Air conditioner load prediction model establishment method and system and storage medium

Abstract

The invention relates to the technical field of air conditioner load prediction, in particular to an air conditioner load prediction model establishment method and system and a storage medium. The method comprises the following steps: acquiring indoor sensing data and outdoor meteorological data, and performing data fusion to obtain a multi-source sensing data set; mode recognition is conducted based on the multi-source sensing data set, and resident air conditioner use behavior mode data are obtained; building a resident personalized load prediction model based on the resident air conditioner use behavior mode data; according to the resident personalized load prediction model, environment load correlation adjustment is conducted on the multi-source sensing data set, and an indoor air conditioner load prediction model is obtained; and real-time outdoor meteorological data and real-time resident interaction data are obtained, air conditioner equipment operation strategy self-adaptive optimization is carried out, and therefore an air conditioner equipment optimized operation strategy is obtained and uploaded to the air conditioner equipment management cloud platform to execute an air conditioner equipment control task. The method can effectively improve the precision of air conditioner load prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of air-conditioning load prediction, and particularly to a method, a system and a storage medium for establishing an air-conditioning load prediction model. Background Art

[0002] With the rapid development of social economy and the improvement of residents' living standards, air conditioners have become an indispensable and important equipment in modern families and commercial buildings. While providing a comfortable indoor environment, the air-conditioning system also consumes a large amount of electric power resources. Especially during the high-temperature period in summer, the increase in air-conditioning load will exert great pressure on the power system. Therefore, accurately predicting the air-conditioning load has become an important task for energy conservation and emission reduction, ensuring stable power supply and optimizing power dispatching. Most of the early air-conditioning load prediction models adopted linear regression or simple time series methods, which usually relied only on single meteorological variables such as temperature. However, the change of air-conditioning load is not only affected by temperature, but also closely related to multiple factors such as humidity, wind speed, solar radiation and the heat load of buildings. Overly simplified models cannot comprehensively consider these multi-dimensional factors, resulting in large deviations in prediction results. Traditional prediction methods often ignore the influence of household behavior on air-conditioning load. Factors such as household air-conditioning usage habits, the number of people and the configuration of household equipment have significant differences, which makes the air-conditioning loads of different families or buildings may vary greatly even under similar climatic conditions. Traditional models do not take these personalized factors into account in the prediction, thus reducing the prediction accuracy. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a method, a system and a storage medium for establishing an air-conditioning load prediction model to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for establishing an air-conditioning load prediction model includes the following steps:

[0005] Step S1: Obtain indoor sensing data and outdoor meteorological data, and perform indoor sensing data fusion on the indoor sensing data to obtain indoor fused sensing data; perform time series merging on the indoor fused sensing data and the outdoor meteorological data to obtain a multi-source sensing data set;

[0006] Step S2: Perform indoor temperature and humidity pattern analysis based on the multi-source sensing data set to obtain indoor temperature and humidity pattern data; perform household air-conditioning usage behavior pattern recognition according to the indoor temperature pattern data to obtain household air-conditioning usage behavior pattern data;

[0007] Step S3: Quantify the influence of household usage behavior on the household air-conditioning usage behavior pattern data to obtain household usage behavior influence amount data, and construct a household personalized load prediction model based on the household usage behavior influence amount data;

[0008] Step S4: Perform environment-load correlation modeling on the multi-source sensing data set according to the household personalized load prediction model to obtain an environment-load correlation matrix, and adjust the household personalized load prediction model according to the environment-load correlation matrix to obtain an indoor air-conditioning load prediction model;

[0009] Step S5: Obtain real-time outdoor meteorological data and real-time household interaction data, and perform air-conditioning load prediction on the real-time outdoor meteorological data and real-time household interaction data through the indoor air-conditioning load prediction model to obtain indoor air-conditioning load prediction data; perform adaptive optimization of the air-conditioning equipment operation strategy based on the indoor air-conditioning load prediction data to obtain an optimized operation strategy for the air-conditioning equipment, and upload it to the air-conditioning equipment management cloud platform to execute the air-conditioning equipment control task.

[0010] By obtaining and fusing indoor sensing data and outdoor meteorological data, the present invention can comprehensively consider multi-dimensional influencing factors, thereby providing more accurate basic data for subsequent air-conditioning load prediction. The fusion of indoor sensing data can eliminate the errors of a single sensor, enhance the reliability of the data, and further improve the stability of model prediction. The time-series merging of outdoor meteorological data and indoor sensing data ensures the consistency of the data in terms of time, helps capture real-time changing environmental factors, and further improves the effectiveness of the multi-source sensing data set. Conducting indoor temperature and humidity pattern analysis based on the multi-source sensing data set can reflect the changing trends of the indoor environment from multiple perspectives, providing data support for identifying the air-conditioning usage behavior patterns of residents. The changes in indoor temperature patterns are closely related to the air-conditioning usage behavior of residents. Through the analysis of these patterns, the temperature control requirements and usage habits of residents can be accurately grasped, providing the necessary inputs for the construction of a personalized load prediction model. The identification of residents' air-conditioning usage behavior patterns not only depends on environmental factors but also fully considers the unique needs of each family, thus being more in line with the actual usage situation and avoiding the neglect of residents' differences in traditional methods. The quantification of residents' usage behavior further improves the accuracy of personalized load prediction. By extracting key factors such as the air-conditioning usage frequency, load period duration, and temperature demand pattern of residents, the changing rules of air-conditioning loads of residents under different environmental conditions can be clearly described. These quantified data help predict air-conditioning loads more accurately and provide more personalized and targeted input information for the load prediction model. The personalized load prediction model constructed based on these quantification results can significantly improve the prediction accuracy, avoid the defect of neglecting individual differences in traditional models, and accurately predict the air-conditioning load requirements of each family or building under different environments and times. Environment-load correlation modeling enables the prediction model to be dynamically synchronized with changes in the external environment. By obtaining the environment-load correlation matrix and adjusting the personalized load prediction model according to environmental changes, the model can better reflect the complex relationship between the environment and air-conditioning loads. Such adjustment can make the model more flexible and adaptable, enhancing its stability and effectiveness in practical applications. In addition, this adjustment process can consider the impact of outdoor climate change on indoor air-conditioning loads in real time, thereby optimizing the accuracy and timeliness of air-conditioning load prediction. By analyzing real-time meteorological data and residents' interaction data and combining with the air-conditioning load prediction model, the operating strategy of air-conditioning equipment can be adjusted in real time. The optimized strategy will help improve the energy efficiency of air-conditioning, avoid energy waste, and ensure the comfort of the indoor environment. For example, by dynamically adjusting the working mode and operating time of the air-conditioning, the peak load can be reduced, and the efficiency during low-load periods can be improved, achieving the goal of energy conservation and emission reduction. After uploading these optimized strategies to the air-conditioning equipment management cloud platform, unified control of multiple air-conditioning equipment can be achieved, ensuring the efficient operation and real-time feedback of the system, and further improving the overall energy management level and the accuracy of load prediction.Overall, this technical solution for air-conditioning load prediction based on multi-source data and personalized analysis can effectively improve the accuracy of load prediction, optimize the operation strategy of air-conditioning equipment, reduce energy consumption, enhance comfort, and at the same time ensure the stability and energy-saving effect of the power system, with strong practical application value and market potential.

[0011] Optionally, step S1 is specifically as follows:

[0012] Step S11: Obtain indoor sensing data and outdoor meteorological data, and respectively perform data preprocessing on the indoor sensing data and outdoor meteorological data to obtain the indoor sensing data to be analyzed and the outdoor meteorological data to be analyzed;

[0013] Step S12: Construct an indoor coordinate system based on the indoor sensing data to be analyzed to obtain three-dimensional coordinate system data of the indoor space;

[0014] Step S13: Calibrate the positions of sensors for the indoor sensing data to be analyzed according to the three-dimensional coordinate system data of the indoor space to obtain an indoor sensor position calibration acquisition data set;

[0015] Step S14: Perform Kalman filtering on the indoor sensing data based on the indoor sensor position calibration acquisition data set to merge the indoor sensing data and obtain indoor fused sensing data;

[0016] Step S15: Merge the timestamp data according to the indoor fused sensing data and the outdoor meteorological data to be analyzed to obtain a multi-source sensing data set.

[0017] The present invention acquires indoor sensing data and outdoor meteorological data and performs preprocessing, which can convert the collected raw data into a standardized data format, thereby reducing data noise and eliminating inconsistencies, providing a more reliable input for subsequent analysis. This step ensures the high quality of the data and lays a solid foundation for subsequent in-depth analysis. A three-dimensional coordinate system for the indoor space is constructed. Through this spatial positioning method, clear physical coordinate information can be provided for the measurement data of each sensor, which not only improves the accuracy of the data but also provides important support for subsequent spatial analysis and data fusion. The position calibration of the indoor sensors further precisifies the spatial distribution of the sensing data. By applying the Kalman filtering algorithm, the data collected by different sensors can be effectively fused, reducing the deviation caused by sensor errors, making the finally obtained indoor fused sensing data more accurate and stable. This process utilizes the advantages of the Kalman filter to dynamically update and correct the data, further improving the accuracy of the fused data. The indoor sensing data merged by the Kalman filter is merged with the external meteorological data by timestamp, enabling the change data of the indoor and outdoor environments to be consistent in the time dimension. This time-series merging method not only ensures the synchronization of the data but also enhances the response ability to dynamic environmental changes, making the multi-source sensing data set more timely and relevant. This data set not only fuses the changes in the indoor environment but also combines the real-time changes in the external climate, providing comprehensive and accurate input data for the subsequent air-conditioning load prediction model.

[0018] Optionally, step S2 is specifically as follows:

[0019] Step S21: Extract indoor temperature features and indoor humidity features from the multi-source sensing data set to obtain indoor temperature data and indoor humidity data;

[0020] Step S22: Analyze the change trends of the indoor temperature data and the indoor humidity data in time windows respectively to obtain time-window temperature trend data and time-window humidity trend data;

[0021] Step S23: Identify the indoor temperature pattern based on the time-window temperature trend data to obtain indoor temperature pattern data; identify the indoor humidity pattern based on the time-window humidity trend data to obtain indoor humidity pattern data;

[0022] Step S24: Correlate the change patterns of the indoor temperature pattern data and the indoor humidity pattern data to obtain indoor temperature and humidity pattern data;

[0023] Step S25: Identify the household air-conditioning usage behavior pattern based on the indoor temperature and humidity pattern data to obtain household air-conditioning usage behavior pattern data.

[0024] Through the extraction of indoor temperature and humidity characteristics from multi-source sensing data sets, the present invention can effectively extract the most critical temperature and humidity information from complex environmental data, which is crucial for air-conditioning load prediction. The accurate acquisition of temperature and humidity data lays a foundation for subsequent pattern analysis, enabling the indoor environmental changes to be more clearly reflected. This process improves the usability of the data, ensures its representativeness, and reduces the interference of irrelevant information. After analyzing the changing trends of temperature and humidity data in time windows, the trend information of environmental changes can be better captured. This trend analysis method can extract the dynamic change rules of temperature and humidity according to the changes in different time windows. This process can discover the short-term and long-term change trends of the indoor environment, thereby providing a strong trend basis for the prediction of air-conditioning load and making a more accurate judgment on the usage demand of household air conditioners. Based on the temperature and humidity trend data in time windows, the identification of indoor temperature and humidity patterns can be carried out to deeply understand the typical change patterns of the indoor environment. This pattern recognition provides a specific reference basis for the operation strategy of air conditioners, thus enabling the better prediction of the change of air-conditioning load. For example, by identifying the change patterns of temperature and humidity, the possible air-conditioning usage demands of residents can be estimated, so as to conduct load prediction and scheduling arrangements in advance, reduce energy waste and improve comfort. Further, by correlating the change patterns of indoor temperature patterns and humidity patterns, the mutual influence between temperature and humidity and their combined effect on air-conditioning load can be revealed. The correlation analysis of temperature and humidity enables the prediction model to consider the synergistic effect of these two factors under different environmental conditions, thus making the air-conditioning load prediction more comprehensive and accurate. This method helps to capture the change rules of air-conditioning usage under different patterns, improving the reliability and effectiveness of the prediction. Based on the indoor temperature and humidity pattern data, the identification of the air-conditioning usage behavior patterns of residents can be further carried out to identify the personalized air-conditioning usage behaviors of residents, such as the frequency, time period, and usage habits of air-conditioning use. By identifying these behavior patterns, more accurate air-conditioning load prediction can be achieved, and more personalized air-conditioning usage strategies can be formulated according to the actual needs and preferences of residents. This not only improves the energy efficiency of the air-conditioning system but also enhances the comfort experience of residents, effectively promoting the realization of the energy conservation and emission reduction goals.

[0025] Optionally, step S25 is specifically as follows:

[0026] Step S251: Obtain the operation data of the household air conditioner, and perform data preprocessing on the operation data of the household air conditioner to obtain the operation data of the household air conditioner to be analyzed;

[0027] Step S252: Calculate the operation load of the air conditioner for the operation data of the household air conditioner to be analyzed to obtain the operation load data of the air conditioner, and perform time window load division on the operation load data of the air conditioner to obtain high-load time window data and low-load time window data;

[0028] Step S253: Extract the household set temperature characteristics from the household air conditioner operation data to be analyzed, so as to obtain the household set temperature data, and perform a temperature difference time window statistics on the household set temperature data and the outdoor meteorological data to be analyzed, so as to obtain the indoor high temperature difference time window data and the indoor low temperature difference time window data;

[0029] Step S254: Perform a time window intersection operation on the indoor high temperature difference time window data and the high load time window data, so as to obtain the air conditioner high load usage time window data; perform a time window intersection operation on the indoor low temperature difference time window data and the low load time window data, so as to obtain the air conditioner low load usage time window data;

[0030] Step S255: Identify the peak operation mode of the air conditioner for the household air conditioner operation data to be analyzed according to the air conditioner high load usage time window data, so as to obtain the peak operation mode data of the air conditioner; identify the idle operation mode of the air conditioner for the household air conditioner operation data to be analyzed according to the air conditioner low load usage time window data, so as to obtain the idle operation mode data of the air conditioner;

[0031] Step S256: Integrate the peak operation mode data of the air conditioner and the idle operation mode data of the air conditioner to obtain the air conditioner usage mode data, and identify the household air conditioner usage behavior mode according to the indoor temperature and humidity mode data and the air conditioner usage mode data, so as to obtain the household air conditioner usage behavior mode data.

[0032] By removing noise and filling in missing values, the present invention makes the analysis more accurate, ensuring that the input data of the model has a high degree of accuracy and consistency. Immediately afterwards, the calculation of the air-conditioning operation load and the division of the load in the time window can effectively identify the high-load and low-load periods of the air-conditioning load, providing a detailed division of the air-conditioning load. This lays the foundation for the subsequent identification of the air-conditioning usage behavior pattern, helps to identify the characteristics of the air-conditioning usage of the householders at different time periods, and provides a basis for energy-saving optimization. By extracting the characteristics of the temperature set by the householders and conducting a temperature difference time window statistics with the outdoor meteorological data, the impact of the temperature difference on the air-conditioning usage can be analyzed in depth. This analysis helps to reveal the relationship between the temperature set by the householders and the actual outdoor air temperature, and then infer the air-conditioning usage behavior pattern of the householders. For example, when the temperature difference is large, the householders may be more inclined to turn on the air-conditioning, which provides important reference data for the air-conditioning load prediction and operation scheduling. The time window intersection operation further precisifies the definition of the high-load and low-load usage times of the air-conditioning, thus helping to clarify the usage period of the air-conditioning. By matching the high-load and low-load periods, the peak and trough periods of the air-conditioning can be effectively identified, providing refined data support for optimizing the energy scheduling and load management. By analyzing the data in the high-load usage time window, the peak operation mode of the air-conditioning usage can be identified, and then it provides an important basis for optimizing the air-conditioning operation efficiency and reducing energy waste. At the same time, the identification of the low-load usage time window helps to understand the idle operation mode of the air-conditioning, provides a direction for optimizing the idle management of the air-conditioning equipment, and reduces unnecessary power consumption. The integration of the air-conditioning usage patterns helps to further improve the intelligence of the air-conditioning operation strategy. By fusing the data of the peak and idle operation modes, a complete air-conditioning usage pattern data is formed, and then it provides support for the identification of the air-conditioning usage behavior pattern of the householders. By combining the indoor temperature and humidity pattern data, the air-conditioning usage habits, demands and preferences of the householders can be more accurately reflected, providing more personalized characteristic information for predicting the air-conditioning load of the householders. This can not only improve the accuracy of the air-conditioning usage, but also automatically adjust the air-conditioning operation strategy according to the demands of the householders, and then achieve the dual goals of energy conservation and improved comfort.

[0033] Optionally, step S256 is specifically as follows:

[0034] Integrate the air-conditioning peak operation mode data and the air-conditioning idle operation mode data to obtain the air-conditioning usage mode data;

[0035] Extract the air-conditioning usage behavior characteristics from the air-conditioning usage mode data to obtain the air-conditioning switch frequency data and the load period duration data;

[0036] Extract the indoor temperature and humidity fluctuation mode according to the indoor temperature and humidity pattern data to obtain the indoor temperature and humidity fluctuation mode data;

[0037] Analyze the household temperature demand based on the air conditioner switching frequency data and the indoor temperature and humidity fluctuation pattern data, so as to obtain the household temperature demand pattern data;

[0038] Analyze the household air conditioner usage preference based on the load period duration data and the indoor temperature and humidity fluctuation pattern data, so as to obtain the household air conditioner usage preference data;

[0039] Identify the user air conditioner usage behavior pattern based on the household air conditioner usage preference data and the household temperature demand pattern data, so as to obtain the household air conditioner usage behavior pattern data.

[0040] The present invention can comprehensively master the operation rules of the air conditioner in different usage states by integrating the peak operation mode data and the idle operation mode data of the air conditioner, and then generate a comprehensive air conditioner usage mode data. This integration can not only reveal the fluctuation of the air conditioner load, but also effectively divide the peak and valley usage periods of the air conditioner, providing a clear direction for subsequent optimization. Extracting the behavioral characteristics in the air conditioner usage mode data, such as the air conditioner switching frequency data and the load period duration data, helps to more deeply understand the household air conditioner usage habits. These characteristics provide basic data for analyzing the peak and idle periods of the air conditioner load, and at the same time help to predict future load changes, facilitating the implementation of targeted energy-saving measures. By analyzing the household temperature demand based on the air conditioner switching frequency data and the indoor temperature and humidity fluctuation pattern data, the household temperature demand pattern data can be obtained. This step helps to more accurately predict the air conditioner demand of the household under different climate conditions, thereby optimizing the air conditioner usage strategy. This not only improves the comfort level but also effectively avoids energy waste. By combining the load period duration data and the indoor temperature and humidity fluctuation pattern data, the household air conditioner usage preference can be analyzed, so as to better capture the air conditioner usage characteristics of the household at different time periods. The household air conditioner usage preference directly affects the distribution of the air conditioner load. Analyzing these preference data helps to improve the accuracy of the air conditioner load prediction. Identifying the air conditioner usage behavior pattern based on the household air conditioner usage preference data and the temperature demand pattern data can not only accurately grasp the personalized needs of the household, but also provide personalized data support for subsequent load prediction and intelligent scheduling. By identifying and analyzing these patterns, a more efficient air conditioner operation strategy can be formulated, improving the energy efficiency of the system and providing a more comfortable and customized air conditioner usage experience for the household.

[0041] Optionally, step S3 is specifically as follows:

[0042] Step S31: Extract the influencing factors of the household usage behavior from the household air conditioner usage behavior pattern data, so as to obtain the air conditioner usage frequency factor, the load period duration factor, the temperature demand pattern factor, and the air conditioner usage preference factor;

[0043] Step S32: Perform associated factor clustering on the air conditioner usage frequency factor, load period duration factor, temperature demand pattern factor, and air conditioner usage preference factor, so as to obtain a set of household usage behavior impact factors;

[0044] Step S33: Perform factor impact weight statistics on the set of household usage behavior impact factors according to the air conditioner operation load data, so as to obtain factor impact weight data;

[0045] Step S34: Quantify the household behavior impact factors based on the factor impact weight data and the set of household usage behavior impact factors, so as to obtain household usage behavior impact quantity data;

[0046] Step S35: Construct a household personalized load prediction model based on the household usage behavior impact quantity data.

[0047] Through the extraction of household usage behavior impact factors from the household air conditioner usage behavior pattern data, the present invention can identify key factors related to air conditioner usage, such as air conditioner usage frequency, load period duration, temperature demand pattern, and air conditioner usage preference. The extraction of these factors lays a foundation for subsequent behavior pattern analysis, and can more accurately grasp the core characteristics of household air conditioner usage, providing necessary data support for personalized load prediction. Performing associated factor clustering on these impact factors helps to find the internal relationships between different factors, and establish a comprehensive set of impact factors for each household's air conditioner usage behavior. This clustering process enables the quantification of the interaction between different factors, and provides a basis for data integration for subsequent analysis, facilitating the overall understanding of the regularity of household behavior. Performing factor impact weight statistics on the set of household usage behavior impact factors according to the air conditioner operation load data can reveal the relative importance of each factor in air conditioner load prediction. Assigning weights to factors enables different factors to be accurately quantified in the personalized load prediction model, avoiding overemphasis or neglect of certain factors, and improving the accuracy and practicality of prediction. The step of quantifying the behavior impact factors based on the factor impact weight data and the set of household usage behavior impact factors further deepens the actual impact of the factors. It can provide a precise dataset of usage behavior impact quantities for each household. Through the analysis of these quantified data, the change trend of household air conditioner usage behavior can be predicted, and accurate input can be provided for the load prediction model. Constructing a personalized load prediction model based on the household usage behavior impact quantity data enables the air conditioner load prediction to not only reflect the changes in meteorological conditions, but also fully consider the personalized needs and usage habits of households. Through this personalized prediction, the air conditioner system can more accurately predict load changes, thereby achieving more efficient and energy-saving air conditioner management.

[0048] Optionally, step S4 is specifically as follows:

[0049] Step S41: Extract environmental sensing time series features from multi-source sensing data to obtain environmental sensing time series data;

[0050] Step S42: Perform air-conditioning load prediction on the environmental sensing time series data through the household personalized load prediction model to obtain air-conditioning load prediction data;

[0051] Step S43: Calculate the model prediction error of the air-conditioning operation load data and the air-conditioning load prediction data to obtain the model load error data;

[0052] Step S44: Calculate the environmental variable-air conditioning load correlation based on the model load error data and the environmental sensing time series data to obtain the environmental variable-air conditioning load correlation data;

[0053] Step S45: Perform correlation matrix conversion on the environmental variable-air conditioning load correlation data to obtain the environmental load association matrix;

[0054] Step S46: Combine the correlation coefficients of the environmental load association matrix and the household usage behavior impact factor set to obtain the dynamic adjustment factor;

[0055] Step S47: Iteratively adjust the model parameters of the household personalized load prediction model according to the dynamic adjustment factor to obtain the indoor air-conditioning load prediction model.

[0056] The present invention extracts the environmental sensing time series features from multi-source sensing data, which can accurately capture the temporal changes of environmental variables and provide more detailed environmental data analysis. This feature extraction not only helps to identify the changing trends of environmental variables, but also provides the basic data support for subsequent load prediction, enabling the air-conditioning load prediction to consider more dynamic factors in the time series and enhancing the timeliness and accuracy of the prediction. Using the personalized load prediction model of residents to predict the air-conditioning load for the environmental sensing time series data can combine the environmental data with the behavior patterns of residents, thereby generating load prediction data based on environmental changes. This step enables the air-conditioning load prediction to reflect the actual impact of environmental changes on the air-conditioning load and further improves the personalization and accuracy of the prediction. By calculating the model prediction error between the air-conditioning operation load data and the air-conditioning load prediction data, the error magnitude of the prediction model can be measured and feedback can be provided for subsequent optimization. This error calculation can reveal the deviation of the current prediction model in practical applications, thereby providing a basis for the improvement and optimization of the model to ensure that the prediction results are more in line with the actual operating conditions. Based on the model load error data and the environmental sensing time series data, the correlation calculation between the environmental variables and the air-conditioning load can be carried out to discover the internal relationship between the environmental variables and the air-conditioning load. This step enables the prediction model to identify which environmental factors have a greater impact on the change of the air-conditioning load, thereby providing key information for subsequent adjustment and optimization. By converting the environmental variable-air-conditioning load correlation data into a correlation matrix, the complex environmental load relationship can be presented in matrix form, making these data more systematic and facilitating subsequent analysis and adjustment. This matrix can reveal the influence degree of different environmental factors on the air-conditioning load and become the basis for further optimizing the model. Combining the environmental load correlation matrix with the set of influence factors of residents' usage behavior to perform dynamic factor combination of correlation coefficients can comprehensively consider the influence of environmental factors and residents' behavior patterns and provide a more accurate dynamic adjustment factor. The combination of this factor will further improve the accuracy of the load prediction model, enabling it to adapt to different environmental conditions and residents' behaviors and achieving a higher prediction accuracy. According to the dynamic adjustment factor, iterative adjustment of the model parameters of the personalized load prediction model of residents is carried out to continuously optimize the air-conditioning load prediction model and enhance its adaptability to environmental and behavioral factors. This iterative adjustment process ensures that the load prediction model can be continuously improved with the changes of the environment and residents' behaviors, thereby realizing a more accurate and dynamic air-conditioning load prediction and ultimately optimizing the operation and energy-saving effect of air-conditioning equipment.

[0057] Optionally, step S5 is specifically as follows:

[0058] Step S51: Obtain the real-time outdoor meteorological data and the real-time resident interaction data, and calculate the indoor temperature difference between the real-time meteorological data and the multi-source sensing data, so as to obtain the indoor temperature difference data;

[0059] Step S52: Extract the household behavior characteristics from the real-time household interaction data to obtain the real-time household behavior characteristic data;

[0060] Step S53: Perform air-conditioning load prediction on the indoor temperature difference data and the real-time household behavior characteristic data through the indoor air-conditioning load prediction model to obtain the indoor air-conditioning load prediction data;

[0061] Step S54: Identify the household air-conditioning usage pattern for the indoor air-conditioning load prediction data based on the household air-conditioning usage behavior pattern data to obtain the real-time household air-conditioning usage pattern data;

[0062] Step S55: Obtain the real-time indoor air-conditioning operation mode data, and select the air-conditioning operation strategy for the real-time indoor air-conditioning operation mode data according to the real-time household air-conditioning usage pattern data to obtain the air-conditioning equipment operation strategy;

[0063] Step S56: Perform adaptive strategy optimization on the air-conditioning equipment operation strategy to obtain the optimized air-conditioning equipment operation strategy, and upload it to the air-conditioning equipment management cloud platform to execute the air-conditioning equipment control task.

[0064] The present invention calculates the indoor temperature difference by obtaining real-time outdoor meteorological data and real-time household interaction data and combining multi-source sensing data, which can reflect the indoor-outdoor temperature difference in real time. This calculation provides a basis for real-time temperature changes in air-conditioning load prediction, helping the model to make more accurate load predictions under different meteorological conditions, thereby optimizing the efficiency and comfort of air-conditioning operation. Extracting household behavior characteristics from real-time household interaction data enables the air-conditioning load prediction model to not only rely on environmental factors but also consider the actual behavior patterns of households, such as the frequency of turning on and off the air conditioner and the usage time period. The extraction of these behavior characteristics further refines the prediction model, enabling it to adapt to the personalized needs of different households and improve the accuracy and adaptability of the prediction. Using the indoor air-conditioning load prediction model to predict the air-conditioning load based on the temperature difference data and household behavior characteristic data enables the model to comprehensively consider the dual impacts of external environmental changes and household behavior on the air-conditioning load. This multi-dimensional prediction can predict the air-conditioning load in real time, ensuring that the air-conditioning system always operates at the best working state, thereby avoiding energy waste and improving the energy-saving effect. Through the identification of the household air-conditioning usage behavior pattern, the actual air-conditioning usage pattern of the household can be further refined according to the air-conditioning load prediction result. This identification process helps to predict the air-conditioning demand of the household during a specific period, provides a more accurate air-conditioning usage pattern identification, and helps the model to optimize its personalized load prediction ability. According to the real-time air-conditioning usage pattern data of the household, an air-conditioning operation strategy is selected for the real-time indoor air-conditioning operation mode data, enabling the air-conditioning equipment to make intelligent decisions according to the real-time air-conditioning usage pattern, thereby achieving more flexible and accurate control. This strategy selection process takes into account the changing needs of the household and the operating state of the air-conditioning equipment, helping to avoid unnecessary energy consumption and ensure the stable operation of the air-conditioning system. Adaptive strategy optimization of the air-conditioning equipment operation strategy can continuously adjust and improve the air-conditioning operation strategy, enabling it to adapt to changes in the environment and household behavior in real time. This optimization process can improve the operating efficiency of the air-conditioning, reduce energy consumption, and enhance comfort. Finally, by uploading to the air-conditioning equipment management cloud platform to execute control tasks, it ensures that the system operates according to the optimal strategy, thereby achieving more intelligent air-conditioning equipment management.

[0065] Optionally, the present specification also provides an air-conditioning load prediction model establishment system for performing the air-conditioning load prediction model establishment method described above. The air-conditioning load prediction model establishment system includes:

[0066] A sensing data fusion module for obtaining indoor sensing data and outdoor meteorological data, performing indoor sensing data fusion on the indoor sensing data to obtain indoor fused sensing data; and performing time-series merging on the indoor fused sensing data and the outdoor meteorological data to obtain a multi-source sensing data set;

[0067] A usage behavior pattern recognition module is used to perform indoor temperature and humidity pattern analysis based on a multi-source sensing data set to obtain indoor temperature and humidity pattern data; and perform household air conditioner usage behavior pattern recognition based on the indoor temperature pattern data to obtain household air conditioner usage behavior pattern data.

[0068] A prediction model construction module is used to quantify the influence of household usage behavior on the household air conditioner usage behavior pattern data to obtain household usage behavior influence amount data, and construct a household personalized load prediction model based on the household usage behavior influence amount data.

[0069] An environmental load adjustment module is used to perform environment-load correlation modeling on the multi-source sensing data set according to the household personalized load prediction model to obtain an environment load correlation matrix, and perform environment load correlation adjustment on the household personalized load prediction model according to the environment load correlation matrix to obtain an indoor air conditioner load prediction model.

[0070] An operation strategy optimization module is used to obtain real-time outdoor meteorological data and real-time household interaction data, and perform air conditioner load prediction on the real-time outdoor meteorological data and real-time household interaction data through the indoor air conditioner load prediction model to obtain indoor air conditioner load prediction data; perform adaptive optimization of the air conditioner equipment operation strategy based on the indoor air conditioner load prediction data to obtain an optimized operation strategy for the air conditioner equipment, and upload it to the air conditioner equipment management cloud platform to execute the air conditioner equipment control task.

[0071] Optionally, this specification also provides a storage medium storing a computer program, and when the computer program is executed by a processor, it implements the air conditioner load prediction model establishment method as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Other features, objects, and advantages of the present invention will become more apparent by reading the detailed description of the non-limiting embodiments with reference to the following drawings.

[0073] Figure 1 It is a schematic flow chart of the steps of the air conditioner load prediction model establishment method of the present invention.

[0074] Figure 2 It is a detailed schematic flow chart of step S1 in the present invention.

[0075] Figure 3 It is a detailed schematic flow chart of step S2 in the present invention.

[0076] The realization, functional characteristics, and advantages of the object of the present invention will be further described with reference to the embodiments and the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] The technical method of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative work belong to the scope of protection of the present invention.

[0078] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0079] It should be understood that although the terms "first", "second", etc. may be used here to describe each unit, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed associated items.

[0080] To achieve the above object, please refer to Figures 1 to 3 , the present invention provides a method for establishing an air-conditioning load prediction model, and the method includes the following steps:

[0081] Step S1: Obtain indoor sensing data and outdoor meteorological data, and perform indoor sensing data fusion on the indoor sensing data to obtain indoor fused sensing data; perform time-series merging on the indoor fused sensing data and the outdoor meteorological data to obtain a multi-source sensing data set;

[0082] In this embodiment, when obtaining indoor sensing data and outdoor meteorological data, multiple sensors (such as temperature, humidity, light, and air quality sensors) are used to comprehensively monitor the indoor environment, and data such as outdoor temperature, humidity, and wind speed are obtained from a weather station or a weather API. To enhance the accuracy of indoor data, a data fusion algorithm (such as the weighted average method or the Kalman filter algorithm) is used to fuse the indoor sensing data to obtain more accurate indoor environment data. Then, the indoor fusion data and the outdoor meteorological data are merged in time series using timestamps to form a multi-source sensing data set. This data set contains indoor data (such as temperature, humidity, etc.) collected by different sensors and external environmental factors (such as outdoor temperature, humidity, wind speed, etc.). The timestamps ensure the timeliness and accuracy of the data, providing basic data support for subsequent analysis.

[0083] Step S2: Perform indoor temperature and humidity pattern analysis based on the multi-source sensing data set to obtain indoor temperature and humidity pattern data; identify the household air conditioner usage behavior pattern based on the indoor temperature pattern data to obtain the household air conditioner usage behavior pattern data.

[0084] In this embodiment, when analyzing the multi-source sensing data set, first, based on the indoor temperature and humidity data, a time series analysis method (such as moving average or Fourier transform) is used to identify the indoor temperature and humidity change patterns and extract the key indoor temperature and humidity pattern data. These data can reflect the indoor temperature and humidity fluctuations in different time periods. Then, based on these indoor temperature pattern data, a machine learning algorithm (such as support vector machine or decision tree) is used to identify the air conditioner usage behavior pattern of the household, such as the time, frequency, and temperature setting when the air conditioner is turned on. Through these data, the air conditioner usage behavior pattern of the household can be determined, providing data support for subsequent personalized load prediction.

[0085] Step S3: Quantify the impact of the household air conditioner usage behavior pattern data on the household usage behavior to obtain the household usage behavior impact data, and construct a household personalized load prediction model based on the household usage behavior impact data.

[0086] In this embodiment, based on the data of the household air conditioner usage behavior patterns, first, through statistical analysis methods (such as weighted average or regression analysis), the influencing factors of the household air conditioner usage behavior are extracted, specifically including the air conditioner usage frequency factor (such as the number of on / off times per hour), the load period duration factor (such as the duration of each air conditioner startup), the temperature demand pattern factor (such as the set temperature range), and the air conditioner usage preference factor (such as the preferred air conditioner mode). Through these data, the impact of household behavior on the air conditioner load is further quantified. For example, some households frequently turn on the air conditioner in hot weather, and some households are more inclined to set a low-temperature environment. Based on these impact data, a personalized load prediction model is constructed using machine learning algorithms (such as neural networks or random forests) to accurately predict the air conditioner load requirements of different households under different environmental conditions.

[0087] Step S4: Perform environment-load association modeling on the multi-source sensing data set according to the household personalized load prediction model to obtain an environment-load association matrix, and perform environment-load association adjustment on the household personalized load prediction model according to the environment-load association matrix to obtain an indoor air conditioner load prediction model;

[0088] In this embodiment, when making predictions based on the household personalized load prediction model, first, the multi-source sensing data set is combined with the household air conditioner usage behavior data for environment-load association modeling. By analyzing the relationship between indoor and outdoor environmental variables (such as indoor temperature and humidity and outdoor climate conditions) and the air conditioner load, an environment-load association matrix is established, which can represent the influence degree of various environmental factors (such as temperature, humidity, etc.) on the air conditioner load. Then, the environment-load association matrix is used to perform environment-load association adjustment on the personalized load prediction model, so that the model can more accurately reflect the impact of external environmental changes on the household air conditioner load. Finally, an updated indoor air conditioner load prediction model is obtained, which can adjust the prediction results in real time according to environmental changes and household behavior.

[0089] Step S5: Obtain real-time outdoor meteorological data and real-time household interaction data, and perform air conditioner load prediction on the real-time outdoor meteorological data and real-time household interaction data through the indoor air conditioner load prediction model to obtain indoor air conditioner load prediction data; perform adaptive optimization of the air conditioner equipment operation strategy based on the indoor air conditioner load prediction data to obtain an optimized operation strategy for the air conditioner equipment, and upload it to the air conditioner equipment management cloud platform to execute the air conditioner equipment control task.

[0090] In this embodiment, after obtaining the real-time outdoor meteorological data and real-time household interaction data, first, an indoor air-conditioning load prediction model is used to predict the air-conditioning load based on the real-time outdoor meteorological data (such as the current outdoor temperature, humidity, wind speed, etc.) and real-time household interaction data (such as the user's behavior of controlling the air conditioner through a smart device). The model comprehensively considers the current climate conditions, the real-time needs of the household, and their air-conditioning usage behavior, and outputs the predicted indoor air-conditioning load data. Based on these data, further adaptive optimization of the air-conditioning equipment operation strategy is carried out. An optimization algorithm (such as a genetic algorithm or a particle swarm algorithm) is used to adjust the operation strategy of the air conditioner, such as adjusting the air-conditioning startup time, temperature setting, etc., to maximize energy efficiency. The optimized strategy will be uploaded to the air-conditioning equipment management cloud platform for the air-conditioning control system to execute, ensuring that the air-conditioning equipment operates according to the optimal strategy, thereby effectively improving energy efficiency and reducing energy consumption.

[0091] Optionally, step S1 is specifically as follows:

[0092] Step S11: Obtain indoor sensing data and outdoor meteorological data, and perform data preprocessing on the indoor sensing data and outdoor meteorological data respectively, so as to obtain the indoor sensing data to be analyzed and the outdoor meteorological data to be analyzed;

[0093] In this embodiment, a variety of sensors (such as temperature and humidity sensors, light sensors, CO2 concentration sensors, etc.) installed at different positions indoors are used to collect indoor environmental data. The outdoor meteorological data is obtained through a weather station or an online meteorological API (such as OpenWeatherMap), covering information such as outdoor temperature, humidity, wind speed, and radiation intensity. Both the indoor sensing data and the outdoor meteorological data are initially preprocessed. For indoor data, outliers are processed through denoising and smoothing (such as the moving average method); for outdoor data, interpolation methods are used to fill in missing data. The finally obtained indoor sensing data to be analyzed includes the time series data of each sensor, and the outdoor meteorological data to be analyzed contains the meteorological information corresponding to the indoor data, ensuring the integrity and timeliness of the data.

[0094] Step S12: Construct an indoor coordinate system based on the indoor sensing data to be analyzed, so as to obtain the three-dimensional coordinate system data of the indoor space;

[0095] In this embodiment, based on the indoor sensing data to be analyzed, a known building floor plan and the installation position data of the sensors are used to construct an indoor coordinate system based on the relative positions of the sensors. Assume that the indoor space is a three-dimensional coordinate system, where the X-axis and Y-axis represent the horizontal plane, and the Z-axis represents the vertical direction. By measuring the indoor building layout, the specific positions of each sensor are determined. According to the position data of the sensors, an accurate three-dimensional coordinate system of the indoor space is established through geometric transformation, ensuring that the data of each sensor can be accurately mapped to the corresponding physical position, helping to further analyze the spatial distribution of the indoor environment.

[0096] Step S13: Calibrate the positions of the indoor sensors based on the three-dimensional coordinate system data of the indoor space, so as to obtain an indoor sensor position calibration acquisition data set;

[0097] In this embodiment, based on the three-dimensional coordinate system of the indoor space, by calibrating the positions of the sensors in the physical space, it is ensured that the data collected by each sensor can accurately correspond to the actual spatial positions. Use high-precision positioning technologies (such as laser rangefinders or 3D scanners) to accurately measure the installation positions of each sensor to obtain the actual coordinate positions of each sensor. Then, by comparing the differences between the theoretical positions and the actual positions of the sensors, position calibration adjustments are made to ensure that the collected data of all sensors can be accurately mapped to specific areas of the indoor space, thereby realizing accurate data analysis.

[0098] Step S14: Perform Kalman filtering on the indoor sensor data based on the indoor sensor position calibration acquisition data set to obtain indoor fused sensor data;

[0099] In this embodiment, according to the data set collected by the indoor sensor position calibration, the Kalman filtering algorithm is used to merge the indoor sensor data. Kalman filtering can, in the presence of noise, fuse data from multiple sensors by weighted averaging, reduce sensor errors, and improve the accuracy of the data. Specifically, Kalman filtering will dynamically adjust the error of each sensor according to the data after sensor position calibration, and combine the previous measurement values and the current observation values to predict the true state of the indoor environment. Finally, after being processed by Kalman filtering, indoor fused sensor data is obtained, and these data will more accurately reflect the actual indoor environment.

[0100] Step S15: Perform timestamp data merging based on the indoor fused sensor data and the outdoor meteorological data to be analyzed, so as to obtain a multi-source sensor data set.

[0101] In this embodiment, according to the indoor fused sensor data and the outdoor meteorological data to be analyzed, data merging is performed after timestamp alignment. Specifically, first ensure that the indoor sensor data and the outdoor meteorological data are accurately aligned on the time axis, that is, all data have a unified timestamp. Then, use timestamp merging technology to combine the indoor environmental data and the outdoor meteorological data to form a multi-source sensor data set. To ensure the accuracy of data merging, interpolation methods can be used to process the missing values in the data to ensure that each timestamp has complete indoor and outdoor data. Finally, the merged data set can provide comprehensive indoor and outdoor environmental data, providing complete input data for subsequent analysis and model establishment.

[0102] Optionally, step S2 is specifically:

[0103] Step S21: Extract indoor temperature features and indoor humidity features from the multi-source sensing data set, so as to obtain indoor temperature data and indoor humidity data;

[0104] In this embodiment, the indoor temperature and humidity data obtained from the multi-source sensing data set are processed by a feature extraction algorithm. For example, statistical methods such as mean, variance, maximum, minimum, etc. are used to extract the basic features of temperature and humidity. At the same time, frequency domain analysis methods such as Fourier transform can also be used to extract the periodic features in the data. Finally, through the temperature and humidity feature data extracted by these algorithms, the indoor temperature data (such as temperature readings at different time points) and humidity data (such as humidity change trends) can be obtained, providing basic data for subsequent analysis.

[0105] Step S22: Perform time window change trend analysis on the indoor temperature data and the indoor humidity data respectively, so as to obtain time window temperature trend data and time window humidity trend data;

[0106] In this embodiment, when performing time window change trend analysis on the indoor temperature data and humidity data, the indoor temperature and humidity data are first sliced into multiple fixed time windows according to time. For example, a time window of every 5 minutes is selected, and statistical analysis is performed on the temperature and humidity data within each window. Trend analysis methods such as linear regression or exponential smoothing are used to calculate the change trends of temperature and humidity within each time window, and the increase or decrease trends of temperature changes and the fluctuation conditions of humidity are recorded. Through these analyses, time window temperature trend data and time window humidity trend data are obtained, providing input for subsequent pattern recognition.

[0107] Step S23: Perform indoor temperature pattern recognition based on the time window temperature trend data, so as to obtain indoor temperature pattern data; perform indoor humidity pattern recognition based on the time window humidity trend data, so as to obtain indoor humidity pattern data;

[0108] In this embodiment, according to the time window temperature trend data, the temperature data is classified and analyzed by a pattern recognition algorithm. Specifically, the K-means clustering algorithm can be used to divide the temperature data into different temperature patterns according to the trend changes (such as rising, falling, stable), such as high temperature pattern, low temperature pattern, etc. Similarly, the humidity data can also be used for humidity pattern recognition in the same way, for example, identifying the rising or falling pattern of humidity through a time window-based clustering method. Finally, indoor temperature pattern data and humidity pattern data are obtained respectively through these methods, and these data reflect the change characteristics of temperature and humidity in different time periods.

[0109] Step S24: Perform change pattern correlation on the indoor temperature pattern data and the indoor humidity pattern data to obtain the indoor temperature and humidity pattern data;

[0110] In this embodiment, when performing change pattern correlation on the indoor temperature pattern data and the humidity pattern data, it is first necessary to compare and analyze the patterns of temperature and humidity to identify their common change rules. Specifically, the correlation coefficient between the two can be calculated, or a cross-pattern analysis method can be used to detect the association between temperature changes and humidity changes. For example, in a high-temperature mode, there may be a downward trend in humidity, or when the humidity increases, the temperature may be stable. Through such pattern correlation analysis, the comprehensive pattern of indoor temperature and humidity changes can be extracted to obtain the indoor temperature and humidity pattern data, revealing the overall change characteristics of the indoor environment.

[0111] Step S25: Identify the household air conditioner usage behavior pattern based on the indoor temperature and humidity pattern data to obtain the household air conditioner usage behavior pattern data.

[0112] In this embodiment, based on the indoor temperature and humidity pattern data, through the air conditioner usage behavior pattern recognition algorithm, the air conditioner usage of the household in different temperature and humidity patterns is analyzed. For example, it is assumed that in a high-temperature and high-humidity mode, the household may tend to turn on the air conditioner for cooling and dehumidification, while in a situation with moderate temperature and low humidity, the air conditioner usage frequency is low. Machine learning algorithms such as decision trees and support vector machines are used to train based on the household's historical air conditioner usage data to identify the household's air conditioner usage behavior pattern in different temperature and humidity patterns. Finally, the household air conditioner usage behavior pattern data obtained through this process can be used to predict the household's air conditioner usage behavior in future similar environments.

[0113] Optionally, step S25 is specifically:

[0114] Step S251: Obtain the household air conditioner operation data and perform data preprocessing on the household air conditioner operation data to obtain the household air conditioner operation data to be analyzed;

[0115] In this embodiment, the operation data is obtained from the household air conditioner management platform, such as the on / off time, operation power, temperature setting, etc. of the air conditioner. Then, preprocessing is performed on these data, including removing outliers and filling in missing data. The mean filling or interpolation method can be used to fill in the missing values, and the sliding window algorithm is used for outlier detection. After processing, the household air conditioner operation data set to be analyzed is obtained, which contains the detailed operation records of the air conditioner, providing a basis for subsequent load calculation and pattern recognition.

[0116] Step S252: Calculate the air-conditioning operation load for the household air-conditioning operation data to be analyzed, so as to obtain the air-conditioning operation load data, and perform time-window load division on the air-conditioning operation load data, so as to obtain high-load time-window data and low-load time-window data;

[0117] In this embodiment, for the household air-conditioning operation data to be analyzed, first calculate the air-conditioning operation load, mainly according to the power data of the air-conditioning and the air-conditioning operation time, calculate the load for each time period. For example, assuming that the air-conditioning operates at a power of 200W for 1 hour in a certain time period, then the air-conditioning load for this time period is 200W·h. Then, divide the calculated load data into different time windows (such as every hour or every 30 minutes) according to time, and divide them into high-load and low-load time windows according to the load size. For example, the time window with a load greater than 300W·h is divided into a high-load time window, and the time window with a load less than 100W·h is divided into a low-load time window.

[0118] Step S253: Extract the household set temperature characteristics from the household air-conditioning operation data to be analyzed, so as to obtain the household set temperature data, and perform temperature difference time window statistics on the household set temperature data and the outdoor meteorological data to be analyzed, so as to obtain indoor high temperature difference time window data and indoor low temperature difference time window data;

[0119] In this embodiment, when extracting the household set temperature characteristics from the household air-conditioning operation data, the set temperature at each air-conditioning operation can be extracted, and the temperature setting changes in each time period can be recorded. Then, combined with the outdoor meteorological data (such as external temperature and humidity), perform temperature difference time window statistics. Specifically, calculate the temperature difference between indoor and outdoor in each time period, and divide it into two categories: high temperature difference and low temperature difference. For example, in high-temperature weather, the temperature difference between the indoor set temperature and the external temperature may be large, so the time window data can be marked as a high temperature difference time window.

[0120] Step S254: Perform a time window intersection operation on the indoor high temperature difference time window data and the high-load time window data to obtain the air-conditioning high-load usage time window data; perform a time window intersection operation on the indoor low temperature difference time window data and the low-load time window data to obtain the air-conditioning low-load usage time window data;

[0121] In this embodiment, when performing an intersection operation on the indoor high temperature difference time window data and the high load time window data, a matching algorithm can be used to extract the overlapping part of the high temperature difference time window and the high load time window. For example, assume that the high temperature difference window is from 12:00 to 14:00, and during this time period, the load of the air conditioner is greater than 300 W·h. Then this time period can be marked as the high load usage time window of the air conditioner. Similarly, when performing an intersection operation on the indoor low temperature difference time window data and the low load time window data, the low load usage time period is extracted, such as the time window when the air conditioner load is lower than 100 W·h.

[0122] Step S255: Identify the peak operation mode of the air conditioner for the household air conditioner operation data to be analyzed according to the high load usage time window data of the air conditioner, so as to obtain the peak operation mode data of the air conditioner; identify the idle operation mode of the air conditioner for the household air conditioner operation data to be analyzed according to the low load usage time window data of the air conditioner, so as to obtain the idle operation mode data of the air conditioner.

[0123] In this embodiment, according to the high load usage time window data of the air conditioner, a pattern recognition algorithm (such as K-means clustering or decision tree) can be used to identify the peak operation mode of the air conditioner. For example, assume that the high load time window usually appears from 1 pm to 3 pm in the afternoon, and during this period, the air conditioner load remains high, and the peak operation mode of the air conditioner can be identified. Similarly, according to the low load usage time window data of the air conditioner, the idle operation mode of the air conditioner is identified, such as when the air conditioner load is low at night, or when the air conditioner operation load is low when the temperature is moderate.

[0124] Step S256: Integrate the peak operation mode data of the air conditioner and the idle operation mode data of the air conditioner to obtain the air conditioner usage mode data, and identify the household air conditioner usage behavior mode according to the indoor temperature and humidity mode data and the air conditioner usage mode data, so as to obtain the household air conditioner usage behavior mode data.

[0125] In this embodiment, when integrating the peak operation mode data of the air conditioner and the idle operation mode data of the air conditioner, data fusion technology can be used to combine the recognition results of the two to obtain a complete air conditioner usage mode data. For example, according to the high load and low load modes of the air conditioner usage time period, the air conditioner operation mode of the whole day can be divided into different states. For example, when in the high load mode, the air conditioner works for a long time, while in the idle mode, the air conditioner hardly runs. According to the indoor temperature and humidity mode data and the air conditioner usage mode data, the household air conditioner usage behavior mode is further identified through a behavior pattern recognition algorithm (such as a support vector machine), such as whether the household tends to use the air conditioner to cool down during the day, or whether there is a habit of setting the air conditioner to the automatic mode at night.

[0126] Optionally, step S256 is specifically as follows:

[0127] Integrate the air conditioner usage peak operation mode data and the air conditioner usage idle operation mode data to obtain the air conditioner usage mode data;

[0128] In this embodiment, the air conditioner usage peak operation mode data and the air conditioner usage idle operation mode data are integrated. By analyzing the air conditioner usage data in different time windows, the high-load and low-load operation modes are combined, and the time period is weighted using an algorithm (such as weighted average) to obtain the complete air conditioner usage mode data. Suppose that during high-load periods, the operation duration and frequency of the air conditioner are high, while during low-load periods, the usage duration of the air conditioner is short and the operation power is low. These pieces of information are integrated into a unified air conditioner usage mode, representing the air conditioner usage characteristics of the household at different time periods.

[0129] Extract the air conditioner usage behavior characteristics from the air conditioner usage mode data to obtain the air conditioner switch frequency data and the load period duration data;

[0130] In this embodiment, the air conditioner usage behavior characteristics are extracted based on the integrated air conditioner usage mode data. By counting the air conditioner switch frequency, the number of times the air conditioner is switched on and off per hour or per day by the household is extracted. For example, if a household switches on and off the air conditioner twice from 7 am to 8 am and three times from 9 pm to 11 pm every day, the air conditioner switch frequency data will be recorded. At the same time, the load period duration is analyzed to calculate the duration of continuous operation of the air conditioner in the high-load mode and the duration of operation in the low-load mode. For example, the air conditioner runs continuously for 2 hours from 8 am to 10 am. After extracting these behavior characteristics, the usage habits of the household can be better understood.

[0131] Extract the indoor temperature and humidity fluctuation mode from the indoor temperature and humidity mode data to obtain the indoor temperature and humidity fluctuation mode data;

[0132] In this embodiment, the temperature and humidity fluctuation mode is extracted by analyzing the indoor temperature and humidity mode data. Using time series analysis methods such as Fourier transform or wavelet transform, the fluctuation characteristics of indoor temperature and humidity in different time windows are extracted. For example, suppose that in summer, the indoor temperature drops in the morning and evening, and rises at noon. This temperature fluctuation mode can be obtained by statistically analyzing the fluctuation of data within the time period. After extracting the temperature and humidity fluctuation mode, the operation requirements of the air conditioner during these fluctuations can be predicted more accurately.

[0133] Analyze the household temperature demand based on the air conditioner switch frequency data and the indoor temperature and humidity fluctuation mode data to obtain the household temperature demand mode data;

[0134] In this embodiment, the air conditioner switching frequency data and the indoor temperature and humidity fluctuation mode data are combined to analyze the temperature demand of the household. According to the relationship between the change in the switching frequency and the indoor temperature and humidity fluctuation, the temperature demand mode of the household is judged. For example, if the air conditioner switching frequency is high during a high-temperature period and the indoor temperature fluctuates greatly, it may indicate that the household is more sensitive to temperature, and the air conditioner starts frequently to maintain a comfortable indoor environment. This analysis helps to provide a personalized temperature demand solution for the household.

[0135] Based on the load period duration data and the indoor temperature and humidity fluctuation mode data, the air conditioner usage preference analysis of the household is carried out to obtain the air conditioner usage preference data of the household;

[0136] In this embodiment, the air conditioner usage preference analysis of the household is carried out by combining the load period duration data and the indoor temperature and humidity fluctuation mode data. For example, by analyzing the duration of the load period and the indoor temperature and humidity fluctuation law, the preference of the household for the air conditioner at different time periods is inferred. For example, a certain household prefers to start the air conditioner in the early morning with a lower temperature and maintain a lower load operation, while reducing the air conditioner usage time at noon when it is hot. By analyzing these characteristics, the air conditioner usage preference of the household in different situations can be identified.

[0137] Based on the air conditioner usage preference data of the household and the household temperature demand mode data, the user's air conditioner usage behavior mode is identified to obtain the air conditioner usage behavior mode data of the household.

[0138] In this embodiment, the user's air conditioner usage behavior mode is identified based on the air conditioner usage preference data of the household and the temperature demand mode data. Using machine learning algorithms (such as decision trees or support vector machines), combined with data such as the air conditioner usage preference, switching frequency, and load period duration of the household, the air conditioner usage behavior mode of the household is identified. For example, by analyzing the usage situation of the household in different seasons and time periods, it is judged whether the household is more inclined to frequently adjust the temperature or is used to setting the air conditioner to a fixed temperature and rarely adjusting it.

[0139] Optionally, step S3 is specifically:

[0140] Step S31: Extract the influencing factors of the household usage behavior from the air conditioner usage behavior mode data of the household to obtain the air conditioner usage frequency factor, the load period duration factor, the temperature demand mode factor, and the air conditioner usage preference factor;

[0141] In this embodiment, the data of the household air conditioner usage behavior patterns is analyzed to extract key influencing factors. First, the daily air conditioner switching frequency of the household is counted from the data of the household air conditioner usage behavior patterns. It is assumed that the household switches the air conditioner 3 to 5 times at different time periods every day. Next, the load period duration factor can be extracted by analyzing the continuous working time of the household air conditioner during high-load periods, and the duration of the air conditioner running continuously within each load period is calculated. For example, the high-load period duration of a certain household is 2 hours. The temperature demand pattern factor is extracted based on the fluctuation of the temperature set by the household. It is assumed that a certain household prefers to set the temperature at 25°C at noon and in the evening, while it is set at 28°C in the morning and at night. The air conditioner usage preference factor is based on the household's preference for using the air conditioner. For example, some households prefer a low-temperature state, while other households like to set it at medium-high temperatures. By extracting these four factors, a foundation can be laid for subsequent behavior analysis.

[0142] Step S32: Perform associated factor clustering on the air conditioner usage frequency factor, the load period duration factor, the temperature demand pattern factor, and the air conditioner usage preference factor, so as to obtain a set of household usage behavior influencing factors;

[0143] In this embodiment, associated factor clustering is performed on the four extracted factors. Through a clustering algorithm (such as K-means), the air conditioner usage frequency factor, the load period duration factor, the temperature demand pattern factor, and the air conditioner usage preference factor are clustered. For example, a certain household continuously uses the air conditioner for a long time during the load period and frequently adjusts the temperature setting. Such households will be clustered as high-frequency adjustment users, while other households have a lower air conditioner switching frequency and a shorter load period duration and are classified as low-frequency adjustment users. After clustering, a set of household usage behavior influencing factors is obtained, which helps to perform differential analysis on different types of households.

[0144] Step S33: Perform factor influence weight statistics on the set of household usage behavior influencing factors according to the air conditioner operation load data, so as to obtain influence factor weight data;

[0145] In this embodiment, the factor influence weights of the household usage behavior influence factor set are statistically analyzed according to the air conditioner operation load data. The load consumption of each household under different factors is analyzed through the air conditioner operation load data. For example, during high load periods, frequent switching of the air conditioner by households may result in higher energy consumption, while during low load periods, the load consumption is lower. Based on this data, the influence weight of each factor on the load consumption is calculated. For example, the weight of the air conditioner usage frequency factor is 40%, the weight of the load period duration factor is 30%, the weight of the temperature demand pattern factor is 20%, and the weight of the air conditioner usage preference factor is 10%. Through these weight data, the influence degree of each factor on the household air conditioner usage behavior can be accurately described. Suppose the air conditioner load of household A is 4 kW during the high temperature period in summer, while the load of household B during the same period is 3 kW. Household A has a higher load consumption due to frequently adjusting the air conditioner set temperature. By analyzing the behavior and load data of the households, the influence weights of each factor on the load are obtained. For example, the influence weight of the air conditioner usage frequency factor on the load may be 50%, the weight of the load period duration factor is 30%, the weight of the temperature demand pattern factor is 15%, and the weight of the air conditioner usage preference factor is 5%. The calculation of these weights is based on the air conditioner usage behavior patterns of the households and their actual load consumption, which can provide quantitative support for the subsequent model.

[0146] Step S34: Quantify the household behavior influence factors based on the influence factor weight data and the household usage behavior influence factor set, so as to obtain the household usage behavior influence amount data;

[0147] In this embodiment, based on the factor weight data and the household usage behavior influence factor set, the behavior influence factors of the households are quantified. By quantifying the influence factors of each household according to the weights, for example, if a household's air conditioner continuously works for 2 hours and frequently adjusts the temperature during the load period, it will be given a higher behavior influence amount, and the calculated behavior influence amount of the household is 80%. This quantification result helps to provide personalized air conditioner usage data for each household and further improve the accuracy of the prediction model. Suppose the weight of the air conditioner usage frequency factor of household A is 0.5, the weight of the load period duration factor is 0.3, the weight of the temperature demand pattern factor is 0.15, and the weight of the air conditioner usage preference factor is 0.05. In actual operation, if the air conditioner of household A is switched 5 times during the high load period, the duration is 3 hours, and there is a preference for large fluctuations in the temperature setting, the system will calculate the influence amount of household A to be 80%. This quantification method calculates a comprehensive behavior influence value through weighted calculation of each factor, providing strong data support for the household air conditioner load prediction.

[0148] Step S35: Construct a household personalized load prediction model based on the household usage behavior influence amount data.

[0149] In this embodiment, a personalized load prediction model is constructed based on the data of the influence amount of household usage behavior. By combining the data of the influence amount of household behavior with the air-conditioning usage load data, a personalized load prediction model is constructed using regression analysis or machine learning methods, such as random forest or neural network. For example, through this model, the air-conditioning load demand of the household in a specific time period can be accurately predicted, and the model is dynamically adjusted according to the quantification result of the household behavior influence factor, making the prediction result more accurate and capable of providing real-time optimization control for the air-conditioning management system.

[0150] Optionally, step S4 is specifically as follows:

[0151] Step S41: Extract the environmental sensing time series features from the multi-source sensing data to obtain the environmental sensing time series data;

[0152] In this embodiment, the environmental sensing time series features are extracted from the multi-source sensing data. Specifically, by performing time series analysis on the environmental sensing data such as indoor and outdoor temperature and humidity, illumination, and carbon dioxide concentration, the trend, fluctuation, and periodic features of the data are extracted. For example, the outdoor temperature fluctuates greatly during the daytime in summer, while the indoor temperature is relatively stable in the morning and evening. By extracting these features, rich time series data support can be provided for subsequent air-conditioning load prediction.

[0153] Step S42: Perform air-conditioning load prediction on the environmental sensing time series data through the household personalized load prediction model to obtain the air-conditioning load prediction data;

[0154] In this embodiment, the air-conditioning load prediction is performed on the environmental sensing time series data through the household personalized load prediction model. Specifically, when implementing, a household personalized load prediction model (such as a machine learning model based on historical data) is used, and the extracted environmental sensing time series data (such as temperature change, humidity fluctuation, etc.) is input into the model for load prediction. For example, when the model analyzes that the outdoor temperature is 30°C and the indoor temperature has risen to 28°C, and the household usually sets the air conditioner to 25°C, the air-conditioning load demand in this period is predicted. The model predicts the air-conditioning load demand data by training historical data and provides accurate load prediction.

[0155] Step S43: Calculate the model prediction error of the air-conditioning operation load data and the air-conditioning load prediction data to obtain the model load error data;

[0156] In this embodiment, the model prediction error of the air conditioner operation load data and the air conditioner load prediction data is calculated. Specifically, by comparing the actual air conditioner operation load in the air conditioner operation load data with the air conditioner load prediction data predicted by the model, the error between the two is calculated. For example, in a certain time period, the actual load of the air conditioner is 3 kW, while the load given by the prediction model is 2.8 kW, and the error is 0.2 kW. By continuously calculating the model error, the accuracy of the model prediction can be understood, and the error can be analyzed to further improve the model.

[0157] Step S44: Calculate the environmental variable - air conditioner load correlation based on the model load error data and the environmental sensing time series data, so as to obtain the environmental variable - air conditioner load correlation data;

[0158] In this embodiment, the environmental variable - air conditioner load correlation is calculated based on the model load error data and the environmental sensing time series data. Specifically, multivariate analysis is performed according to historical load data and environmental sensing data (such as indoor and outdoor temperature, humidity, etc.). For example, it is found that the error of the air conditioner load is larger during high-temperature periods in summer, which is related to the drastic change of outdoor temperature or the change of humidity. By calculating the correlation between environmental variables and load, it is possible to identify which environmental factors have a greater impact on the air conditioner load, providing a basis for subsequent model optimization. Use the Pearson correlation coefficient, Spearman correlation coefficient or other statistical methods to analyze the correlation between the model load error and the environmental sensing data. Assume that the error data is error(t) and the environmental variable is outdoor_temp(t), and calculate cor(error(t),outdoor_temp(t)) to analyze the impact of temperature on the load error. If the correlation coefficient is 0.8, it indicates that the outdoor temperature has a greater impact on the air conditioner load prediction error, and more factors related to outdoor temperature fluctuations can be considered to be added to the model.

[0159] Step S45: Perform a correlation matrix transformation on the environmental variable - air conditioner load correlation data, so as to obtain the environmental load association matrix;

[0160] In this embodiment, the correlation data between environmental variables and air-conditioning load is converted into a correlation matrix. Specifically, first, the correlation data between each environmental variable (such as temperature, humidity, light, etc.) and the air-conditioning load is converted into a correlation matrix. For example, the correlation between temperature and load is 0.85, and the correlation between humidity and load is 0.6. In a matrix form, the influence degree of each environmental variable on the air-conditioning load can be clearly shown, providing a clear basis for subsequent data analysis and model adjustment. For example, an item in the matrix can be cor(outdoor_temp(t), air_conditioning_load(t)) = 0.85, indicating a strong positive correlation between outdoor temperature and air-conditioning load. By constructing a correlation matrix, it is possible to quickly identify which environmental factors most significantly affect the air-conditioning load, thus providing a specific direction for optimizing the model.

[0161] Step S46: Dynamically combine the correlation coefficients of the environmental load correlation matrix and the influencing factor set of the household usage behavior to obtain a dynamic adjustment factor;

[0162] In this embodiment, a new dynamic adjustment factor is calculated by combining the correlation data in the environmental load correlation matrix with the household behavior characteristic data (such as air conditioner switching frequency, load period duration, etc.). For example, assume that outdoor temperature and humidity have a high positive correlation with air conditioner load, the household prefers a low temperature setting, and has a high air conditioner switching frequency. The correlation matrix can show that cor(outdoor_temp, air_conditioning_load) is 0.85 and cor(humidity, air_conditioning_load) is 0.72, and the household behavior characteristic such as preferring a low temperature setting leads to an increase in load. At this time, the influence factor of household behavior should be increased. When combining the correlation data in the environmental load correlation matrix with the household behavior characteristic data, a weighted average or weighted synthesis method can be used: for example, let the correlation of environmental variables be C_env and the correlation of household behavior characteristics be C_behavior, then the new dynamic adjustment factor C_adjusted can be obtained by weighted summation: C_{\text{adjusted}} = w_{1}\times C_{env}+w_{2}\times C_{behavior}; where w_1 and w_2 are dynamic weights that reflect the relative importance of environmental variables and household behavior characteristics. During the weighting process, the values of w_1 and w_2 can be adjusted according to real-time feedback. For example, if the external temperature suddenly rises, then w_1 increases; if the household increases the air conditioner usage frequency, then w_2 increases. This dynamic adjustment ensures more accurate prediction of air conditioner load, can reasonably adjust the air conditioner operation strategy in high-temperature weather, and achieve a balance between energy conservation and comfort. Assume that in a high-temperature summer day, the outdoor temperature is 35°C and the humidity is 80%, and the household prefers the air conditioner to be set at 22°C. At this time, it is found that the environmental load correlation matrix shows a high correlation (0.9) between outdoor temperature and air conditioner load, but the influence of humidity on the load is low (0.5). At the same time, the household frequently turns on the air conditioner, and the air conditioner load is high during high-temperature periods. According to this information, dynamically adjust the weight of the environmental variable temperature to improve the accuracy of load prediction and ensure that it can accurately reflect the air conditioner load demand of the household in high-temperature weather.

[0163] Step S47: Iteratively adjust the model parameters of the household personalized load prediction model according to the dynamic adjustment factor, so as to obtain an indoor air conditioner load prediction model.

[0164] In this embodiment, the model parameters of the personalized load prediction model for households are iteratively adjusted according to the dynamic adjustment factor. The aforementioned dynamic adjustment factor is applied to the personalized load prediction model to adjust the prediction parameters of the model. For example, when it is identified that a certain household usually sets the air conditioner to a low temperature (such as 20°C) in hot weather and the outdoor temperature fluctuation has a greater impact on the air conditioner load, the model will automatically adjust the relevant parameters to make it more sensitive to temperature fluctuations and pay more attention to the low-temperature needs of the household. Through continuous iterative optimization, the model can continuously improve the prediction accuracy, especially in extreme climate conditions or when the air conditioner usage behavior of the household changes.

[0165] Optionally, step S5 is specifically as follows:

[0166] Step S51: Obtain real-time outdoor meteorological data and real-time household interaction data, and calculate the indoor temperature difference between the real-time meteorological data and the multi-source sensing data, so as to obtain the indoor temperature difference data;

[0167] In this embodiment, the real-time outdoor meteorological data is obtained through meteorological sensors (such as temperature, humidity, and wind speed sensors) installed outside the building or a local meteorological website. The household interaction data collected from smart home devices (such as smart thermostats, smart sockets, etc.) can be transmitted to local smart devices through APIs. Inside the room, real-time indoor data is collected through deployed temperature and humidity sensors (such as wall sensors or smart thermostats). After these data are fused and processed, the indoor temperature difference is calculated. For example, when the outdoor temperature is 32°C and the indoor temperature is 28°C, the indoor temperature difference is 4°C, and this data is used to evaluate the air conditioner load demand and indicate the need to adjust the air conditioner intensity to maintain a comfortable indoor environment.

[0168] Step S52: Extract the household behavior characteristics from the real-time household interaction data, so as to obtain the household real-time behavior characteristic data;

[0169] In this embodiment, the household behavior characteristics are extracted by using the household interaction data collected from smart home devices (such as smart thermostats, smart sockets, etc.). For example, a certain household usually sets the air conditioner to 20°C in the morning and has a high air conditioner switching frequency, while in the evening, it tends to maintain 22°C. By analyzing this data, behavior characteristics such as the temperature control habits, air conditioner switching frequency, and adjustment amplitude of a specific household are extracted to form a personalized behavior data set for subsequent load prediction.

[0170] Step S53: Perform air conditioner load prediction on the indoor temperature difference data and the household real-time behavior characteristic data through the indoor air conditioner load prediction model, so as to obtain the indoor air conditioner load prediction data;

[0171] In this embodiment, the real-time indoor temperature difference data and the household behavior characteristic data are input into an air-conditioning load prediction model (such as a regression model based on machine learning) for air-conditioning load prediction. For example, it is assumed that based on the parameters obtained from historical data by a machine learning model, combined with the current temperature difference and behavior data, the model predicts that the air-conditioning load will increase to 600 W under such conditions. The load prediction value helps the system accurately judge the air-conditioning load demand in the future for a period of time for optimized control.

[0172] Step S54: Based on the household air-conditioning usage behavior pattern data, identify the household air-conditioning usage pattern for the indoor air-conditioning load prediction data, so as to obtain the real-time household air-conditioning usage pattern data;

[0173] In this embodiment, pattern recognition is performed based on the household air-conditioning usage behavior pattern data (such as long-term use of the air conditioner during peak hours, preference for low-temperature environments). For example, a certain household turns on the air conditioner and sets a low temperature during the high-temperature period in the afternoon every day. The model identifies the household's air-conditioning usage pattern (such as high-load long-term use) by analyzing historical usage data and the predefined household air-conditioning usage behavior pattern data. The combination of this pattern data and the air-conditioning load prediction result forms a real-time identification of the current air-conditioning usage behavior, which is convenient for predicting future demands and making optimized adjustments.

[0174] Step S55: Obtain the real-time indoor air-conditioning operation mode data, and select an air-conditioning operation strategy for the real-time indoor air-conditioning operation mode data according to the real-time household air-conditioning usage pattern data, so as to obtain the air-conditioning equipment operation strategy;

[0175] In this embodiment, the real-time indoor air-conditioning operation mode data (such as the actual air-conditioning operation power, wind speed, set temperature, etc.) is obtained through the air-conditioning equipment management cloud platform and compared with the real-time household air-conditioning usage pattern data. For example, if a household sets a low temperature for a long time during a high-load period, the air-conditioning operation mode can be selected to be adjusted to an energy-saving mode to reduce load consumption and extend the service life of the air conditioner. This adjustment is achieved through the control platform of the intelligent air-conditioning equipment.

[0176] Step S56: Perform adaptive strategy optimization on the air-conditioning equipment operation strategy to obtain the optimized air-conditioning equipment operation strategy, and upload it to the air-conditioning equipment management cloud platform to execute the air-conditioning equipment control task.

[0177] In this embodiment, after the operation strategy of the air-conditioning equipment is optimized, adaptive strategy adjustment is performed based on real-time environment and usage pattern data. Assuming that it is detected that the outdoor temperature is continuously rising and the load is too high, the optimization algorithm (genetic algorithm, annealing algorithm, adaptive optimization algorithm, etc.) can automatically lower the air-conditioning output power or increase the wind speed to ensure that the air-conditioning operates in a more efficient manner. After the optimization is completed, the strategy is uploaded through the air-conditioning equipment management cloud platform for the execution of control tasks, and the optimization strategy is continuously adjusted according to the data feedback, so as to achieve fully automatic and intelligent air-conditioning management.

[0178] Optionally, this specification also provides a system for establishing an air-conditioning load prediction model, which is used to execute the method for establishing an air-conditioning load prediction model as described above. The system for establishing an air-conditioning load prediction model includes:

[0179] A sensing data fusion module, which is used to obtain indoor sensing data and outdoor meteorological data, perform indoor sensing data fusion on the indoor sensing data to obtain indoor fused sensing data; perform time-series merging on the indoor fused sensing data and the outdoor meteorological data to obtain a multi-source sensing data set;

[0180] A usage behavior pattern recognition module, which is used to perform indoor temperature and humidity pattern analysis based on the multi-source sensing data set to obtain indoor temperature and humidity pattern data; perform household air-conditioning usage behavior pattern recognition based on the indoor temperature pattern data to obtain household air-conditioning usage behavior pattern data;

[0181] A prediction model construction module, which is used to quantify the influence of household usage behavior on the household air-conditioning usage behavior pattern data to obtain household usage behavior influence amount data, and construct a household personalized load prediction model based on the household usage behavior influence amount data;

[0182] An environmental load adjustment module, which is used to perform environment-load correlation modeling on the multi-source sensing data set according to the household personalized load prediction model to obtain an environment-load correlation matrix, and perform environment-load correlation adjustment on the household personalized load prediction model according to the environment-load correlation matrix to obtain an indoor air-conditioning load prediction model;

[0183] An operation strategy optimization module, which is used to obtain real-time outdoor meteorological data and real-time household interaction data, perform air-conditioning load prediction on the real-time outdoor meteorological data and the real-time household interaction data through the indoor air-conditioning load prediction model to obtain indoor air-conditioning load prediction data; perform adaptive optimization of the air-conditioning equipment operation strategy based on the indoor air-conditioning load prediction data to obtain an optimized operation strategy for the air-conditioning equipment, and upload it to the air-conditioning equipment management cloud platform to execute the air-conditioning equipment control task.

[0184] The air-conditioning load prediction model establishment system of the present invention can implement any one of the air-conditioning load prediction model establishment methods of the present invention, and is used as a medium for coordinating operations and signal transmission among various modules to complete the air-conditioning load prediction model establishment method. The internal modules of the system cooperate with each other, thereby effectively improving the accuracy of air-conditioning load prediction.

[0185] Optionally, this specification also provides a storage medium storing a computer program, and when the computer program is executed by a processor, the above-mentioned air-conditioning load prediction model establishment method is implemented.

[0186] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application document are intended to be embraced within the present invention.

[0187] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A method for establishing an air-conditioning load prediction model, characterized in that, it includes the following steps: Step S1: Obtain indoor sensing data and outdoor meteorological data, and perform indoor sensing data fusion on the indoor sensing data to obtain indoor fused sensing data; Perform time-series merging on the indoor fused sensing data and the outdoor meteorological data to obtain a multi-source sensing data set; Step S2: Perform indoor temperature and humidity pattern analysis based on the multi-source sensing data set to obtain indoor temperature and humidity pattern data; perform household air-conditioning usage behavior pattern recognition based on the indoor temperature pattern data to obtain household air-conditioning usage behavior pattern data; Step S3: Quantify the influence of household usage behavior on the household air-conditioning usage behavior pattern data to obtain household usage behavior influence amount data, and construct a household personalized load prediction model based on the household usage behavior influence amount data; Step S4: Perform environment-load correlation modeling on the multi-source sensing data set according to the household personalized load prediction model to obtain an environment-load correlation matrix, and perform environment-load correlation adjustment on the household personalized load prediction model according to the environment-load correlation matrix to obtain an indoor air-conditioning load prediction model; Step S5: Obtain real-time outdoor meteorological data and real-time household interaction data, and perform air-conditioning load prediction on the real-time outdoor meteorological data and the real-time household interaction data through the indoor air-conditioning load prediction model to obtain indoor air-conditioning load prediction data; perform adaptive optimization of the air-conditioning equipment operation strategy based on the indoor air-conditioning load prediction data to obtain an optimized air-conditioning equipment operation strategy, and upload it to the air-conditioning equipment management cloud platform to execute the air-conditioning equipment control task.

2. The method for establishing an air-conditioning load prediction model according to claim 1, characterized in that, Step S1 is specifically: Step S11: Obtain indoor sensing data and outdoor meteorological data, and perform data preprocessing on the indoor sensing data and the outdoor meteorological data respectively to obtain indoor sensing data to be analyzed and outdoor meteorological data to be analyzed; Step S12: Construct an indoor coordinate system according to the indoor sensing data to be analyzed to obtain indoor space three-dimensional coordinate system data; Step S13: Calibrate the sensor positions of the indoor sensing data to be analyzed according to the indoor space three-dimensional coordinate system data to obtain an indoor sensor position calibration acquisition data set; Step S14: Perform Kalman filter indoor sensing data merging based on the indoor sensor position calibration acquisition data set to obtain indoor fused sensing data; Step S15: Perform timestamp data merging according to the indoor fused sensing data and the outdoor meteorological data to be analyzed to obtain a multi-source sensing data set.

3. The method for establishing an air-conditioning load prediction model according to claim 1, characterized in that, Step S2 is specifically: Step S21: Extract indoor temperature features and indoor humidity features from the multi-source sensing data set to obtain indoor temperature data and indoor humidity data; Step S22: Perform time-window change trend analysis on the indoor temperature data and the indoor humidity data respectively to obtain time-window temperature trend data and time-window humidity trend data; Step S23: Perform indoor temperature pattern recognition based on the time-window temperature trend data to obtain indoor temperature pattern data; perform indoor humidity pattern recognition based on the time-window humidity trend data to obtain indoor humidity pattern data; Step S24: Perform change pattern association on the indoor temperature pattern data and the indoor humidity pattern data to obtain indoor temperature and humidity pattern data; Step S25: Perform recognition of the household air conditioner usage behavior pattern based on the indoor temperature and humidity pattern data to obtain the household air conditioner usage behavior pattern data.

4. The method for establishing an air conditioner load prediction model according to claim 3, wherein, Step S25 is specifically as follows: Step S251: Obtain the household air conditioner operation data, and perform data preprocessing on the household air conditioner operation data to obtain the household air conditioner operation data to be analyzed; Step S252: Perform air conditioner operation load calculation on the household air conditioner operation data to be analyzed to obtain the air conditioner operation load data, and perform time-window load division on the air conditioner operation load data to obtain the high-load time-window data and the low-load time-window data; Step S253: Extract the household set temperature characteristics from the household air conditioner operation data to be analyzed to obtain the household set temperature data, and perform temperature difference time-window statistics on the household set temperature data and the outdoor meteorological data to be analyzed to obtain the indoor high temperature difference time-window data and the indoor low temperature difference time-window data; Step S254: Perform time-window intersection operation on the indoor high temperature difference time-window data and the high-load time-window data to obtain the air conditioner high-load usage time-window data; Perform time-window intersection operation on the indoor low temperature difference time-window data and the low-load time-window data to obtain the air conditioner low-load usage time-window data; Step S255: Perform recognition of the air conditioner usage peak operation mode on the household air conditioner operation data to be analyzed according to the air conditioner high-load usage time-window data to obtain the air conditioner usage peak operation mode data; Perform recognition of the air conditioner usage idle operation mode on the household air conditioner operation data to be analyzed according to the air conditioner low-load usage time-window data to obtain the air conditioner usage idle operation mode data; Step S256: Perform integration of the air conditioner usage peak operation mode data and the air conditioner usage idle operation mode data to obtain the air conditioner usage mode data, and perform recognition of the household air conditioner usage behavior pattern according to the indoor temperature and humidity pattern data and the air conditioner usage mode data to obtain the household air conditioner usage behavior pattern data.

5. The method for establishing an air conditioner load prediction model according to claim 4, wherein, Step S256 is specifically as follows: Perform integration of the air conditioner usage peak operation mode data and the air conditioner usage idle operation mode data to obtain the air conditioner usage mode data; Extract the air conditioner usage behavior characteristics from the air conditioner usage mode data to obtain the air conditioner switch frequency data and the load period duration data; Extract the indoor temperature and humidity fluctuation pattern according to the indoor temperature and humidity pattern data to obtain the indoor temperature and humidity fluctuation pattern data; Analyze the temperature demand of households based on the air conditioner switching frequency data and the indoor temperature and humidity fluctuation pattern data, so as to obtain the household temperature demand pattern data; Analyze the air conditioner usage preferences of households based on the load period duration data and the indoor temperature and humidity fluctuation pattern data, so as to obtain the household air conditioner usage preference data; Identify the user air conditioner usage behavior pattern based on the household air conditioner usage preference data and the household temperature demand pattern data, so as to obtain the household air conditioner usage behavior pattern data.

6. The method for establishing an air conditioner load prediction model according to claim 1, characterized in that, Step S3 is specifically as follows: Step S31: Extract the influencing factors of household usage behavior from the household air conditioner usage behavior pattern data, so as to obtain the air conditioner usage frequency factor, the load period duration factor, the temperature demand pattern factor, and the air conditioner usage preference factor; Step S32: Cluster the associated factors of the air conditioner usage frequency factor, the load period duration factor, the temperature demand pattern factor, and the air conditioner usage preference factor, so as to obtain the set of influencing factors of household usage behavior; Step S33: Statistically calculate the factor influence weights of the set of influencing factors of household usage behavior according to the air conditioner operation load data, so as to obtain the influence factor weight data; Step S34: Quantify the influencing factors of household behavior based on the influence factor weight data and the set of influencing factors of household usage behavior, so as to obtain the data of the influencing amount of household usage behavior; Step S35: Construct a personalized household load prediction model based on the data of the influencing amount of household usage behavior.

7. The method for establishing an air conditioner load prediction model according to claim 1, characterized in that, Step S4 is specifically as follows: Step S41: Extract the time series characteristics of environmental sensing from the multi-source sensing data, so as to obtain the environmental sensing time series data; Step S42: Predict the air conditioner load based on the environmental sensing time series data through the personalized household load prediction model, so as to obtain the air conditioner load prediction data; Step S43: Calculate the model prediction error of the air conditioner operation load data and the air conditioner load prediction data, so as to obtain the model load error data; Step S44: Calculate the correlation between the environmental variable and the air conditioner load according to the model load error data and the environmental sensing time series data, so as to obtain the environmental variable-air conditioner load correlation data; Step S45: Convert the environmental variable-air conditioner load correlation data into a correlation matrix, so as to obtain the environmental load association matrix; Step S46: Combine the dynamic factors of the correlation coefficients of the environmental load association matrix and the set of influencing factors of household usage behavior, so as to obtain the dynamic adjustment factor; Step S47: Iteratively adjust the model parameters of the personalized household load prediction model according to the dynamic adjustment factor, so as to obtain the indoor air conditioner load prediction model.

8. The method for establishing an air conditioner load prediction model according to claim 1, characterized in that, Step S5 is specifically as follows: Step S51: Obtain the real-time outdoor meteorological data and the real-time household interaction data, and calculate the indoor temperature difference between the real-time meteorological data and the multi-source sensing data, so as to obtain the indoor temperature difference data; Step S52: Extract the household behavior characteristics from the real-time household interaction data to obtain the real-time household behavior characteristic data; Step S53: Perform air-conditioning load prediction on the indoor temperature difference data and the real-time household behavior characteristic data through the indoor air-conditioning load prediction model to obtain the indoor air-conditioning load prediction data; Step S54: Identify the household air-conditioning usage pattern for the indoor air-conditioning load prediction data based on the household air-conditioning usage behavior pattern data to obtain the real-time household air-conditioning usage pattern data; Step S55: Obtain the real-time indoor air-conditioning operation mode data, and select the air-conditioning operation strategy for the real-time indoor air-conditioning operation mode data according to the real-time household air-conditioning usage pattern data to obtain the air-conditioning equipment operation strategy; Step S56: Perform adaptive strategy optimization on the air-conditioning equipment operation strategy to obtain the optimized air-conditioning equipment operation strategy, and upload it to the air-conditioning equipment management cloud platform to execute the air-conditioning equipment control task.

9. An air-conditioning load prediction model establishment system, characterized in that, it is used to execute the air-conditioning load prediction model establishment method as described in claim 1. This air-conditioning load prediction model establishment system includes: A sensing data fusion module, which is used to obtain indoor sensing data and outdoor meteorological data, and perform indoor sensing data fusion on the indoor sensing data to obtain indoor fused sensing data; perform time-series merging on the indoor fused sensing data and the outdoor meteorological data to obtain a multi-source sensing data set; A usage behavior pattern recognition module, which is used to perform indoor temperature and humidity pattern analysis based on the multi-source sensing data set to obtain indoor temperature and humidity pattern data; identify the household air-conditioning usage behavior pattern according to the indoor temperature pattern data to obtain the household air-conditioning usage behavior pattern data; A prediction model construction module, which is used to quantify the influence of the household usage behavior on the household air-conditioning usage behavior pattern data to obtain the household usage behavior influence amount data, and construct a household personalized load prediction model based on the household usage behavior influence amount data; An environmental load adjustment module, which is used to perform environmental-load correlation modeling on the multi-source sensing data set according to the household personalized load prediction model to obtain an environmental load correlation matrix, and perform environmental load correlation adjustment on the household personalized load prediction model according to the environmental load correlation matrix to obtain the indoor air-conditioning load prediction model; An operation strategy optimization module, which is used to obtain the real-time outdoor meteorological data and the real-time household interaction data, and perform air-conditioning load prediction on the real-time outdoor meteorological data and the real-time household interaction data through the indoor air-conditioning load prediction model to obtain the indoor air-conditioning load prediction data; perform adaptive optimization of the air-conditioning equipment operation strategy based on the indoor air-conditioning load prediction data to obtain the optimized air-conditioning equipment operation strategy, and upload it to the air-conditioning equipment management cloud platform to execute the air-conditioning equipment control task.

10. A storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the air-conditioning load prediction model establishment method as described in any one of claims 1 to 8.

Images