulting in a high false alarm rate.

[0004] Therefore, there is an urgent need for an energy consumption optimization and fault early warning method that can effectively process data from heterogeneous devices, support parallel intelligent analysis, and take into account multiple objective constraints, so as to improve the management efficiency and operational reliability of home appliances. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for energy consumption optimization and fault early warning of heterogeneous home appliances, which solves the problems mentioned in the background art through the following scheme.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for energy consumption optimization and fault early warning of heterogeneous home appliances, comprising:

[0007] S1: Access heterogeneous home appliances in the home network and collect multi-source heterogeneous data of the home appliances in real time. The multi-source heterogeneous data includes operating status data, energy consumption data and related environmental data.

[0008] S2: Clean, align, and standardize the collected multi-source heterogeneous data to form a time series dataset in a unified format;

[0009] S3: Receives time-series datasets and performs parallel intelligent analysis.

[0010] S31: Build and run an energy consumption optimization model, analyze the energy consumption data of home appliances, and generate energy-saving control strategies for target home appliances based on the optimization objectives of the energy consumption data. The optimization objectives include peak and off-peak electricity prices, user comfort constraints, and equipment operating efficiency.

[0011] S32: Build and run a fault early warning model, analyze the operating status data of home appliances, identify abnormal data that deviates from the normal state, and calculate the probability of fault risk of home appliances.

[0012] S4: Execute response operations based on the parallel intelligent analysis results of S3. The response operations include sending the energy-saving control strategy to the corresponding home appliances for execution, and sending an alarm notification to the user terminal when the fault risk probability exceeds a preset threshold.

[0013] The technical effects and advantages of this invention are as follows:

[0014] 1. This invention improves data processing capabilities and data quality by constructing a complete data processing workflow that includes cleaning, alignment, and standardization, thereby achieving dynamic filling, dual anomaly detection, timestamp alignment, and standardization of multi-source data such as operating status, energy consumption, and environment.

[0015] 2. This invention designs a parallel intelligent analysis architecture based on the Spark Streaming distributed streaming computing framework. Through independent data scheduling nodes, energy consumption optimization analysis nodes, and fault early warning analysis nodes, it realizes synchronous sharding and parallel processing of the same time series dataset, and achieves parallel real-time analysis of energy consumption optimization and fault early warning.

[0016] 3. This invention constructs a deep learning model that integrates LSTM temporal feature extraction and Bayesian probability output. It calculates anomaly scores through Mahalanobis distance and maps them to fault risk probabilities based on historical distribution, thereby improving the accuracy and interpretability of fault warning.

[0017] 4. To address the issues of diverse device access protocols and singular optimization objectives, this invention constructs a unified access adaptation framework in S1 that supports multiple communication protocols, and incorporates a device type dictionary and security authentication mechanism, thereby enabling plug-and-play and secure access for heterogeneous devices and enhancing the adaptability of home appliances to multi-objective constraints. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0019] Figure 2 This is a schematic diagram of the device access and data acquisition structure of the present invention.

[0020] Figure 3 This is a schematic diagram of the data cleaning, alignment, and standardization structure of the present invention.

[0021] Figure 4 This is a schematic diagram of the parallel intelligent analysis structure of the present invention.

[0022] Figure 5 This is an optimized diagram of the experimental data on the operating energy consumption of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see Figures 1-5 As shown, this embodiment of the invention provides a method for energy consumption optimization and fault early warning of heterogeneous home appliances, including:

[0025] S1: Access heterogeneous home appliances in the home network and collect multi-source heterogeneous data of the home appliances in real time. The multi-source heterogeneous data includes operating status data, energy consumption data and related environmental data.

[0026] S2: Clean, align, and standardize the collected multi-source heterogeneous data to form a time series dataset in a unified format;

[0027] S3: Receives time-series datasets and performs parallel intelligent analysis.

[0028] S31: Build and run an energy consumption optimization model, analyze the energy consumption data of home appliances, and generate energy-saving control strategies for target home appliances based on the optimization objectives of the energy consumption data. The optimization objectives include peak and off-peak electricity prices, user comfort constraints, and equipment operating efficiency.

[0029] S32: Build and run a fault early warning model, analyze the operating status data of home appliances, identify abnormal data that deviates from the normal state, and calculate the probability of fault risk of home appliances.

[0030] S4: Execute response operations based on the parallel intelligent analysis results of S3. The response operations include sending the energy-saving control strategy to the corresponding home appliances for execution, and sending an alarm notification to the user terminal when the fault risk probability exceeds a preset threshold.

[0031] In S1, the heterogeneous home appliances connected to the home network are collected in real time, and the multi-source heterogeneous data of the home appliances are collected in real time. The multi-source heterogeneous data includes operating status data, energy consumption data and related environmental data.

[0032] S101: Constructing a framework for connecting and adapting heterogeneous home appliances

[0033] Deploy a home edge gateway device. The edge gateway integrates a multi-protocol adapter module, supports commonly used home communication protocols such as WiFi, Bluetooth, Modbus, and MQTT, and has a built-in protocol conversion engine for unified parsing of heterogeneous protocols.

[0034] A type dictionary for home appliances is pre-configured in the edge gateway. The home appliances include air conditioners, refrigerators, washing machines, water heaters, rice cookers, and air purifiers. The device type dictionary includes brand identifiers, model codes, communication protocol types, data transmission formats, and default access parameters.

[0035] A device access security authentication module is set up with a preset two-way authentication mechanism. The edge gateway stores the device ID of the home appliance, the IP address of the home appliance, and the authentication public key of the gateway for subsequent identity verification.

[0036] S102: Access to heterogeneous home appliances

[0037] Enable the device scanning function of the edge gateway and perform periodic scans according to the following rules. The scan cycle can be configured to be 10 seconds per scan:

[0038] For WiFi protocol devices, a device request is sent via UDP broadcast, and a device description file returned by the device is received. The description file contains the home appliance ID, model, IP address, and supported parameter types.

[0039] For Bluetooth devices, enable Bluetooth scanning mode to obtain the MAC address and broadcast packet information of nearby Bluetooth devices, match the Bluetooth device feature code in the device type dictionary, and filter target home appliances.

[0040] For wired Modbus devices, the system iterates through the connections according to the pre-configured COM serial port parameters, sends device query commands, and obtains device response information.

[0041] After receiving the identity information uploaded by the home appliance, the edge gateway initiates the two-way authentication process:

[0042] The gateway sends an authentication request to the home appliance. The home appliance encrypts a random number using a pre-stored key and returns it, along with its device certificate.

[0043] The gateway verifies the validity of the device certificate and decrypts the random number using a key. If the decryption result matches the sent value, the authentication is successful; otherwise, access to the home appliance data is rejected.

[0044] After successful authentication, the edge gateway assigns a unique device access identifier to the device. The access identifier consists of a home ID, a device type code, and a home appliance ID. Based on the device type dictionary, it loads multi-source heterogeneous data of the home appliance to complete the device access.

[0045] For home appliances that are not actively scanned, a passive registration interface is provided. Users can input the device model, communication protocol and connection parameters through a mobile APP. The APP will synchronize the information to the edge gateway, and the gateway will initiate a targeted connection according to the parameters configured by the user.

[0046] S103: Real-time acquisition of multi-source heterogeneous data

[0047] Operating status data collection items, air conditioner: collects operating mode, which includes cooling, heating, and ventilation; user-set temperature, actual temperature, fan speed level, running time, compressor running or stopping, and fault codes;

[0048] Refrigerator: Refrigeration temperature, freezing temperature, operating mode (energy saving or quick cooling), door open / closed status, compressor operating frequency;

[0049] Washing machine: Remaining washing time, water level, spin speed, water inlet and outlet status;

[0050] Energy consumption data collection items: real-time power consumption of home appliances Cumulative electricity consumption ,Voltage Current Power factor ;

[0051] The associated environmental data collection items involve collecting environmental data around the home appliance through built-in sensors and external environmental sensors. This environmental data includes temperature. ,humidity PM2.5 concentration ;

[0052] Users can customize data collection items through the app, and can add specific data collection items, such as the temperature of the air conditioner vent. Modify the sampling frequency or disable unnecessary sampling items, such as the washing machine's voltage. Current data .

[0053] S104: Data Acquisition Status Monitoring and Anomaly Handling

[0054] The gateway monitors the data acquisition process in real time. If the following abnormal situations occur, the corresponding handling mechanism will be triggered:

[0055] If data acquisition times out and no response is received from the device within 5 seconds after the request is sent, the acquisition will be automatically retried twice. If the acquisition still fails, it will be marked as acquisition timeout and a device communication error reminder will be sent to the user.

[0056] The data format is abnormal. The data type returned by the home appliance does not match. It is marked as abnormal data, and the data collection is skipped. At the same time, the protocol is triggered to recalibrate the data format of the home appliance.

[0057] If the device fails to respond to the heartbeat packet three times in a row, the gateway will automatically attempt to reconnect. If the reconnection fails, the device will be marked as offline, and an offline reminder will be displayed on the user's app.

[0058] In S2, the collected multi-source heterogeneous data is cleaned, aligned, and standardized to form a time series dataset in a unified format.

[0059] S201: Multi-source heterogeneous data reception and structured storage

[0060] It receives multi-source heterogeneous data transmitted from S103, receives data streams from cloud servers, and receives heterogeneous home appliance data transmitted from gateways in real time. The receiving protocol adopts MQTT-SN to ensure that the data transmission delay is ≤50ms.

[0061] The structured storage design stores five core fields in the database: appliance ID, data collection timestamp, data collection item, data value, and data status flag; data partitioning management is implemented, with partitions created according to appliance type and independent indexes set for each partition.

[0062] S202: Data Cleaning Processing

[0063] S2021: Missing value identification and imputation. Missing value identification involves traversing the stored data in the database. If a certain collection item has no data recorded within a continuous time window and the home appliance has no offline marker, it is determined to be a missing value.

[0064] The dynamic filling strategy uses linear interpolation to fill missing data if the missing duration is less than twice the collection period. The interpolation method calculates the intermediate value based on the two most recent valid data points before and after the missing data.

[0065] If the missing duration is more than twice the collection period, the device type mean filling method is used, which is based on the historical mean of the household appliances in the same time period over the past 24 hours; for state-type discrete data filling, the forward filling method is used, which uses the most recent valid state value before the missing data.

[0066] S2022: A dual anomaly detection mechanism is adopted for outlier identification and processing. The first step is based on... Threshold determination is performed based on the principle of calculating the average of valid data for each collection item over the past 7 days. and standard deviation If the current data value exceeds If the range is specified, it is marked as a candidate outlier;

[0067] The second step is to verify the equipment operation rules, filtering out false judgments by combining equipment type and operating status. For example, when the air conditioner is off, a power data of 0 to 5W is normal, while a value exceeding this range is considered incorrect. Mark the range as abnormal;

[0068] S203: Data Alignment Processing

[0069] The alignment granularity is preset to a uniform timestamp granularity of 1 minute, and the alignment reference time is every minute on the hour. For timestamp calibration, for raw data whose collected timestamps deviate from the reference time, such as 00:01:05, the nearest alignment principle is adopted. If the deviation time is ≤30 seconds, it is aligned to the current reference time 00:01:00; if the deviation time is greater than 30 seconds, it is aligned to the next reference time 00:02:00.

[0070] For timestamp errors caused by device clock deviation, such as a difference of more than 1 minute between the device time and the server time, calibration is performed based on the server timestamp, and the calibrated time is marked as a time calibration identifier.

[0071] Multi-source heterogeneous data association and alignment: Home appliances and data collection items are aligned to ensure that the same home appliance ID contains all data collection items of the home appliance under the same base timestamp. For example, an air conditioner under the timestamp of 00:01:00 contains data such as actual temperature, power, and working mode.

[0072] Aligning multi-source heterogeneous data with environmental data, the household environmental data is synchronized to the time series of home appliances according to the base timestamp, forming a related dataset of device data and environmental data.

[0073] S204: Data Standardization

[0074] Standardization aims to eliminate differences in dimensions and scope among multi-source heterogeneous data and to unify data structures;

[0075] Numerical normalization and standardization: Normalization scales multi-source heterogeneous data to a fixed interval [0,1], which is suitable for parameters with well-defined boundaries;

[0076] Standardization transforms heterogeneous data from multiple sources into a normal distribution with a mean of 0 and a standard deviation of 1. The formula is... It is suitable for situations where the distribution is unknown or where it is necessary to emphasize the characteristics of the data distribution; classification data encoding converts non-numerical data such as the operating status and mode of home appliances into numerical labels, such as 0, 1, 2 or one-hot encoded vectors; unified data structure encapsulation encapsulates all processed data into records in a unified format, with each record being a data point;

[0077] S205: Construction and Storage of Time Series Datasets

[0078] Windowing processing: To adapt to the subsequent LSTM time series analysis model, the continuous time series stream is cut into fixed-length sliding window samples. Each sample contains data from the past hour, one point per minute, for a total of 60 time steps; persistent storage: The processed and structured time series data is written into a database specifically optimized for time series.

[0079] The dataset is published, forming a high-quality time series dataset that can be directly read by the energy consumption optimization model and fault early warning model in step S3.

[0080] In S3, the received time-series dataset is subjected to parallel intelligent analysis:

[0081] Further explanation is needed regarding parallel intelligent analysis, which is a parallel analysis platform built on the distributed Spark Streaming computing framework. The platform includes two independent analysis task nodes: an energy consumption optimization node and a fault early warning node, and one data scheduling node.

[0082] The data scheduling strategy receives the unified format time series dataset output by S2, divides the data into pieces according to the home appliance ID and time window, and sends the pieced data to two analysis task nodes to achieve parallel execution of S31 and S32.

[0083] Computing resources are dynamically allocated, with CPU allocated according to task priority. Energy optimization nodes occupy 40%-50% of CPU computing resources, fault early warning nodes occupy 30%-40% of CPU computing resources, and the remaining resources are reserved for data scheduling and emergency processing.

[0084] The parallel synchronization mechanism uses ZooKeeper distributed locks to synchronize the states of two nodes, ensuring that the analysis results of the same home appliance are consistent in the time dimension.

[0085] S31: Build and run an energy consumption optimization model, analyze the energy consumption data of home appliances, and generate energy-saving control strategies for target home appliances based on the optimization objectives of the energy consumption data. The optimization objectives include peak and off-peak electricity prices, user comfort constraints, and equipment operating efficiency.

[0086] S311: Problem Modeling and Input Definition

[0087] The optimization cycle is usually based on a 24-hour day, divided into T time periods, such as T=96, where each time period is 15 minutes.

[0088] Decision variables: Define the operating state of each controllable home appliance, such as an air conditioner, refrigerator, or scheduleable washing machine, at each time period T. This can be a continuous variable, such as the set temperature of the air conditioner. Target speed of washing machine Environmental parameters, such as indoor temperature ,humidity ;

[0089] Input data for time-of-use electricity prices for the next 24 hours. (Peak / Side / Valley); User constraints, comfort range, i.e., the allowable range of indoor temperature. ;

[0090] Device usage time window, such as electric vehicles needing to be charged to the target battery level before 8:00 AM the following day. Equipment parameters, rated power Energy efficiency ratio Thermal inertia parameters, minimum operating / downtime, etc.;

[0091] Environmental and Load Forecasting: Outdoor Temperature Forecast for the Next 24 Hours Household basic load forecast ;

[0092] S312: Build and run the energy consumption optimization model

[0093] The objective function and constraints are constructed to minimize the total electricity cost or perform peak shaving and valley filling while satisfying user comfort and equipment physical limitations. Minimizing the total electricity cost is the optimization objective for energy consumption data, and the specific calculation formula is as follows: ,in Represents a set of dispatchable home appliances. Indicates home appliances During the period Power consumption Indicates the length of the time period;

[0094] Key constraints, user comfort constraints, and the calculation formula for the dynamic indoor temperature model using air conditioning as an example are as follows: ,in express Indoor temperature during the period The parameters represent the thermal resistance and heat capacity of the building envelope; the comfort boundary constraints are... ;

[0095] Equipment operating efficiency and physical constraints, based on the upper and lower limits of household appliance power. The specific formula for calculating the minimum continuous operating time of home appliances is as follows:

[0096]

[0097] Indicates home appliances If it starts at any time, then it will run at least once in the future. Each time period;

[0098] Task completion constraints, taking electric vehicles as an example, are calculated using the following formula:

[0099] in Indicates the time window during which electric vehicles are allowed to charge. Indicates the charging efficiency of electric vehicles, Indicates the initial charge level of the electric vehicle;

[0100] The table below shows the experimental data of S312 running the energy consumption optimization model, including the operating energy consumption of different home appliances at different times, and the optimized operating energy consumption data.

[0101]

[0102] S313: Generation of Energy-Saving Control Strategies

[0103] The solution yields the optimal power plan for each controllable home appliance for each time period over the next 24 hours, and converts the energy consumption data into specific, executable energy-saving control commands, such as:

[0104] Air conditioner: Set temperature from 24℃ to 25℃, fan speed from level 3 to level 2, compressor operating frequency from 50Hz to 45Hz;

[0105] Washing machine: Start during off-peak hours (23:00-7:00), reduce spin speed from 1200rpm to 1000rpm, and keep the washing time unchanged;

[0106] Water heater: During peak hours (8:00-12:00, 17:00-21:00), the heat preservation temperature drops from 60℃ to 55℃, and during off-peak hours, it heats up to 60℃.

[0107] S32: Build and run a fault early warning model, analyze the operating status data of home appliances, identify abnormal data that deviates from the normal state, and calculate the probability of fault risk of home appliances.

[0108] S321: Construction of Operational Fault Early Warning Model

[0109] Dataset selection: Historical fault-free operation status data was used as the training set.

[0110] Model structure: A deep learning model consisting of an LSTM layer, a fully connected layer, and a Bayesian probability output layer is constructed to identify abnormal data and quantify fault risk.

[0111] Input layer: 12-dimensional feature vectors, with a time step of 10, meaning that the input consists of feature data from 10 consecutive time points;

[0112] LSTM layer: 2-layer bidirectional LSTM with 64 and 32 hidden units respectively, to capture the temporal dependencies of time series data;

[0113] Fully connected layer: 2 fully connected layers, with 32 and 16 neurons respectively, and ReLU activation function;

[0114] Output layer: Bayesian probability layer, outputs the probability of failure risk. , range ;

[0115] S322: Real-time anomaly detection and risk probability calculation

[0116] Real-time feature extraction: Using a sliding window approach, the same feature vectors as those extracted during the training phase are extracted from the time-series dataset output by S2. Calculating outlier scores based on Mahalanobis distance The specific calculation formula is as follows:

[0117]

[0118] Calculating outlier scores based on Mahalanobis distance Calculate the probability of failure risk abnormal scores Mapped to Intervals, using anomaly scores from historical normal operating data. Perform distributed calibration; calculate the current The survival function value under this distribution serves as the risk probability. ,in It is the cumulative distribution function of the historical operational status anomaly scores;

[0119] Preset risk probability thresholds based on historical anomaly detection data ,when When the alarm is triggered, the alarm information includes the appliance ID, risk probability value, abnormal characteristics, and fault type.

[0120] In S4, the parallel intelligent analysis results based on S3 are used to perform response operations. The response operations include sending the energy-saving control strategy to the corresponding home appliances for execution, and sending an alarm notification to the user terminal when the fault risk probability exceeds a preset threshold.

[0121] Branch 1: Issuance and Implementation of Energy-Saving Control Strategies:

[0122] The strategy receiving and parsing process receives a set of energy-saving control strategies generated by the S31 energy consumption optimization model. This set of strategies typically includes a timestamp, appliance ID, and target instruction sequence.

[0123] Safety and status verification: Before issuing commands, the status of target home appliances is queried in real time, and safety rule checks are performed to avoid issuing start commands to devices that are already in a faulty state, or to check whether the user is in home mode to avoid turning off safety-related devices when no one is home.

[0124] Command issuance and confirmation: Control commands are sent to target home appliances in real time through the home network. A command confirmation mechanism is established, and the device is waited for a response signal indicating successful execution. The data acquisition channel of S1 is used to verify whether the device status has changed to the target status within the expected time.

[0125] The system performs monitoring and adaptive adjustments, continuously monitoring data such as actual energy consumption and temperature after the strategy is executed. If a fault risk signal for the device is received from S32, the system triggers a strategy replanning, suspends or adjusts the current strategy, and notifies the user.

[0126] Branch 2: Generation and notification of fault warnings:

[0127] Risk assessment involves receiving real-time equipment failure risk probability values ​​and associated anomaly characteristics from the S32 fault early warning model; and then comparing the risk probability with a preset threshold. Compare;

[0128] If the warning threshold If it is, it is judged as high risk, and an alarm will be sent.

[0129] If the warning threshold If the risk is low, the status will only be displayed in the application interface and no push notifications will be sent.

[0130] Integrate the generated alarm information to generate alarm notification content, including: core information, appliance ID, high fault risk detected, and risk probability value; suggested measures, such as contacting maintenance or the equipment has automatically switched to energy-saving mode.

[0131] Multi-channel notification delivery: By integrating message push services, formatted alarm information is sent to the user terminal through multiple channels: mobile APP push, sending high-priority mobile phone pop-up notifications; home hub voice / screen prompts, local voice broadcasting or display through smart speakers or central control screens;

[0132] Alarm status management: Create and maintain an alarm list in the system backend, provide a user feedback interface, and allow users to confirm and mark processed alarms on the APP;

[0133] Set alarm clearing conditions. When the risk probability remains below the threshold for a period of time, the system will automatically change the alarm status from active to recovered and record the log.

[0134] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0135] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for energy consumption optimization and fault early warning of heterogeneous home appliances, characterized in that, include: S1: Access heterogeneous home appliances in the home network and collect multi-source heterogeneous data of the home appliances in real time. The multi-source heterogeneous data includes operating status data, energy consumption data and related environmental data. S1 includes: deploying a home edge gateway to build a heterogeneous home appliance access adaptation framework, which supports the conversion and unified parsing of multiple home communication protocols and pre-configures a device type dictionary and a security authentication module; performing device scanning or receiving user registration through the edge gateway to complete the access and two-way security authentication of heterogeneous home appliances, and assigning a unique access identifier to the authenticated devices; and collecting real-time operating status data, energy consumption data, and associated environmental data of each accessed home appliance based on the device type dictionary. S2: Clean, align, and standardize the collected multi-source heterogeneous data to form a time series dataset in a unified format; S3: Receives time-series datasets and performs parallel intelligent analysis. The parallel intelligent analysis in S3 includes: based on a distributed streaming computing framework, setting up independent data scheduling nodes, energy consumption optimization analysis nodes, and fault early warning analysis nodes; after receiving the time series dataset, the data scheduling node performs data sharding according to device and time window, and simultaneously distributes it to the energy consumption optimization analysis node and the fault early warning analysis node, and dynamically allocates calculations according to task priority; S31: Build and run an energy consumption optimization model, analyze the energy consumption data of home appliances, and generate energy-saving control strategies for target home appliances based on the optimization objectives of the energy consumption data. The optimization objectives include peak and off-peak electricity prices, user comfort constraints, and equipment operating efficiency. The S31 method for constructing and running the energy consumption optimization model includes: defining a preset period as the optimization time period, using the operating status of schedulable home appliances in each time period as decision variables, and inputting time-of-use electricity prices, user comfort constraints, equipment parameters, and environmental prediction data; constructing an objective function with the goal of minimizing total electricity costs, the objective function including a set of constraints including dynamic constraints on user comfort, physical operation constraints on equipment, and task completion constraints; solving the objective function and the set of constraints to obtain the optimal power plan for each home appliance within the period, and converting it into executable energy-saving control instructions; The formula for calculating the objective function is: in Represents a set of schedulable home appliances. Indicates home appliances During the period Power consumption Indicates the length of the time period; The formula for calculating user comfort constraints is: in express Indoor temperature during the period The parameters represent the thermal resistance and heat capacity of the building envelope; the comfort boundary constraints are... ; Equipment operating efficiency and physical constraints, based on the upper and lower limits of household appliance power. The specific formula for calculating the minimum continuous operating time of home appliances is as follows: Indicates home appliances If it starts at any time, then it will run at least once in the future. Each time period; The formula for calculating task completion constraints is: in Indicates the time window during which electric vehicles are allowed to charge. Indicates the charging efficiency of electric vehicles, Indicates the initial charge level of the electric vehicle; S32: Build and run a fault early warning model, analyze the operating status data of home appliances, identify abnormal data that deviates from the normal state, and calculate the probability of fault risk of home appliances. S4: Execute response operations based on the parallel intelligent analysis results of S3. The response operations include sending the energy-saving control strategy to the corresponding home appliances for execution, and sending an alarm notification to the user terminal when the fault risk probability exceeds a preset threshold. The energy consumption optimization model and the fault early warning model run in parallel based on the same time series dataset, and the energy consumption optimization analysis node and the fault early warning analysis node achieve state synchronization through a distributed coordination framework; when the fault risk probability output by the fault early warning model exceeds a preset threshold, the energy consumption optimization model is triggered to replan the current energy-saving control strategy.

2. The method for energy consumption optimization and fault early warning of heterogeneous home appliances according to claim 1, characterized in that, S1 further includes: Monitor the data acquisition process and handle and alert for abnormal situations such as acquisition timeouts, abnormal data formats, or device offline.

3. The method for energy consumption optimization and fault early warning of heterogeneous home appliances according to claim 1, characterized in that, S2 includes: The system receives and structures multi-source heterogeneous data from the gateway in real time, and cleans the multi-source heterogeneous data. The cleaning includes dynamic filling of missing values ​​and dual judgment and processing of outliers. The system performs timestamp alignment and association alignment on multi-source heterogeneous data and associated environmental data at a preset time granularity; the aligned data is then standardized, including numerical normalization / standardization and classification data encoding; the standardized data is then constructed into a time series dataset with a fixed time window and stored.

4. The method for energy consumption optimization and fault early warning of heterogeneous home appliances according to claim 1, characterized in that, The fault early warning model built and run by S32 includes: Based on historical fault-free operation data, a deep learning model containing an LSTM network and a Bayesian probability output layer is constructed; feature vectors matching the model input are extracted in real time, anomaly scores are calculated based on Mahalanobis distance, and the anomaly scores are mapped to fault risk probabilities according to the historical anomaly score distribution.

5. The method for energy consumption optimization and fault early warning of heterogeneous home appliances according to claim 1, characterized in that, In step S4, sending the energy-saving control strategy to the corresponding home appliances for execution includes: Before issuing commands, verify the status and safety rules of the target home appliances; issue control commands through the home network and establish a confirmation mechanism to verify the execution effect; continuously monitor the actual data after execution, and trigger policy replanning when a fault risk signal from an associated device is received.

6. The method for energy consumption optimization and fault early warning of heterogeneous home appliances according to claim 1, characterized in that, The step S4, which involves sending an alarm notification to the user terminal when the probability of the fault risk exceeds a preset threshold, includes: The system determines the warning level based on the comparison between the risk probability value and the threshold; generates alarm information including equipment information, risk probability, and suggested measures; sends alarm notifications through multiple channels such as mobile APP push, home central voice / screen prompts; and manages alarm status in the system background, supporting user feedback and automatic clearing.

Images