AI-based home security hazard intelligent detection and early warning system

Abstract

The application discloses an AI-based home safety hidden danger intelligent detection and early warning system, relates to the technical field of home hidden danger detection, and comprises an early warning center, a data acquisition module, a data processing module, an AI analysis module and an execution module. The operation data of various home devices in a home environment are acquired through the data acquisition module, the operation data of each home device is subjected to data cleaning, numerical normalization processing and feature interval screening through the data processing module, and then the principal component feature data corresponding to each home device is obtained, the AI analysis module is used for constructing an AI analysis model for hidden danger analysis of the home device, the principal component feature data of each home device is input into the AI analysis model, the hidden danger items and the hidden danger coefficient of the home device are obtained, the hidden danger level of the home device is determined according to the hidden danger coefficient, and finally, the execution module adopts corresponding execution operations according to the hidden danger level of the home device.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of home hidden danger detection, and in particular to an AI-based home safety hidden danger intelligent detection and early warning system. BACKGROUND

[0002] With the popularization of smart homes, the types and quantities of home devices are increasing, and real-time monitoring and early warning of their operation safety hidden dangers have become an important issue. Existing home safety systems mostly rely on manual inspection or single sensor threshold alarm, and have the following shortcomings:

[0003] 1. Low data processing efficiency: traditional methods are difficult to cope with the cleaning, normalization and feature extraction of massive real-time data, resulting in delayed analysis;

[0004] 2. Insufficient hidden danger identification accuracy: relying on static rules or simple models, it is difficult to dynamically capture abnormal trends of device voltage, power and temperature;

[0005] 3. Single response mechanism: lacking of graded early warning and intelligent execution strategy, high-risk hidden dangers cannot be quickly blocked, and low-risk hidden dangers are easily ignored.

[0006] In view of the above problems, an AI-based home safety hidden danger intelligent detection and early warning system is needed to improve the real-time performance, accuracy and execution efficiency of home hidden danger detection. SUMMARY

[0007] To solve the above problems, the purpose of the present application is to provide an AI-based home safety hidden danger intelligent detection and early warning system.

[0008] The purpose of the present application can be achieved by the following technical solution: an AI-based home safety hidden danger intelligent detection and early warning system, comprising a warning center, which is communicatively connected with a data acquisition module, a data processing module, an AI analysis module and an execution module;

[0009] The data acquisition module is used to acquire the operation data of each home device in the home environment;

[0010] The data processing module is used to perform data cleaning, numerical normalization processing and feature interval screening on the operation data of each home device, and then obtain the principal component feature data corresponding to each home device;

[0011] The AI analysis module constructs an AI analysis model for hidden danger analysis of home devices, inputs the principal component feature data of each home device into the AI analysis model, obtains the hidden danger items and hidden danger coefficients corresponding to the home devices, and determines the hidden danger level of the home devices according to the hidden danger coefficients;

[0012] The execution module is used to take corresponding execution operations according to the hidden danger level of the home devices.

[0013] Further, the process of collecting the operation data of each home device in the home environment includes:

[0014] A plurality of home devices are numbered and denoted as i, i = 1, 2, 3, …, n, n is a natural number greater than 0, and a data buffer area is constructed for each home device, and the operation data of each home device is stored in the data buffer area after being collected in real time by each home device;

[0015] A data packaging node is created in the data buffer area of each home device;

[0016] The operation data stored in the data buffer area is packaged by the data packaging node, the packaged data amount is set, the data storage amount of the collected operation data in the data buffer area is recorded in real time, when the data storage amount reaches the packaged data amount, the corresponding operation data is packaged to generate a data subset, and the corresponding collection time period is labeled for the data subset, the data collection period corresponding to each home device is set, and all data subsets packaged in the data collection period are uploaded to the data processing module.

[0017] Further, the process of data cleaning and numerical normalization of the operation data of each home device includes:

[0018] The computing power allocation period, the cleaning period and the normalization period are set;

[0019] In the computing power allocation period, according to the number of data subsets uploaded by each home device to the data processing module, the computing power resource is called, and the computing power size for data processing of the operation data corresponding to each home device is allocated;

[0020] In the cleaning period, according to the computing power resource allocated to each home device, the pre-constructed data cleaning program is started, and a plurality of data subsets of each home device are sequentially cleaned, so as to correct the error data in the data subset, eliminate the redundant data in the data subset, and fill the missing data in the data subset, and then construct a plurality of standard data subsets corresponding to each home device in each collection period;

[0021] In the normalization period, the numerical normalization of a plurality of standard data subsets of each home device is performed to obtain the normalization coefficient of the real-time operation data of each home device in different collection periods; specifically including the normalization coefficient of the device voltage, the device power and the device temperature.

[0022] Further, the process of feature interval screening to obtain the principal component feature data corresponding to each home device includes:

[0023] The first screening interval, the second screening interval and the third screening interval are set.

[0024] The first screening interval is used for screening effective to-be-analyzed voltages;

[0025] The second screening interval is used for screening effective to-be-analyzed powers;

[0026] The third screening interval is used for screening effective to-be-analyzed temperatures;

[0027] The effective to-be-analyzed voltages, the effective to-be-analyzed powers, and the effective to-be-analyzed temperatures of each household device are integrated, and then the principal component feature data corresponding to each household device is obtained.

[0028] Further, the process of constructing the AI analysis model for analyzing the hidden dangers of the household device includes:

[0029] Obtaining household device running data of a plurality of data quantities at a historical time node, the household device running data at the historical time node being used for recording the running state of the household device, including household devices with normal running states and corresponding characteristic component data, and household devices with abnormal running states and corresponding characteristic component data;

[0030] According to the AI technology, an initial AI analysis model is constructed, the household device running data at the historical time node is input into the initial AI analysis model, the characteristic component data of each household device is analyzed, and the running state of each household device is output;

[0031] When the running state output by the initial AI analysis model is consistent with the actual running state of the household device, a correct analysis behavior is recorded, otherwise, an incorrect analysis behavior is recorded;

[0032] The model prediction accuracy of the initial AI analysis model is obtained, denoted as Sc;

[0033]

[0034] An iteration threshold is set, denoted as Dd;

[0035] When Sc < Dd, the initial AI analysis model is subjected to model iteration, and the model prediction accuracy of the AI analysis model after each model iteration is compared with the iteration threshold in terms of numerical value;

[0036] When Sc ≥ Dd, the model iteration is stopped, and the final AI analysis model is constructed.

[0037] Further, the process of inputting the principal component feature data of each household device into the AI analysis model to obtain the hidden danger items and the hidden danger coefficients corresponding to the household device includes:

[0038] The principal component feature data of each home device is sequentially input into the AI analysis model;

[0039] The AI analysis model synchronously constructs a time series prediction layer, an anomaly detection layer, and a knowledge graph mapping layer;

[0040] The time series prediction layer is used to sort the effective to-be-analyzed voltage, the effective to-be-analyzed power, and the effective to-be-analyzed temperature included in the principal component feature data according to the running time stamp of the home device in chronological order, and analyze the dynamic change trend of the device voltage, the device power, and the device temperature in the running process of the home device;

[0041] The anomaly detection layer is used to identify and mark the running time stamp of the device voltage, the device power, and the device temperature of the home device deviating from the normal change trend, derive the abnormal device working condition point, and obtain the detailed running data of the home device at the abnormal device working condition point, analyze and derive all hidden trouble items of the home device, and the hidden trouble coefficient corresponding to each hidden trouble item;

[0042] The knowledge graph mapping layer is used to construct the hidden trouble causal chain of each home device, and the structure of the hidden trouble causal chain is: hidden trouble cause-hidden trouble item-hidden trouble coefficient. In the constructed knowledge graph mapping layer, a mapping node is created for each home device, and the hidden trouble causal chain of each home device is mapped into the corresponding mapping node.

[0043] Further, the process of classifying the hidden trouble level of the home device according to the hidden trouble coefficient includes:

[0044] Let the hidden trouble coefficient of the home device be YH;

[0045] Set the first hidden trouble interval and the second hidden trouble interval, respectively, as Δ1 and Δ2;

[0046] When YH∈Δ1, the hidden trouble level of the corresponding home device is classified as a low-risk level;

[0047] When YH∈Δ2, the hidden trouble level of the corresponding home device is classified as a high-risk level.

[0048] Further, the process of taking corresponding execution operations according to the hidden trouble level of the home device includes:

[0049] When the hidden trouble level of the home device is a low-risk level, the hidden trouble causal chain of the home device in the knowledge graph mapping layer is retrieved through the AI analysis model, and a maintenance suggestion corresponding to the current home device is created according to the hidden trouble causal chain;

[0050] When the hidden trouble level of the home device is a high-risk level, an APP notification is pushed to the user corresponding to the home device, and the corresponding home device is powered off, and a relevant maintenance personnel is arranged to perform device maintenance on the powered-off home device.

[0051] Compared with the prior art, the present application has the following advantages:

[0052] 1. The data is dynamically packaged and collected through the data buffer area, the computing power resource allocation is optimized by combining the computing power matching mechanism, the efficiency of data cleaning, numerical normalization processing and feature screening is improved, the feature interval screening is adopted, the redundant data is eliminated, the principal component feature data is retained, the calculation complexity of the AI analysis model is reduced, and the data processing efficiency is effectively improved.

[0053] 2. The AI analysis model with a multi-level relationship of a time series prediction layer, an anomaly detection layer and a knowledge graph mapping layer is constructed to dynamically analyze the equipment operation trend, identify abnormal working conditions and generate a hidden danger causal chain, divide the low-risk and high-risk levels of the household equipment based on the hidden danger coefficient, realize the hidden danger risk quantitative evaluation, and effectively improve the hidden danger identification accuracy.

[0054] 3. The targeted maintenance suggestion is automatically generated at the low-risk level, the APP notification is immediately pushed at the high-risk level, the related maintenance personnel is dispatched after remote power-off, a "monitoring-analysis-response" closed loop is formed, the accident rate is effectively reduced, and the rapid blocking of high-risk hidden dangers and the daily monitoring of low-risk hidden dangers in household equipment are realized. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 The schematic diagram of the present application. DETAILED DESCRIPTION

[0056] As shown in Figure 1 , the AI-based household safety hidden danger intelligent detection and early warning system comprises a warning center, which is communicatively connected with a data acquisition module, a data processing module, an AI analysis module and an execution module;

[0057] The data acquisition module is used for acquiring the operation data of each household equipment in the household environment.

[0058] The data processing module is used for data cleaning, numerical normalization processing and feature interval screening of the operation data of each household equipment, and then the principal component feature data corresponding to each household equipment is obtained.

[0059] The AI analysis module constructs an AI analysis model for hidden danger analysis of household equipment, inputs the principal component feature data of each household equipment into the AI analysis model, obtains the hidden danger items and the hidden danger coefficient corresponding to the household equipment, and divides the hidden danger level of the household equipment according to the hidden danger coefficient.

[0060] The execution module is used for taking corresponding execution operation according to the hidden danger level of the household equipment.

[0061] It needs to be further explained that in the specific implementation process, the process of collecting the operation data of each household device in the home environment includes:

[0062] A plurality of household devices are numbered, and the number is denoted as i, i.e. i = 1, 2, 3, …, n, where n is a natural number greater than 0, and a data buffer area corresponding to each household device is constructed, and the operation data of each household device is stored in the data buffer area after being collected in real time by the household device;

[0063] A data packaging node is created in the data buffer area of each household device;

[0064] The operation data stored in the data buffer area is packaged by the data packaging node, the packaged data amount is set, the data storage amount of the collected operation data in the data buffer area is recorded in real time, and when the data storage amount reaches the packaged data amount, the corresponding operation data is packaged to generate a data subset, and the data subset is labeled with the corresponding collection time period;

[0065] The data collection period corresponding to each household device is set, and all data subsets packaged in the data collection period are uploaded to the data processing module for subsequent processing operation by the data processing module;

[0066] It needs to be noted that if the data storage amount of the remaining operation data in the data buffer area of the household device has not reached the packaged data amount at the end of the data collection period of the household device, the amount difference between the packaged data amount and the data storage amount is calculated, and the empty character data corresponding to the amount difference value is filled into the remaining operation data to reach the packaged data amount, and the data subset is constructed.

[0067] It needs to be further explained that in the specific implementation process, the process of data cleaning and value normalization processing of the operation data of each household device includes:

[0068] The computing power allocation period, the cleaning period and the normalization period are set;

[0069] In the computing power allocation period, the number of data subsets uploaded by each household device to the data processing module is used to retrieve computing power resources, and the computing power size for data processing of the operation data of each household device is allocated;

[0070] The number of data subsets of the operation data of the household device numbered i is denoted as S[i], the unit computing power is set, the value of the unit computing power is denoted as cal, the unit computing power is used for data processing of one data subset, the total operation computing power of the household device numbered i is denoted as Cal[i], and the calculation formula of Cal[i] is as follows:

[0071] Cal[i] = S[i] x cal;

[0072] During the cleaning period, according to the computing power resources allocated to each household device, a pre-constructed data cleaning program is started to sequentially clean the data of each household device, and then correct the error data in the data subset, eliminate the redundant data in the data subset, and fill in the missing data in the data subset;

[0073] Then, a number of standardized data subsets corresponding to each household device in each collection period are constructed;

[0074] During the normalization period, the numerical normalization processing is performed on the number of standardized data subsets of each household device, and then the normalization coefficient corresponding to the real-time running data of each household device in different collection periods is obtained;

[0075] The specific data normalization processing process is as follows:

[0076] For a household device, the instantaneous running data of a collection time corresponding to the record of the standardized data subset of the household device in each collection period is denoted as D[t], wherein t is a collection time in a collection period, and the instantaneous running data of each collection time of the household device includes device instantaneous voltage, device instantaneous power and device instantaneous temperature;

[0077] The instantaneous running data D[t] is expressed as follows:

[0078] D[t] = [V[t]-s, P[t]-s, C[t]-s];

[0079] Wherein, V[t]-s, P[t]-s and C[t]-s are the device instantaneous voltage, device instantaneous power and device instantaneous temperature corresponding to the household device at t collection time;

[0080] Taking a household device as an example, the instantaneous running data of each collection time in each collection period is sequentially normalized, which is: obtaining the voltage extreme value, power extreme value and temperature extreme value of the household device in each collection period, combining the device instantaneous voltage, device instantaneous power and device instantaneous temperature of each collection time in each collection period, and calculating the normalization coefficient of the current household device corresponding to the running data in each collection period;

[0081] Wherein, the voltage extreme value includes the maximum voltage and the minimum voltage, denoted as V-max and V-min respectively;

[0082] The power extreme value includes the maximum power and the minimum power, denoted as P-max and P-min respectively;

[0083] The temperature extreme value includes the highest temperature and the lowest temperature, denoted as C-max and C-min respectively;

[0084] The normalized coefficient of the device voltage corresponding to the instantaneous running data of the household equipment numbered i at each collection time point under a certain collection period is denoted as follows:

[0085] The normalized coefficient of the device voltage is denoted as a[i];

[0086] The normalized coefficient of the device power is denoted as β[i];

[0087] The normalized coefficient of the device temperature is denoted as γ[i];

[0088] a[i] = (V[t]-s-V-min) / (V-max-V-min);

[0089] β[i] = (P[t]-s-P-min) / (P-max-P-min);

[0090] γ[i] = (C[t]-s-C-min) / (C-max-C-min).

[0091] It needs to be further explained that, in the specific implementation process, the process of screening the feature interval and then obtaining the principal component feature data corresponding to each household equipment includes:

[0092] A first screening interval, a second screening interval, and a third screening interval are set;

[0093] The first screening interval is used to screen effective to-be-analyzed voltages;

[0094] The second screening interval is used to screen effective to-be-analyzed powers;

[0095] The third screening interval is used to screen effective to-be-analyzed temperatures;

[0096] The first screening interval, the second screening interval, and the third screening interval are denoted as Ω1, Ω2, and Ω3 respectively, wherein Ω1 = [0.2, 0.8], Ω2 = [0.3, 0.75], and Ω3 = [0.4, 0.9];

[0097] All device voltages with a[i] ∈ Ω1 are screened as effective to-be-analyzed voltages of the household equipment, and all device voltages with a[i] ∉ Ω1 are excluded; All device powers with β[i] ∈ Ω2 are screened as effective to-be-analyzed powers of the household equipment, and all device powers with β[i] ∉ Ω2 are excluded;

[0098] All device temperatures with γ[i] ∈ Ω3 are screened as effective to-be-analyzed temperatures of the household equipment, and all device temperatures with γ[i] ∉ Ω3 are excluded;

[0099] All device temperatures with γ[i] ∈ Ω3 are screened as effective to-be-analyzed temperatures of the household equipment, and all device temperatures with γ[i] ∉ Ω3 are excluded; ​all the equipment temperatures of the household equipment are rejected;

[0100] The effective voltages to be analyzed, the effective powers to be analyzed, and the effective temperatures to be analyzed obtained by each household equipment are integrated to obtain the principal component feature data corresponding to each household equipment.

[0101] It should be further explained that, in the specific implementation process, the process of constructing the AI analysis model for analyzing the hidden dangers of the household equipment includes:

[0102] Obtain household equipment operation data of a plurality of data amounts corresponding to the household equipment at a historical time node, the household equipment operation data of the plurality of data amounts at the historical time node being used to record the running state of the household equipment, including household equipment whose running state is normal and the corresponding characteristic component data thereof, and also including household equipment whose running state is abnormal and the corresponding characteristic component data thereof;

[0103] Construct an initial AI analysis model according to AI technology;

[0104] The AI analysis model is used for analyzing hidden dangers of the household equipment;

[0105] Input the household equipment operation data of the household equipment at the historical time node into the initial AI analysis model, analyze the characteristic component data of each household equipment by the initial AI analysis model, and then output the running state of each household equipment, when the running state output by the initial AI analysis model is consistent with the actual running state of the household equipment, record a correct analysis behavior, otherwise, record an error analysis behavior;

[0106] Obtain the model prediction accuracy of the initial AI analysis model, and denote the model prediction accuracy as Sc;

[0107]

[0108] Set an iteration threshold, and denote the iteration threshold as Dd;

[0109] When Sc < Dd, set a training set, a test set, and a validation set, perform model iteration on the initial AI analysis model, and obtain the model prediction accuracy of the AI analysis model after each model iteration, and perform numerical size judgment on the model prediction accuracy and the iteration threshold;

[0110] When Sc ≥ Dd, stop model iteration, and construct a final AI analysis model.

[0111] It should be further explained that, in the specific implementation process, the process of inputting the principal component feature data of each household equipment into the AI analysis model to obtain the hidden danger items and the hidden danger coefficients corresponding to the household equipment includes:

[0112] The principal component feature data of each household device is sequentially input into the AI analysis model;

[0113] The AI analysis model synchronously constructs a time series prediction layer, an anomaly detection layer, and a knowledge graph mapping layer;

[0114] The time series prediction layer is used to sort the effective to-be-analyzed voltage, the effective to-be-analyzed power, and the effective to-be-analyzed temperature included in the principal component feature data according to the running time stamp of the household device in chronological order, and analyze the dynamic change trend of the device voltage, the device power, and the device temperature in the running process of the household device;

[0115] The anomaly detection layer is used to identify and mark the running time stamp of the device voltage, the device power, and the device temperature of the household device deviating from the normal change trend, thereby obtaining an abnormal device working condition point and acquiring detailed running data of the household device corresponding to the abnormal device working condition point;

[0116] According to the detailed running data of the household device at the abnormal device working condition point, all hidden trouble items of the household device and the respective hidden trouble coefficients of each hidden trouble item are analyzed and obtained;

[0117] The knowledge graph mapping layer is used to construct a hidden trouble causal chain of each household device, and the structure of the hidden trouble causal chain is: hidden trouble cause-hidden trouble item-hidden trouble coefficient. In the constructed knowledge graph mapping layer, a mapping node is created for each household device, and the hidden trouble causal chain of each household device is mapped into the respective mapping node.

[0118] The hidden trouble cause is the specific accident reason corresponding to the abnormality of the household device, the hidden trouble item is the specific accident type corresponding to the abnormality of the household device, and the hidden trouble coefficient represents the probability of the occurrence of various specific accident types of the household device.

[0119] It needs to be further explained that, in the specific implementation process, the process of dividing the hidden trouble level of the household device according to the hidden trouble coefficient includes:

[0120] The hidden trouble coefficient of the household device is denoted as YH, and YH takes a real number in the interval (0.4, 1);

[0121] A first hidden trouble interval and a second hidden trouble interval are set;

[0122] The first hidden trouble interval and the second hidden trouble interval are denoted as Δ1 and Δ2, respectively;

[0123] Δ1 = (0.4, 0.7], and Δ2 = (0.7, 1);

[0124] When YH ∈ Δ1, the hidden trouble level of the corresponding household device is divided as a low-risk level;

[0125] When YH∈Δ2, the hidden danger level of the corresponding home device is determined as a high-risk level.

[0126] It should be further explained that, in the specific implementation process, the process of taking corresponding execution operations according to the hidden danger level of the home device includes:

[0127] When the hidden danger level of the home device is a low-risk level, the hidden danger causal chain of the home device in the knowledge graph mapping layer is called through the AI analysis model, and the maintenance suggestion corresponding to the current home device is created according to the hidden danger causal chain;

[0128] Among them, the maintenance suggestion includes checking circuit load, checking mechanical performance, checking equipment function completion degree and checking equipment appearance, etc.

[0129] When the hidden danger level of the home device is a high-risk level, an APP notification is pushed to the user corresponding to the home device, and the corresponding home device is powered off, and the related maintenance personnel are arranged to carry out equipment maintenance on the powered-off home device.

[0130] The above embodiments are only used to illustrate the technical method of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. An AI-based home safety hazard intelligent detection and early warning system, comprising an early warning center, characterized in that, The early warning center is communicatively connected with a data acquisition module, a data processing module, an AI analysis module and an execution module; The data acquisition module is configured to acquire operation data of each household device in the household environment; The data processing module is configured to perform data cleaning, value normalization processing and feature interval screening on the operation data of each household device, and then obtain principal component feature data corresponding to each household device; The AI analysis module constructs an AI analysis model for hidden danger analysis of the household device, inputs the principal component feature data of each household device into the AI analysis model, obtains hidden danger items and a hidden danger coefficient corresponding to the household device, and determines a hidden danger level of the household device according to the hidden danger coefficient; The execution module is configured to take corresponding execution operations according to the hidden danger level of the household device; The process of inputting the principal component feature data of each household device into the AI analysis model to obtain hidden danger items and a hidden danger coefficient corresponding to the household device includes: The principal component feature data of each household device is sequentially input into the AI analysis model; The AI analysis model synchronously constructs a time series prediction layer, an anomaly detection layer and a knowledge graph mapping layer; The time series prediction layer is configured to sort effective to-be-analyzed voltages, effective to-be-analyzed powers and effective to-be-analyzed temperatures included in the principal component feature data according to the running time stamps of the household device in chronological order, and analyze the dynamic change trend of the device voltage, the device power and the device temperature in the running process of the household device; The anomaly detection layer is configured to identify and mark the running time stamps at which the device voltage, the device power and the device temperature of the household device deviate from the normal change trend, obtain abnormal device working condition points, acquire detailed running data of the household device at the abnormal device working condition points, analyze all hidden danger items of the household device, and obtain a hidden danger coefficient corresponding to each hidden danger item; The knowledge graph mapping layer is configured to construct a hidden danger causal chain of each household device, and the structure of the hidden danger causal chain is: hidden danger cause-hidden danger item-hidden danger coefficient; a mapping node is created for each household device in the constructed knowledge graph mapping layer, and the hidden danger causal chain of each household device is mapped into the corresponding mapping node. 2.The AI-based home safety hazard intelligent detection and early warning system according to claim 1, characterized in that, The process of acquiring operation data of each household device in the household environment includes: A plurality of household devices are numbered and denoted as i, i=1, 2, 3, …, n, n is a natural number greater than 0, and a data buffer area is constructed for each household device, and the running data of each household device is stored in the data buffer area after being acquired by the household device; A data packaging node is created in the data buffer area of each household device; The data packaging node packages the running data stored in the data buffer area, sets a packaging data amount, and records the data amount of the acquired running data in the data buffer area in real time, when the data amount reaches the packaging data amount, the corresponding running data is packaged to generate a data subset, and the data subset is labeled with the corresponding acquisition time period, and the data acquisition period of each household device is set, and all data subsets packaged in the data acquisition period are uploaded to the data processing module. 3.The AI-based home safety hazard intelligent detection and early warning system according to claim 2, characterized in that, The data cleaning and value normalization process of the operation data of each household device includes: setting a computing power matching period, a cleaning period, and a normalization period; in the computing power matching period, according to the number of data subsets uploaded by each household device to the data processing module, the computing power resource is called to match the computing power size for data processing of the operation data of each household device; in the cleaning period, according to the computing power resource allocated to each household device, the pre-constructed data cleaning program is started to sequentially clean the data subsets of each household device, thereby correcting the error data in the data subsets, eliminating the redundant data in the data subsets, and filling the missing data in the data subsets, and thereby constructing a plurality of standardized data subsets corresponding to each household device in each collection period; in the normalization period, the plurality of standardized data subsets of each household device are subjected to value normalization to obtain the normalization coefficient of the real-time operation data of each household device in different collection periods; specifically including the normalization coefficient of the device voltage, the device power, and the device temperature. 4.The AI-based home safety hazard intelligent detection and early warning system according to claim 3, characterized in that, The process of feature interval screening to obtain the principal component feature data corresponding to each household device includes: setting a first screening interval, a second screening interval, and a third screening interval; the first screening interval is used to screen effective to-be-analyzed voltage; the second screening interval is used to screen effective to-be-analyzed power; the third screening interval is used to screen effective to-be-analyzed temperature; the effective to-be-analyzed voltage, the effective to-be-analyzed power, and the effective to-be-analyzed temperature of each household device are integrated to obtain the principal component feature data corresponding to each household device. 5.The AI-based home safety hazard intelligent detection and early warning system according to claim 4, characterized in that, The process of constructing an AI analysis model for hidden danger analysis of household devices includes: obtaining a plurality of data quantities of household device operation data at historical time nodes, the household device operation data at the historical time nodes being used to record the running state of the household device, including household devices with normal running state and their corresponding feature component data, and household devices with abnormal running state and their corresponding feature component data; constructing an initial AI analysis model according to AI technology, inputting the household device operation data at the historical time nodes into the initial AI analysis model, analyzing the feature component data of each household device, and outputting the running state of each household device; when the running state output by the initial AI analysis model is consistent with the actual running state of the household device, a correct analysis behavior is recorded, otherwise, an error analysis behavior is recorded; obtaining the model prediction accuracy of the initial AI analysis model, denoted as Sc; ； setting an iteration threshold, denoted as Dd; when Sc < Dd, the initial AI analysis model is subjected to model iteration, and the model prediction accuracy of the AI analysis model after each model iteration is compared with the iteration threshold in terms of numerical size; when Sc ≥ Dd, the model iteration is stopped, and the final AI analysis model is constructed. 6.The AI-based home safety hazard intelligent detection and early warning system according to claim 5, characterized in that, The process of delimiting the hidden danger level of the household device according to the hidden danger coefficient includes: denoting the hidden danger coefficient of the household device as YH; setting a first hidden danger interval and a second hidden danger interval, denoted as Δ1 and Δ2, respectively; When YH∈∆1, the hidden danger level of the corresponding home equipment is classified as a low-risk level; When YH∈∆2, the hidden danger level of the corresponding home equipment is classified as a high-risk level. 7.The AI-based home safety hazard intelligent detection and early warning system according to claim 6, characterized in that, The process of taking corresponding execution operations according to the hidden danger level of the home equipment includes: When the hidden danger level of the home equipment is a low-risk level, the hidden danger causal chain of the home equipment in the knowledge graph mapping layer is called through the AI analysis model, and the maintenance suggestion corresponding to the current home equipment is created according to the hidden danger causal chain; When the hidden danger level of the home equipment is a high-risk level, an APP notification is pushed to the user to which the home equipment belongs, the corresponding home equipment is powered off, and the relevant maintenance personnel are arranged to perform equipment maintenance on the powered-off home equipment.

Images