Smart home operation optimization method and system based on optimal combination of equipment genes

Abstract

The invention discloses a smart home equipment operation optimization method and system based on equipment gene optimal combination, and the method comprises the steps: converting the multi-dimensional original information of smart home equipment into four equipment bases A, C, T and U, and generating a three-base gene segment representing the operation of the equipment; screening out long-life equipment from historical equipment, and performing genetic treatment on the long-life equipment to obtain a long-life gene chain; performing distillation fine adjustment on the pre-trained large model by adopting a long-life gene chain, so that the large model learns to identify and apply an optimal gene combination; obtaining the similarity between the real-time working condition sequence and the long-life gene chain through a fine-tuned large model, scanning key indexes based on the similarity, and carrying out nearest neighbor search to obtain an equipment operation recommendation threshold value; and storing the longevity gene chain locally, comparing the deviation degree between the real-time gene chain corresponding to the current working condition sequence of the smart home equipment and the longevity gene chain, and pulling the equipment operation parameters back to the equipment operation recommendation threshold value range when the deviation is relatively large, so that the life optimization of the smart home equipment is realized.

Description

Technical Field

[0001] This invention belongs to the field of smart home technology, specifically, it relates to a smart home operation optimization method and system based on the optimal combination of device genes. Background Technology

[0002] Traditional smart home device status monitoring solutions employ an independent monitoring model of "sensor acquisition - feature extraction - status judgment". Various sensors (vibration, temperature, current, acoustic) collect device operating data separately, extract single features such as RMS value, peak value, and temperature through signal processing, and then determine the current status of the device (normal / abnormal) based on thresholds.

[0003] Traditional predictive maintenance of equipment mainly adopts the "sensor monitoring - feature engineering - machine learning modeling - life prediction" model. It extracts time domain, frequency domain, and time-frequency domain features from multi-sensor data such as vibration, temperature, and current, and then uses algorithms such as random forest, support vector machine, and neural network to build a prediction model for the remaining life of the equipment.

[0004] Traditional solutions focus more on when failures will occur, rather than how to extend the lifespan of equipment. Summary of the Invention

[0005] The purpose of this invention is to overcome the limitations of existing technologies that focus on fault prediction, and to propose a smart home operation optimization method and system based on device gene combination. It establishes a gene coding system for smart home appliances, and through distillation fine-tuning of pre-trained large models, enables the large models to learn to recognize and apply longevity gene combinations, thereby providing professional longevity operation management for the devices. By drawing on biological gene coding, a new paradigm of proactive longevity optimization is realized.

[0006] The present invention is implemented using the following technical solutions:

[0007] A smart home operation optimization method based on the optimal combination of device genes is proposed, including:

[0008] S1, the gene expression of building smart home devices;

[0009] S2, identify the longevity gene chain of smart home devices;

[0010] S3 uses the longevity gene chain as the teacher signal to distill and fine-tune the pre-trained large model;

[0011] S4, based on the longevity gene chain and the fine-tuned large model, generates recommended thresholds for device operation;

[0012] S5 calculates the deviation between the current gene chain and the longevity gene chain in real time. When the deviation exceeds the set threshold, it adjusts the device's operating parameters to align with the recommended operating threshold.

[0013] In some embodiments of the present invention, S1 specifically includes:

[0014] Collect multi-dimensional raw information from smart home devices;

[0015] The raw information is converted into four types of device base values: A, C, T, and U, using a set base calculation formula. Among them, A represents the stability gene, C represents the connectivity gene, T represents the time-dependent gene, and U represents the adaptability gene.

[0016] The three-base gene segments of the characterization device are generated according to the set combination rules.

[0017] In some embodiments of the present invention, generating the tribase gene segment of the characterization device according to a set combination rule specifically includes:

[0018] The base combination is defined as [position 1][position 2][position 3]; where position 1 corresponds to a mechanical system, position 2 corresponds to an electrical system, and position 3 corresponds to a thermal system.

[0019] For mechanical systems, the maximum value between A and U bases is selected as position 1 of the gene segment;

[0020] For electrical systems, the maximum value between C and T bases is selected as position 2 of the gene segment;

[0021] For thermal systems, the maximum value between A and T bases is selected as the location of the gene segment.

[0022] In some embodiments of the present invention, S2 specifically includes:

[0023] Long-life equipment that meets the longevity threshold is selected from the historical equipment group; the longevity threshold is: continuous operation for ≥ N years and failure rate in the Nth year ≤ a set ratio;

[0024] Genetically modify longevity devices to obtain a set of longevity genes;

[0025] Based on frequency and correlation analysis, longevity gene segments were extracted from the longevity gene set;

[0026] The extracted longevity gene segments are linked together in descending order of frequency to form a longevity gene chain.

[0027] In some embodiments of the present invention, S3 specifically includes:

[0028] Constructing a training set includes: using M longevity devices as the data source, obtaining the real-time operating condition sequence X for each device over a continuously set time; using the longevity gene chain as the teacher signal Y, calculating the edit distance d between the real-time operating condition sequence X and the teacher signal Y, and following... Obtain lifespan label Construct triples (X, Y, ... ) serves as regression supervision for pre-trained large models; among which, For the equipment's reference lifespan, This represents the total number of segments in the longevity gene chain.

[0029] Distillation fine-tuning is performed on the pre-trained large model, and the model outputs lifespan predictions and the similarity between the current working condition sequence and the longevity gene chain; the loss function is defined as: ; This is the predicted lifespan value. This represents the mean squared error loss between the predicted lifetime value and the lifetime label. The similarity between the real-time operating condition sequence X and the longevity gene chain Y is used. for and Cosine similarity; The standard vector of the longevity gene chain. These are the weighting coefficients.

[0030] In some embodiments of the present invention, S4 specifically includes:

[0031] The similarity between real-time operating condition sequences and longevity gene chains was obtained using a finely tuned large model.

[0032] Nearest neighbor search is performed based on key similarity metrics.

[0033] The parameters ranked first are combined to form a chain-parameter mapping table, which serves as the recommended threshold for device operation.

[0034] In some embodiments of the present invention, the method further includes:

[0035] Collect operational data and behavior logs from smart home devices;

[0036] The large model is periodically fine-tuned based on the collected data.

[0037] A smart home operation optimization system based on the optimal combination of device genes is proposed, including:

[0038] Gene coding units are used to construct gene expression for smart home devices;

[0039] The longevity gene chain mining unit is used to determine the longevity gene chain of smart home devices;

[0040] The large model fine-tuning unit is used to distill and fine-tune the pre-trained large model using the longevity gene chain as the teacher signal.

[0041] The threshold generation unit is used to generate recommended thresholds for the device based on the longevity gene chain and the fine-tuned large model.

[0042] The operation optimization unit is used to calculate the deviation between the current gene chain of the device and the longevity gene chain in real time. When the deviation exceeds the set threshold, the device operation parameters are adjusted to align with the recommended operating threshold of the device.

[0043] Compared with existing technologies, the advantages and positive effects of this invention are as follows: The smart home device operation optimization method and system proposed in this invention, based on the optimal combination of device genes, converts the multi-dimensional original information of smart home devices into four types of device bases (A, C, T, and U) to generate a three-base gene segment characterizing device operation; it screens out long-lived devices from historical devices and performs genetic processing on these devices to obtain a long-lived gene chain; it uses the long-lived gene chain and historical operating data of the long-lived devices to fine-tune a pre-trained large model, enabling the large model to learn to recognize and apply the optimal gene combination; it obtains the similarity between the real-time operating sequence and the long-lived gene chain through the fine-tuned large model, and performs nearest neighbor search based on this similarity to scan key indicators to obtain the recommended threshold for device operation; it stores the long-lived gene chain locally and compares the deviation between the real-time gene chain corresponding to the current operating sequence of the smart home device and the long-lived gene chain. When the deviation is large, it pulls the device operation parameters back to the recommended threshold range, thereby achieving lifespan optimization of smart home devices and breaking through the limitations of existing technologies that focus on fault prediction.

[0044] The system of this invention solves the feature fragmentation problem through base mapping and discovers the optimal feature combination through gene expression, shifting from passive monitoring and alarm to active identification and guidance, and realizing closed-loop optimization of the entire life cycle of design, operation, maintenance and user.

[0045] Other features and advantages of the present invention will become clearer after reading the detailed description of the embodiments of the present invention in conjunction with the accompanying drawings. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This diagram illustrates the implementation steps of the smart home large-scale model enhancement method based on the discovery of optimal combination of device genes proposed in this invention.

[0048] Figure 2 This diagram illustrates the execution process of the smart home large-scale model enhancement method based on the discovery of optimal combination of device genes proposed in this invention. Detailed Implementation

[0049] 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.

[0050] The purpose of this invention is to overcome the limitations of existing technologies that focus primarily on fault prediction, and to achieve a new paradigm of proactive longevity optimization by drawing on the coding of biological genes. Based on this idea, this invention proposes a smart home large-scale model enhancement method based on the discovery of device gene combinations, which mainly consists of four parts:

[0051] 1. Drawing inspiration from the coding methods of biological genes, the physical characteristics of smart home appliances are mapped to four types of device bases: A (vibration stability), C (current stability), T (mechanical wear), and U (adaptability). By combining three bases, device gene sequences such as AAC, UTT, and ACU are formed, thereby realizing the genetic expression of device characteristics.

[0052] 2. Identify long-life equipment with low failure rates within a set timeframe from among tens of thousands of devices, analyze the genetic code of these long-life equipment, and find the longevity gene chain.

[0053] 3. Distill and fine-tune the pre-trained large model to enable it to learn to recognize and apply longevity gene chains, and output predicted lifespan and similarity with longevity gene chains based on the current state of the device and the longevity gene chain.

[0054] 4. Construct a chain-parameter mapping table for the longevity gene chain and store the longevity gene chain and the chain-parameter mapping table locally. Compare the deviation between the real-time gene string corresponding to the real-time operating condition sequence of the device and the longevity gene chain. When the deviation is large, trigger PID correction to pull the device operating parameters back to the range of the chain-parameter mapping table, so that the lifespan of smart home appliances can be improved through software algorithms.

[0055] Specifically, such as Figure 1 As shown, it includes:

[0056] S1: Gene expression for building smart home devices.

[0057] First, collect multi-dimensional raw information from smart home devices; this multi-dimensional raw information includes:

[0058] (1) Operation status data.

[0059] Power consumption data: real-time power consumption, standby power consumption, peak power consumption.

[0060] Response data: command response time, execution success rate, and error code frequency.

[0061] Sensor data: temperature, humidity, vibration, noise, electromagnetic interference.

[0062] (2) User interaction data.

[0063] Usage frequency data: Daily number of times the device is turned on and the distribution of usage time.

[0064] Time pattern data: usage habits in the morning, noon and evening, and differences between weekends and weekdays.

[0065] Operation sequence data: common function combinations and operation process preferences.

[0066] Environmental data: Changes in usage patterns in different seasons / weathers.

[0067] (3) Equipment environment data.

[0068] Working environment data: temperature range, humidity variation, air quality.

[0069] Spatial location data: room orientation, floor height, and obstruction status.

[0070] Power grid quality: voltage stability, harmonic interference, momentary power outages.

[0071] Next, these raw data are converted into gene base expression through feature engineering.

[0072] (1) Define the bases of the device gene as shown in Table 1 below:

[0073] Table 1: Definitions of Basic Bases

[0074] base Physical meaning Engineering Explanation A Stability genes Overall system stability C Connectivity genes Network connectivity reliability T Time-sensitive genes Response execution speed U Adaptive genes User habit adaptability

[0075] (2) Define the base combination form as: [position 1][position 2][position 3], as shown in Table 2 below:

[0076] Table 2 Base Combination Forms

[0077] Location system Base function 1st place Mechanical system A / C / T / U = Mechanical Stability / Connectivity / Aging / Adaptability 2nd place Electrical system A / C / T / U = Electrical stability / Connectivity / Timeliness / Adaptability 3rd place thermal system A / C / T / U = Thermal stability / Connectivity / Aging / Adaptability

[0078] (3) Define mapping rules.

[0079] Based on the above definition, in this invention, a tetrabase is mapped to a three-position base combination, and the mapping rules are shown in Table 3:

[0080] Table 3 Mapping from three- and four-base pairs to the third position

[0081] Location system Choose strategy For example 1st place Mechanical system Choose the maximum value between A and U. Max(A,U) → Stability priority or adaptation priority 2nd place Electrical system Choose the maximum value between C and T. Max(C,T) → Connection priority or time-sensitive priority 3rd place thermal system Choose the maximum value between A and T. Max(A,T) → Stability priority or timeliness priority

[0082] (4) Define the formula for calculating the base combination of the equipment: Since [position 1], [position 2], and [position 3] in the base combination represent different system performances, the formulas used for base calculation are also different, as shown in Table 4:

[0083] Table 4: Formulas for calculating bases at different positions

[0084] Position\base A (Stable) C (connection) T (Time-bound) U (Adaptation) 1st Machine 1 - 0.3 × vibration - 0.4 × temperature rise - 0.3 × noise 1 - 0.5 × packet loss - 0.5 × retransmission 1 - 0.6 × response delay - 0.4 × number of start / stop cycles 1 - 0.5 × frequency offset - 0.5 × load offset 2nd Electric 1 - 0.4 × current fluctuation - 0.3 × voltage drop - 0.3 × harmonic distortion 1 - 0.3 × packet loss - 0.3 × retransmission - 0.4 × communication jitter 1 - 0.5 × response latency - 0.5 × execution time 1 - 0.6 × habit matching degree - 0.4 × historical error 3rd heat 1 - 0.35 × temperature gradient - 0.35 × thermal inertia - 0.3 × heat dissipation efficiency 1 - 0.4 × packet loss - 0.6 × hotlink packet loss 1 - 0.4 × response delay - 0.3 × compressor uptime - 0.3 × temperature gradient 1 - 0.5 × mode switching frequency - 0.5 × number of user adjustments

[0085] It should be noted that all parameters described above are normalized values.

[0086] Table 5 below provides examples of some gene combinations:

[0087] Table 5 Examples of Tri-Base Combinations

[0088] Gene combination Physical meaning AAC [Mechanical stability][Electrical stability][Reliable thermal connection] AUU [Mechanical stability][Electrical adaptability][Thermal adaptability] ACU [Mechanical stability][Reliable electrical connections][Thermal adaptability] AUC [Mechanical stability][Electrical adaptability][Reliable thermal connection] …… ……

[0089] The following example uses data from the same 1.5 horsepower air conditioner to illustrate the genetic conversion of equipment.

[0090] The original air conditioning data is shown in Table 6:

[0091] Table 6 Original Data (Normalized)

[0092] index value index value index value vibration 0.08 Packet loss 0.02 Temperature gradient 0.07 Temperature rise 0.05 Retransmission 0.03 thermal inertia 0.06 noise 0.06 Response delay 0.10 Heat dissipation efficiency 0.05 Current fluctuation 0.06 Start-stop count 0.04 Compressor uptime 0.08 voltage drop 0.04 Frequency offset 0.07 Mode switching frequency 0.09 Harmonic distortion 0.05 Load offset 0.05 User adjustment times 0.08 Communication jitter 0.03 Habit matching degree 0.12 Execution time 0.09 Historical error 0.06

[0093] Substituting the above values ​​into the formula shown in Table 4, we obtain the base values ​​at each position in Table 7 below:

[0094] Table 7 Base Value Conversion

[0095] Position\base A C T U 1st Machine 1-0.3×0.08-0.4×0.05-0.3×0.06= 0.932 1-0.5×0.02-0.5×0.03= 0.975 1-0.6×0.10-0.4×0.04=0.924 1-0.5×0.07-0.5×0.05=0.940 2nd Electric 1-0.4×0.06-0.3×0.04-0.3×0.05= 0.947 1-0.3×0.02-0.3×0.03-0.4×0.03=0.970 1-0.5×0.10-0.5×0.09=0.905 1-0.6×0.12-0.4×0.06=0.904 3rd heat 1-0.35×0.07-0.35×0.06-0.3×0.05= 0.936 1-0.4×0.02-0.6×0.02= 0.980 1-0.4×0.10-0.3×0.08-0.3×0.07=0.891 1-0.5×0.09-0.5×0.08=0.915

[0096] Based on the mapping rules shown in Table 3, three-position bases were selected, as shown in Table 8:

[0097] Table 8 Screening Results

[0098] Location Candidate score Select 1st place A=0.932, U=0.940 U（0.940）0.932） 2nd place C=0.970, T=0.905 C（0.970）0.905） 3rd place A=0.936, T=0.891 A（0.936）0.891）

[0099] The final base combination obtained is: UCA (mechanical self-adaptation + electrical connection stability + thermal stability).

[0100] S2: Determine the longevity gene chain of smart home devices.

[0101] 1. Select long-life equipment.

[0102] Conduct full lifecycle tracking for no fewer than 10,000 home appliances of the same model, using "continuous operation for ≥N years and failure rate in year N ≤ a set percentage" as the longevity threshold. For example, continuous operation for ≥10 years and failure rate in year 10 ≤ 1.5%, locking in M ​​longevity devices representing a set percentage (e.g., 5%). Record all operational data, sensor data, maintenance records, etc., of these M home appliances throughout their complete lifecycle into a database to form a longevity device dataset.

[0103] 2. Genetically modified long-life equipment.

[0104] For the longevity equipment dataset, fields are extracted according to operating status data, user interaction data, and equipment environment data. Longevity gene sets are generated for each longevity equipment according to the contents of Tables 1 to 4. For example, with a window size of 30d and a step size of 7d, monthly gene strings are obtained: M units × 120 months ≈ 60,000 "longevity month - gene strings". These longevity gene data are recorded for subsequent statistical analysis.

[0105] 3. Gene segment frequency statistics.

[0106] The occurrence frequency of each single gene segment in 60,000 data points was counted, and Topm was selected as the candidate pool, for example, Top20. Then, Pearson correlation coefficient was calculated using the occurrence frequency of single segments as a feature and the actual lifespan in months as a label. Gene segments with r>0.5 were retained to obtain 7-12 highly correlated longevity gene segments. The obtained 7-12 gene segments were then concatenated end to end in descending order of frequency to form a longevity gene chain, for example, AAC-UTT-ACU-UCA-CUT.

[0107] S3: Using the longevity gene chain as a teacher signal, the pre-trained large model is fine-tuned by distillation.

[0108] 1. Construct a training set.

[0109] Using the aforementioned M longevity devices as the data source, a real-time operating condition sequence X is taken for each device over a continuously set time period (e.g., 720 hours). The longevity gene chain obtained in step S2 is used as the teacher signal Y. The edit distance d between the real-time operating condition sequence X and the teacher signal Y is calculated. Obtain lifespan label (Unit: Month), construct triples (X, Y, This serves as regression supervision for pre-trained large models (such as the Transformer model). For the equipment's reference lifespan, This represents the total number of segments in the longevity gene chain. It should be noted that the real-time operating condition sequence X, as input, needs to be encoded into a gene sequence before calculating the edit distance d with the teacher signal Y.

[0110] 2. Distillation fine-tuning is performed on the pre-trained large model to train the model to output lifespan prediction values ​​and the similarity between the current working condition sequence and the longevity gene chain.

[0111] The input prompt provides "aircraft model + real-time operating condition sequence + longevity gene chain", forcing the model to simultaneously output predicted lifespan. and real-time operating condition sequence With the longevity gene chain similarity .

[0112] The loss function is defined as:

[0113] ;

[0114] The model input includes a sequence of operating conditions. With the longevity gene chain Simultaneously output predicted lifetime Sequence-link proximity Supervisory signals Introduced only in the loss layer, this ensures that similarity can be calculated online without real-world lifetimes during the inference phase. The entire training process requires no manual setting of any thresholds or initial weight values. The standard vector of the longevity gene chain. The weight coefficients are automatically optimized using the gradient descent algorithm, when the MSE loss and Training stops when the preset balance ratio is reached. It should be noted that the real-time operating condition sequence... It needs to be encoded into a gene sequence before calculating its correlation with the longevity gene chain. The similarity.

[0115] S4: Recommended threshold for device operation based on longevity gene chains and fine-tuned large models.

[0116] For any new sequence of operating conditions of home appliances Sequence-link proximity was obtained using a finely tuned large model. ;based on By scanning key indicators and performing nearest neighbor retrieval, a chain-parameter mapping table is constructed using the top-1 parameter combinations found. For example, compressor frequency 72Hz, expansion valve reference opening 38%, internal fan speed 850rpm, defrost cycle 42min, oil return interval 8h, and superheat offset +1.2℃. This achieves a closed loop of statistical discovery and generative guidance, and the entire recommendation process does not require human experience.

[0117] S5: Calculates the deviation between the current gene chain and the longevity gene chain of the device in real time. When the deviation exceeds the set threshold, it adjusts the device operating parameters to align with the recommended operating threshold of the device.

[0118] The longevity gene chain is burned into a designated field of the device's main controller, EEPORM. When the home appliance is powered on, real-time operating data is collected at set intervals. The local MCU infers the real-time gene chain corresponding to the current operating sequence and calculates the deviation between the real-time gene chain and the longevity gene chain. If the deviation exceeds a set threshold, PID self-correction is immediately triggered to pull the compressor frequency, expansion valve opening, internal fan speed, etc., back into the chain-parameter mapping table range and record the number of pullbacks and other data for cloud analysis and statistics.

[0119] This invention is designed to provide users with daily longevity and health guidance: translating the longevity gene chain into guidance language, for example, by displaying a prompt message at a fixed time each day via a mobile app such as "Today's gene health score is 92, recommended energy-saving fan temperature 26℃, automatic fan speed." Alternatively, if the app detects that the user frequently manually lowers the air conditioner temperature, it may prompt, "Low temperature and high frequency will accelerate air conditioner degradation and may affect the air conditioner's lifespan. Do you want to enable the child lock?" Finally, after the user confirms with one click, the main controller automatically limits the lowest set temperature.

[0120] A specific embodiment of the method of the present invention is given below.

[0121] System scenario: A certain air conditioning assembly line, which began mass production of a new 1.5 horsepower model with first-class energy efficiency on March 15, 2024.

[0122] 1. Write it into the longevity gene chain.

[0123] The longevity gene chain obtained in step S2, such as AAC-UTT-ACU-UCA-ATT-CUT, and the chain-parameter mapping table obtained in step S4 are solidified into the main control EEPROM (address 0x1800-0x1830, occupying 48 bytes).

[0124] 2. Real-time alignment at the runtime.

[0125] After the air conditioner is powered on and started running, it collects operating data (12-dimensional operating status data, user interaction data, and equipment environment data) every set time. The local edge model infers the current gene string, such as AAC-UTT-ATT-UCA-ATT-CUT. It reads the longevity gene string and compares it with it, and obtains the 3rd degenerate (ACU→ATT). If the deviation is calculated to be 33%, it triggers PID self-correction and adjusts parameters such as compressor operating frequency and expansion valve opening to align with the parameters in the mapping table.

[0126] 3. Daily guidance for users.

[0127] The user's mobile app pushes daily guidance information and, in conjunction with user interactions (such as one-click confirmation), optimizes the lifespan of the air conditioner and sends behavior logs back to the cloud for the next round of fine-tuning.

[0128] It should be noted that, in the specific implementation process, the above methods can be implemented by a hardware processor executing computer-executable instructions in software form stored in memory, which will not be elaborated here. The programs corresponding to the actions executed can all be stored in the computer-readable storage medium of the system in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0129] The computer-readable storage media mentioned above may include volatile memory, such as random access memory; may also include non-volatile memory, such as read-only memory, flash memory, hard disk or solid-state drive; and may also include combinations of the above types of memory.

[0130] The term "processor" as mentioned above can also refer to a collective of multiple processing elements. For example, a processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor, and it can also be a special-purpose processor.

[0131] It should be noted that the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A smart home operation optimization method based on optimal combination of device genes, characterized in that, include: S1, the gene expression of building smart home devices; S2, identify the longevity gene chain of smart home devices; S3 uses the longevity gene chain as the teacher signal to distill and fine-tune the pre-trained large model; S4, based on the longevity gene chain and the fine-tuned large model, generates recommended thresholds for device operation; S5 calculates the deviation between the current gene chain and the longevity gene chain in real time. When the deviation exceeds the set threshold, it adjusts the device's operating parameters to align with the recommended operating threshold.

2. The smart home operation optimization method based on the optimal combination of device genes according to claim 1, characterized in that, S1 specifically includes: Collect multi-dimensional raw information from smart home devices; The raw information is converted into four types of device base values: A, C, T, and U, using a set base calculation formula. Among them, A represents the stability gene, C represents the connectivity gene, T represents the time-dependent gene, and U represents the adaptability gene. The three-base gene segments of the characterization device are generated according to the set combination rules.

3. The smart home operation optimization method based on the optimal combination of device genes according to claim 2, characterized in that, The three-base gene segments for characterization equipment are generated according to the set combination rules, specifically including: The base combination is defined as [position 1][position 2][position 3]; where position 1 corresponds to a mechanical system, position 2 corresponds to an electrical system, and position 3 corresponds to a thermal system. For mechanical systems, the maximum value between A and U bases is selected as position 1 of the gene segment; For electrical systems, the maximum value between C and T bases is selected as position 2 of the gene segment; For thermal systems, the maximum value between A and T bases is selected as the location of the gene segment.

4. The smart home operation optimization method based on the optimal combination of device genes according to claim 1, characterized in that, S2 specifically includes: Long-life equipment that meets the longevity threshold is selected from the historical equipment group; the longevity threshold is: continuous operation for ≥ N years and failure rate in the Nth year ≤ a set ratio; Genetically modify longevity devices to obtain a set of longevity genes; Based on frequency and correlation analysis, longevity gene segments were extracted from the longevity gene set; The extracted longevity gene segments are linked together in descending order of frequency to form a longevity gene chain.

5. The smart home operation optimization method based on the optimal combination of device genes according to claim 1, characterized in that, S3 specifically includes: Constructing a training set includes: using M longevity devices as the data source, obtaining the real-time operating condition sequence X for each device over a continuously set time; using the longevity gene chain as the teacher signal Y, calculating the edit distance d between the real-time operating condition sequence X and the teacher signal Y, and following... Obtain lifespan label Construct triples (X, Y, ... ) serves as regression supervision for pre-trained large models; among which, For the equipment's reference lifespan, This represents the total number of segments in the longevity gene chain. Distillation fine-tuning is performed on the pre-trained large model, and the model outputs lifespan predictions and the similarity between the current working condition sequence and the longevity gene chain; the loss function is defined as: ; This is the predicted lifespan value. This represents the mean squared error loss between the predicted lifetime value and the lifetime label. The similarity between the real-time operating condition sequence X and the longevity gene chain Y is used. for and Cosine similarity; The standard vector of the longevity gene chain. These are the weighting coefficients.

6. The smart home operation optimization method based on the optimal combination of device genes according to claim 5, characterized in that, S4 specifically includes: The similarity between real-time operating condition sequences and longevity gene chains was obtained using a finely tuned large model. Nearest neighbor search is performed based on key similarity metrics. The parameters ranked first are combined to form a chain-parameter mapping table, which serves as the recommended threshold for device operation.

7. The smart home operation optimization method based on the optimal combination of device genes according to claim 6, characterized in that, The method further includes: Collect operational data and behavior logs from smart home devices; The large model is periodically fine-tuned based on the collected data.

8. A smart home operation optimization system based on optimal combination of device genes, characterized in that, include: Gene coding units are used to construct gene expression for smart home devices; The longevity gene chain mining unit is used to determine the longevity gene chain of smart home devices; The large model fine-tuning unit is used to distill and fine-tune the pre-trained large model using the longevity gene chain as the teacher signal. The threshold generation unit is used to generate recommended thresholds for the device based on the longevity gene chain and the fine-tuned large model. The operation optimization unit is used to calculate the deviation between the current gene chain of the device and the longevity gene chain in real time. When the deviation exceeds the set threshold, the device operation parameters are adjusted to align with the recommended operating threshold of the device.

Images