Multi-target intelligent optimization air conditioner control method and system based on multi-dimensional machine learning

Abstract

The invention discloses a multi-target intelligent optimization air conditioner control method and system based on multi-dimensional machine learning, and aims to acquire outdoor environment data by acquiring personnel flow and environment data inside and outside a building in real time and combining weather forecast information and the like, and predict future personnel flow trend and environment change conditions by utilizing a machine learning algorithm, so as to realize intelligent optimization of the air conditioner. And data support is provided for an artificial intelligence control strategy of the air conditioning system. Based on the prediction results, a multi-target optimization control algorithm is used for conducting multi-dimensional parameter regression on the air conditioner operation effect, and operation parameters such as the temperature, the humidity and the wind speed of the air conditioner system are dynamically adjusted. The air conditioning system is suitable for various large buildings such as office buildings, shopping malls and airports, the energy utilization efficiency of the air conditioning system is remarkably improved and the energy consumption is greatly reduced on the premise that the indoor environment comfort is ensured, and the air conditioning system has remarkable energy-saving and carbon-reducing effects, wide application prospects and remarkable economic and environmental protection benefits.

Description

Technical Field

[0001] The present invention relates to the field of air conditioning control technology, and specifically to a multi-objective intelligent optimization air conditioning control method and system based on multi-dimensional machine learning, which can dynamically optimize the operating parameters of the air conditioning system according to personnel flow and environmental changes, and is suitable for energy saving and comfort adjustment of buildings such as office buildings and shopping malls. Background Art

[0002] With the increasing severity of the global energy shortage problem and the continuous improvement of environmental protection requirements, energy conservation and environmental protection of buildings have become an important issue that needs to be urgently addressed in modern society. In building energy consumption, air conditioning systems are one of the main sources of energy consumption, especially in large buildings such as office buildings, shopping malls and airports, where air conditioning systems account for a considerable proportion of energy consumption. At the same time, the comfort of the indoor environment directly affects user experience, work efficiency and satisfaction in public places. Therefore, under the premise of ensuring the comfort of the indoor environment, maximizing the operating efficiency of the air conditioning system and reducing energy consumption are of great significance for energy conservation and emission reduction. However, traditional air conditioning control systems mostly use simple temperature settings and time control, lack intelligent adjustment methods, and are often difficult to adapt to complex and changing environments, resulting in excessive energy consumption and insufficient comfort. They mainly rely on manual control and simple timer settings, lack a dynamic adjustment mechanism for real-time data, and cannot be optimized according to actual conditions, resulting in unnecessary energy consumption. The existing air conditioning intelligent control system mainly focuses on a single factor and relies on a simple indoor temperature sensor. It cannot effectively integrate multi-dimensional environmental data (such as personnel flow, outdoor climate change, air quality, etc.) for comprehensive optimization control. This makes it impossible for the system to predictively regulate future environmental changes, making it difficult to achieve a good balance between energy consumption and indoor comfort.

[0003] In recent years, with the rapid development of information technology and intelligent technology, air conditioning control technology has gradually moved towards intelligence and automation. In particular, with the maturity of algorithms such as machine learning and deep learning, intelligent control technology has begun to show great potential in building energy management. Through real-time data acquisition and learning models, intelligent air conditioning control systems can conduct comprehensive analysis of multi-dimensional data, including personnel flow, indoor and outdoor environmental conditions, weather forecasts, etc. On the one hand, it can make high-precision predictions of the internal and external environmental parameters of buildings that affect the energy efficiency of air conditioning. On the other hand, it can continuously optimize the control strategy based on the predicted data, optimize the operation of the air conditioning system through adaptive control algorithms, improve operating efficiency and comfort, and provide a new solution for the intelligentization of air conditioning control. To this end, we propose a multi-objective intelligent optimization air conditioning control method and system based on multi-dimensional machine learning, which can use machine learning algorithms to predict future changes in personnel flow and environment, and dynamically adjust the operating parameters of the air conditioning system in combination with multi-objective optimization control algorithms to achieve the goals of energy saving and carbon reduction and improving indoor comfort. Summary of the invention

[0004] The present invention provides a multi-objective intelligent optimization air-conditioning control method and system based on multi-dimensional machine learning, aiming to solve the technical problems that traditional air-conditioning control systems adopt simple temperature setting and time control, lack intelligent adjustment means, are difficult to adapt to complex and changeable environments, lead to excessive energy consumption and insufficient comfort, lack a dynamic adjustment mechanism for real-time data, cause unnecessary energy consumption, and are difficult to achieve balanced optimization of energy saving and comfort.

[0005] The present invention aims to provide a multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning, the method comprising the steps of:

[0006] Collecting in real time first initial data of the building operating environment and second initial data of the weather forecast of the building;

[0007] The collected first initial data and second initial data are cleaned, preprocessed, smoothed, and feature values ​​are extracted in sequence to generate a first initial data input set and a second initial data input set, and the first initial data input set and the second initial data input set are randomly divided into a training data set and a validation data set, respectively;

[0008] Training a first training model and a second training model based on an LSTM neural network with a time series output and outputting first prediction data and second prediction data respectively;

[0009] Optimizing the first prediction data and the second prediction data according to a multi-objective optimization control strategy and generating secondary input data;

[0010] According to the secondary input data, real-time feedback is provided to dynamically adjust the operating parameters of the air conditioner;

[0011] Preferably, the first initial data is personnel flow data in the building and indoor and outdoor environmental parameter data;

[0012] The second initial data is real-time weather forecast data of the location of the building;

[0013] Preferably, the smoothing calculation formula of the first initial data is:

[0014]

[0015] Among them, MA t represents the moving average of the first initial data at time t, n represents the number of moving average items, x t-i Represents the first initial data at time ti;

[0016] Preferably, the smoothing calculation formula of the second initial data is:

[0017] S t =αY t +(1-α)S t-1

[0018] Among them, S t represents the smoothed value of the weather forecast data at time t, α is the smoothing coefficient, and Y t Represents the value of the initial weather forecast data at time t;

[0019] Preferably, the formula of the first training model based on the LSTM neural network is:

[0020] h t1 =LSTM(x t1 ,h t1-1 )

[0021] Among them, h t1 represents the hidden state of the first initial data at time t1, x t1 represents the first initial data input value, h t1-1 Represents the hidden state of the first initial data at the previous moment;

[0022] Preferably, the formula of the second training model based on the LSTM neural network is:

[0023] h t2 =LSTM(x t2 ,h t2-1 )

[0024] Among them, h t2 represents the hidden state of the second initial data at time t2, xt2 represents the second initial data input value, h t2-1 Represents the hidden state of the second initial data at the previous moment;

[0025] Preferably, the calculation formula of the multi-objective optimization control strategy is:

[0026] min(α·E+β·C)

[0027] Among them, E represents energy consumption, C represents comfort deviation, α and β are weight coefficients;

[0028] Preferably, the feedback and dynamic adjustment formula is as follows:

[0029]

[0030] Among them, ΔP represents the adjustment amount, K p ,K i , K d They are proportional, integral and differential coefficients, SP represents the set value, and PV represents the process variable;

[0031] Preferably, the step of "real-time feedback and dynamic adjustment of the operating parameters of the air conditioner according to the secondary input data" further includes the following steps:

[0032] Strengthen self-learning, improve the multi-objective optimization control strategy, and enter the step of "optimizing the first prediction data and the second prediction data according to the multi-objective optimization control strategy and generating secondary input data";

[0033] The algorithm formula for strengthening self-learning is:

[0034] Q(s,a)←Q(s,a)+α(r+γmax a Q(s′,a′)-Q(s,a))

[0035] Among them, Q(s,a) represents the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ and a′ represent the next state and the next action respectively.

[0036] Another object of the present invention is to provide a multi-objective intelligent optimization air conditioning control method and system based on multi-dimensional machine learning, wherein the system includes a data acquisition module, a data processing module, a model training module, an optimization control module, and a feedback dynamic adjustment module;

[0037] The data acquisition module is connected to the data processing module, and is used to collect first initial data of the building operation environment and second initial data of the weather forecast of the building in real time, and send the first initial data and the second initial data to the data processing module;

[0038] The data processing module is connected to the data acquisition module and the model training module, and is used to clean, preprocess, smooth and calculate the collected first initial data and second initial data in sequence, extract characteristic values, and generate a first initial data input set and a second initial data input set respectively, and randomly divide the first initial data input set and the second initial data input set into a training data set and a verification data set respectively, and send the training data set and the verification data set in the first initial data input set and the second initial data input set to the model training module respectively;

[0039] The model training module is connected to the data processing module and the optimization control module, and is used to prefabricate a first training model and a second training model based on an LSTM neural network, respectively train the first training model and the second training model to generate first prediction data and second prediction data with time series output, and send the first prediction data and the second prediction data to the optimization control module;

[0040] The optimization control module is connected to the model training module and the feedback dynamic adjustment module, and is used to optimize the first prediction data and the second prediction data according to the multi-objective optimization control strategy and generate secondary input data, and send the secondary input data to the feedback dynamic adjustment module;

[0041] The feedback dynamic adjustment module is connected to the optimization control module, and is used for real-time feedback and dynamic adjustment of the operating parameters of the air conditioner according to the secondary input data.

[0042] The present invention ensures the accuracy of the initial data by cleaning, preprocessing, smoothing calculation, extracting eigenvalues ​​and other processes on the collected initial data in sequence, effectively smoothing out short-term fluctuations and noise, making the data trend clearer, reducing the random fluctuations of the initial data, and averaging the extreme values ​​and abnormal fluctuations in the data, reducing data uncertainty, improving the stability of the initial data, and providing high-quality input data for model training; through time series analysis and machine learning technology, a long short-term memory network (LSTM) training model is used for high-precision prediction. Through mathematical training models and algorithms, the flow of people in a building and changes in the indoor and outdoor environment can be predicted efficiently and accurately, providing a solid foundation for the optimization control strategy of the air-conditioning intelligent system, and can efficiently and accurately predict future weather changes, provide data support for the intelligent control strategy of the air-conditioning system, and achieve the goals of energy saving, carbon reduction and comfort optimization; through The multi-objective optimization control strategy comprehensively considers energy efficiency and indoor comfort, achieves balance through weight adjustment, reduces energy consumption and improves comfort; analyzes historical data and current data through enhanced self-learning, continuously optimizes the control strategy, improves the adaptability and intelligence level of the system, and gradually improves the accuracy and robustness of the control model; analyzes the system's operating data and control effects through a feedback mechanism, continuously optimizes and updates the control model parameters, and the system identifies the effectiveness and shortcomings of the control strategy based on the comparison of historical data and current data, makes targeted adjustments and improvements, and improves the system's prediction accuracy and adaptability; in theoretical application, enhanced self-learning is widely applicable to the optimization of air-conditioning systems and energy conservation and carbon reduction in intelligent buildings. Through self-learning and optimized control, the system can respond to changes in real time under different environments, provide personalized control solutions, improve energy efficiency, reduce carbon emissions, and achieve the goals of intelligent and green buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning of the present invention;

[0044] Figure 2 It is a schematic diagram of the machine learning model of each module of the multi-objective intelligent optimization air-conditioning control system based on multi-dimensional machine learning of the present invention;

[0045] Figure 3 It is a structural schematic diagram of the multi-objective intelligent optimization air-conditioning control system based on multi-dimensional machine learning of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention.

[0047] Embodiment 1

[0048] like Figure 1 and Figure 2 The flowchart of the multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning of the present invention is shown, and the method comprises the steps of:

[0049] S101, collecting first initial data of a building operation environment and second initial data of a weather forecast of the building in real time;

[0050] The first initial data is personnel flow data in the building and indoor and outdoor environmental parameter data;

[0051] Collecting real-time data on the flow of people in the building through cameras and crowd flow sensors. The collection methods include installing high-definition cameras and high-sensitivity sensors at key entrances and main passages to capture and record the number of people and their movements in real time;

[0052] By installing thermometers, hygrometers, air quality sensors, and carbon dioxide concentration sensors in key areas of the building, indoor and outdoor environmental parameter data are monitored in real time. The collected indoor and outdoor environmental parameter data include indoor and outdoor temperature, humidity, air quality (such as PM2.5 concentration), and carbon dioxide concentration. These indoor and outdoor environmental parameter data can reflect the operating effect of the air-conditioning system and the comfort of the indoor and outdoor environment;

[0053] The second initial data is real-time weather forecast data of the location of the building, the weather forecast data is acquired in real time, and the second initial data is combined with the first initial data to optimize the air conditioning control strategy of the building;

[0054] The second initial data acquisition method refers to obtaining real-time outdoor weather forecast data near the area where the building is located by integrating with the API interface of a third-party weather forecast service provider, wherein the weather forecast data includes current and future weather forecast information, such as temperature, humidity, rainfall, wind speed and wind direction, etc., providing an important reference for the air conditioning control strategy of the building on the impact of external environmental changes on the internal environment of the building;

[0055] The weather forecast data generally includes current weather data and weather forecast information within the next 24 hours;

[0056] Use standard HTTP requests to obtain real-time weather forecast data in JSON format through the API interface to ensure the real-time and accuracy of weather forecast data. This part of real-time weather forecast data is defined as unprocessed initial weather forecast data;

[0057] S102, sequentially cleaning, preprocessing, smoothing calculation, and extracting characteristic values ​​of the collected first initial data and second initial data, and generating a first initial data input set and a second initial data input set respectively, and randomly dividing the first initial data input set and the second initial data input set into a training data set and a validation data set respectively;

[0058] The cleaning and preprocessing of the first initial data is to remove noise and abnormal values ​​in the first initial data, such as obviously abnormal temperature and humidity, so as to ensure the accuracy of the first initial data;

[0059] The cleaning and preprocessing of the second initial data is to clean the acquired initial real-time weather forecast data to remove invalid data and abnormal values, such as missing weather data or erroneous weather values;

[0060] The smoothing calculation of the first initial data is a moving average calculation, which effectively smooths out short-term fluctuations and noise, makes the data trend clearer, reduces the random fluctuations of the first initial data, and at the same time, the extreme values ​​and abnormal fluctuations in the data are averaged, reducing data uncertainty and improving the stability of the first initial data;

[0061] The formula for calculating the first initial data by the moving average method is:

[0062]

[0063] Among them, MA t represents the moving average of the first initial data at time t, n represents the number of moving average items, x t-i represents the first initial data at the ti-th moment; extracting feature values ​​from the first initial data refers to extracting key features such as time features, spatial features and environmental features; the time features include time periods of days, weeks and months, reflecting the periodicity of personnel flow; the spatial features include the density and distribution of personnel in different areas of the building; the environmental features include external environmental factors such as weather and temperature;

[0064] The smoothing calculation of the second initial data is an exponential smoothing algorithm. The second initial data after the exponential smoothing calculation reduces the fluctuation of the real-time weather forecast data and improves the stability of the weather forecast data. The formula of the exponential smoothing method is:

[0065] S t =αY t +(1-α)S t-1

[0066] Among them, S t represents the smoothed value of the weather forecast data at time t, α is the smoothing coefficient, and Y t represents the value of the initial weather forecast data at time t; St-1 represents the value of the initial weather forecast data at time t-1; the second initial data extraction feature refers to extracting time features and environmental features from the initial weather forecast data; the time features include daily, weekly, and seasonal changes, reflecting the periodicity of the weather; the environmental features include current temperature, humidity, wind speed, etc., as well as historical weather data; feature engineering technology is used to extract the above feature values ​​from the initial weather forecast data to provide high-quality input data for model training;

[0067] Based on the above processing, the first initial data and the second initial data are cleaned, preprocessed, smoothed, and feature values ​​are extracted to obtain a first initial data input set and a second initial data input set;

[0068] Performing random proportion division on the first initial data input set and the second initial data input set respectively;

[0069] Preferably, a random ratio of 8:2 is adopted; the first initial data input set is divided into a training data set and a verification data set of the first initial data input set according to the random ratio; similarly, the second initial data input set is divided into a training data set and a verification data set of the second initial data input set according to the random ratio; the training data set and the verification data set are used for model training and verification;

[0070] In model training, the training data set is used to train the model and optimize its parameters. The validation data set is used to evaluate the model and the hyperparameters of the model are adjusted according to the performance of the validation data set. The training and evaluation are iterated continuously until the model achieves satisfactory results on the validation data set. By using the training data set to optimize the model during the training process and using the validation data set to monitor and adjust the model, it is ensured that the model can generalize well on new data, and finally a model with stable performance in practical applications can be obtained.

[0071] In machine learning training, after a large amount of statistical analysis of experimental data, the random ratio of 8:2 is used to divide the training data set and the validation data set, which has better results. The advantages and rationality are mainly reflected in:

[0072] Sufficient training data set: 80% of the data is used to train the model, which ensures that the model has enough data to learn the features in the data and reduce the risk of overfitting;

[0073] Moderate validation data set: 20% of the data is used to verify the performance of the model. This ratio provides sufficient information for model parameter adjustment without causing too much deviation in model performance evaluation.

[0074] Balance: In model training and validation, the random ratio of 8:2 is a balanced and reasonable choice; it will not lead to unstable evaluation results due to a small validation data set, nor will it lead to poor model performance due to a small training data set;

[0075] S103, training a first training model and a second training model based on an LSTM neural network with time series output, and outputting first prediction data and second prediction data respectively through the first training model and the second training model;

[0076] First, a model structure is prefabricated. Models have many types and different structures, which are suitable for different tasks and scenarios. A model structure is designed and prefabricated according to the specific task of this embodiment. However, at this time, the parameters in the model structure are random or original and are not suitable for the data to be predicted in this embodiment.

[0077] The model structure designed and prefabricated by the present invention is a first training model and a second training model based on an LSTM neural network, and the first training model and the second training model are trained respectively to generate first prediction data and second prediction data with time series output;

[0078] The first training model and the second training model of the original model structure are trained on the training data set and the verification data set of the 8:2 ratio respectively. The training process uses the data of the training data set and the verification data set at the same time, that is, first extract the features of a wave of data on the training data set, and then generate a set of corresponding parameters, and then use the generated set of parameters to test the effect on the verification data set. If the effect is good, continue to optimize in this direction. If the effect is not good, discard this set of parameters and optimize in the opposite direction. Then, select a group of data on the training data set and the verification data set and repeat this process continuously, and finally find a set of parameters that are most suitable for the data set of this embodiment, that is, a set of parameters obtained on the training data set and perform better on the verification data set. This is the trained model finally obtained in this embodiment (that is, the first training model and the second training model based on the LSTM neural network with time series output are trained). The trained first training model and the second training model are essentially the model structure or the original model structure, but all the parameters are replaced by the parameters trained by this embodiment.

[0079] The time series model is a model for analyzing and predicting data that changes over time. The model uses time as a parameter. The model predicts by studying the time dependence and regularity of the first initial data input set and the second initial data input set. Time series data are observations arranged in chronological order, such as daily temperature, stock prices, economic indicators, sensor data, etc.

[0080] Based on the LSTM neural network algorithm, the long-term dependencies in time series data are captured to provide high-precision prediction results. The prediction model continuously optimizes and adjusts parameters by training historical data to improve the accuracy and generalization of predictions, obtain the prediction results of personnel density and environmental changes in the future, and provide a basis for control decisions;

[0081] The training process refers to using a machine learning algorithm to perform real-time time series prediction on the first training model and the second training model based on the LSTM neural network according to the first initial data input set and the second initial data input set;

[0082] Repeat steps S101 and S102 to collect personnel flow data, indoor and outdoor environmental parameter data, and weather forecast data of the operating environment in the target building, or / and obtain sufficient first initial data and second initial data after a period of data collection, clean, pre-process, smooth and extract characteristic values ​​of the first initial data and the second initial data, and gradually establish sufficient first initial data input set and second initial data input set;

[0083] The training process of the first training model refers to optimizing the parameters of the first training model according to the training data set in the first initial data input set; the parameter optimization refers to optimizing the parameters of the first training model using cross-validation and grid search technology to improve the parameter optimization degree and prediction accuracy of the first training model; the first training model is a training model based on an LSTM neural network, that is, a long short-term memory network training model (LSTM), and the formula of the long short-term memory network training model (LSTM) is:

[0084] h t1 =LSTM(x t1 ,h t1-1 )

[0085] Among them, h t1 represents the hidden state of the first initial data at time t1, x t1 represents the first initial data input value, h t1-1Represents the hidden state of the first initial data at the previous moment; the time series real-time prediction refers to inputting the training data set in the first initial data input set into the first training model that is gradually trained and improved, and outputting the first prediction data, the first prediction data is the personnel flow data and indoor and outdoor environmental parameter data in the future period, that is, the personnel flow prediction results and indoor and outdoor environmental parameter prediction results in the future period; the first prediction data is used for subsequent air-conditioning intelligent adjustment and control; finally, the personnel flow prediction results and indoor and outdoor environmental parameter prediction results are displayed in the form of visual charts, including time series charts, heat maps, etc.; the personnel flow prediction results include personnel density, flow trends and peak hours in the future period, etc.; the indoor and outdoor environmental parameter prediction results include indoor and outdoor environmental parameter data in the future period, including indoor and outdoor temperature, humidity, air quality (such as PM2.5 concentration) and carbon dioxide concentration, etc.; through data visualization technology, users can intuitively understand and use the prediction results, which can be applied to office buildings, shopping centers, In large public places such as airports and stations, the operating parameters of the air-conditioning system are optimized to improve energy efficiency and comfort by predicting the flow of people and changes in indoor and outdoor environments in different time periods. The theoretical basis includes time series analysis and machine learning technology. The long short-term memory network (LSTM) training model is used for high-precision prediction. Through mathematical training models and algorithms, the flow of people in buildings and changes in indoor and outdoor environments can be predicted efficiently and accurately, providing a solid foundation for the optimization control strategy of the air-conditioning intelligent system. The cross-validation refers to improving the stability of model evaluation through multiple data divisions, and the grid search technology optimization refers to optimizing hyperparameters. The combination of the two can significantly improve the performance and generalization ability of the model. The long short-term memory network (LSTM) training model has the ability to handle long-term dependencies and effectively captures long-term and short-term features in sequence data through memory and forgetting mechanisms. It is suitable for tasks such as time series prediction and natural language processing. Its advantages include reducing the gradient vanishing problem, processing variable-length input sequences, and performing well in complex time series patterns.

[0086] The training process of the second training model refers to optimizing the parameters of the second training model according to the training data set in the second initial data input set; the parameter optimization refers to optimizing the parameters of the second training model using cross-validation and grid search technology to improve the parameter optimization degree and prediction accuracy of the second training model; the second training model is a training model based on LSTM neural network, that is, long short-term memory network training model (LSTM), and the formula of long short-term memory network training model (LSTM) is:

[0087] h t2 =LSTM(x t2 ,h t2-1 )

[0088] Among them, h t2 represents the hidden state of the second initial data at time t2, x t2 represents the second initial data input value, h t2-1 Indicates the hidden state of the second initial data at the previous moment; the time series real-time prediction refers to inputting the training data set in the second initial data input set into the second training model that is gradually trained and improved, and outputting the second prediction data, the second prediction data is the weather forecast result for a period of time in the future, including the temperature, humidity, rainfall, wind speed, temperature change rate, humidity change trend and rainfall probability of the location of the building for a period of time in the future, and the weather forecast result is used for the subsequent air conditioning intelligent adjustment control strategy;

[0089] The model training process is performed on both the training data set and the validation data set. This process is iterated continuously, for example, 20,000 times of training. The model with the best performance after training is the final result model obtained in this embodiment. Then, the result model is used to input the current operating environment parameters of the air conditioner to obtain the prediction data.

[0090] In the weather prediction stage, the real-time weather forecast data and extracted feature values ​​are input into the trained second training model (i.e., the prediction training model) to output the future weather change trend. The prediction process uses the optimized training model and combines real-time data to perform high-precision weather prediction, and the obtained weather prediction results can be used for subsequent air conditioning intelligent adjustment control strategies.

[0091] In the result output stage, the weather forecast results are displayed in the form of visual charts, including the forecast of weather changes within the next 24 hours inside and outside the building that directly affects the air conditioning operating environment. The specific contents include temperature, humidity, rainfall probability, wind speed, etc. The forecast results are displayed in the form of charts using data visualization technology to facilitate user understanding and use.

[0092] Through the above detailed description and workflow, future weather changes can be predicted efficiently and accurately, providing data support for the intelligent control strategy of the air-conditioning system, achieving the goals of energy saving, carbon reduction and comfort optimization;

[0093] S104, optimizing the first prediction data and the second prediction data according to a multi-objective optimization control strategy and generating secondary input data;

[0094] The first prediction data includes population density, flow trend and peak hours, etc.; the second prediction data includes temperature change rate, humidity change trend and rainfall probability, etc. in weather prediction data;

[0095] S104 can firstly customize the strategy customization function, and the administrator can customize the air conditioning multi-objective optimization control strategy requirements, which can not only realize fully automatic unsupervised control, but also customize the subsequent air conditioning control strategy through automatic or semi-automatic mode, which can not only realize fully automatic unsupervised control, but also manually modify the control range or degree in the administrator interface when necessary, so as to set or select the range for subsequent multi-objective optimization;

[0096] The multi-objective optimization control strategy refers to the system using a multi-objective optimization control algorithm based on the crowd flow prediction data and weather forecast data, and based on the prediction results, automatically adjusting the operating parameters of the air-conditioning system, including temperature, humidity, and wind speed, etc., taking into account energy efficiency and indoor comfort, and the optimization algorithm taking into account energy efficiency and indoor comfort, achieving balance through weight adjustment, and generating secondary input data, i.e., generating the operating parameters for adjusting the air-conditioning system;

[0097] The goal of the optimization control strategy is to reduce energy consumption and improve comfort. The optimization process is to train and optimize parameters such as temperature setting value, wind speed setting value and running time through historical data. The calculation formula (objective function) of the multi-objective optimization control strategy is:

[0098] min(α·E+β·C)

[0099] Among them, E represents energy consumption, C represents comfort deviation, α and β are weight coefficients used to balance energy consumption and comfort;

[0100] S105, providing real-time feedback and dynamically adjusting the operating parameters of the air conditioner according to the secondary input data;

[0101] After the optimization control strategy (objective function) is determined, the optimization control strategy is sent to the air-conditioning equipment through a standardized interface (such as BACnet and Modbus). The system dynamically adjusts the operating parameters of the air-conditioning in real time and enters the execution stage. The air-conditioning equipment dynamically adjusts the temperature, humidity, wind speed and other settings of the air-conditioning equipment according to the control instructions of the multi-objective optimization control strategy received (that is, the air-conditioning equipment according to the received real-time feedback parameters) to ensure that the system operates in the best state;

[0102] During the execution process, the system monitors the operating status and environmental parameters of the air conditioner in real time, and provides feedback and dynamic adjustment to the air conditioner through sensor feedback data. The formula for the feedback and dynamic adjustment is as follows:

[0103]

[0104] Among them, ΔP represents the adjustment amount, K p ,K i , K dare proportional, integral and differential coefficients respectively, SP represents the set value, PV represents the process variable, and dt represents the derivative at time t;

[0105] S106, strengthening self-learning, improving the multi-objective optimization control strategy, and entering step S104;

[0106] The execution phase also includes enhanced self-learning, which refers to continuously improving the multi-objective optimization control strategy by analyzing historical data and current data, improving the adaptability and intelligence level of the system, and using enhanced self-learning to gradually improve the accuracy and robustness of the control model. The algorithm formula for enhanced self-learning is:

[0107] Q(s,a)←Q(s,a)+α(r+γmax a Q(s′,a′)-Q(s,a))

[0108] Where Q(s,a) represents the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ and a′ represent the next state and the next action respectively;

[0109] Strengthening self-learning also includes a feedback mechanism, which continuously optimizes and updates the control model parameters by analyzing the system's operating data and control effects. The system identifies the effectiveness and deficiencies of the control strategy based on the comparison of historical data and current data, and makes targeted adjustments and improvements. The core of the feedback mechanism is to continuously adjust the model parameters to improve the system's prediction accuracy and adaptability.

[0110] In terms of theoretical application, enhanced self-learning is widely applicable to the optimization of air conditioning systems and energy conservation and carbon reduction in intelligent buildings. Through self-learning and optimized control, the system can respond to changes in real time under different environments, provide personalized control solutions, improve energy efficiency, reduce carbon emissions, and achieve the goals of intelligent and green buildings.

[0111] Step S105 also includes real-time monitoring, remote management and maintenance management of the air conditioning system, and data analysis and report output;

[0112] The remote management refers to real-time monitoring of the operating status and environmental parameters of the air-conditioning system through the cloud platform and remote operation. The system uses data visualization technology to display the monitoring data in the form of charts and dashboards. Users can view the operating status of the air-conditioning system in real time through web pages or mobile applications. Commonly used visualization tools include time series graphs, heat maps and pie charts, which intuitively display temperature changes, energy consumption and equipment operating status;

[0113] Remote operation means that users and administrators can remotely operate and manually adjust the air conditioning control through the Internet or mobile devices. The system provides a remote control interface, allowing users to send control instructions through mobile applications or web clients at any location. These instructions include temperature setting, wind speed adjustment, mode switching, etc. The remote control interface uses secure encryption technology to ensure the security and reliability of data transmission. The control instructions are transmitted to the local controller through the cloud platform, and the controller performs specific operations;

[0114] The data analysis and output reports refer to the system regularly generating detailed operation reports and optimization suggestions to help users and administrators understand the performance and energy consumption of the system. The report content includes energy consumption statistics of the air-conditioning system, operation efficiency analysis, environmental parameter change trends and energy-saving optimization suggestions, etc. By analyzing these reports, users can identify the peak hours of energy use and the reasons for inefficient operation, and take corresponding measures to optimize.

[0115] This embodiment ensures the accuracy of the initial data by cleaning, preprocessing, smoothing calculation, extracting eigenvalues, etc., and effectively smoothing out short-term fluctuations and noise, making the data trend clearer and reducing the random fluctuations of the initial data. At the same time, the extreme values ​​and abnormal fluctuations in the data are averaged, reducing data uncertainty, improving the stability of the initial data, and providing high-quality input data for model training; by randomly dividing the initial data input set into proportions, the parameter optimization degree and prediction accuracy of the training model are improved through parameter optimization; through data visualization technology, users can intuitively understand and use the prediction results, which can be applied to large public places such as office buildings, shopping centers, airports and stations, and can predict the flow of people in different time periods. The flow of personnel and changes in the indoor and outdoor environment can be used to optimize the operating parameters of the air-conditioning system and improve energy efficiency and comfort. Through time series analysis and machine learning technology, the long short-term memory network (LSTM) training model is used for high-precision prediction. Through mathematical training models and algorithms, the flow of personnel in the building and changes in the indoor and outdoor environment can be predicted efficiently and accurately, providing a solid foundation for the optimization control strategy of the air-conditioning intelligent system. It can efficiently and accurately predict future weather changes, provide data support for the intelligent control strategy of the air-conditioning system, and achieve the goals of energy saving, carbon reduction and comfort optimization. Through multi-objective optimization control strategies, energy efficiency and indoor comfort are comprehensively considered, and balance is achieved through weight adjustment to reduce energy consumption and improve comfort. By strengthening Self-learning analyzes historical data and current data, continuously optimizes control strategies, improves the adaptability and intelligence level of the system, and gradually improves the accuracy and robustness of the control model; analyzes the system's operating data and control effects through a feedback mechanism, continuously optimizes and updates control model parameters, and identifies the effectiveness and deficiencies of control strategies based on the comparison of historical data and current data, and makes targeted adjustments and improvements to improve the system's prediction accuracy and adaptability; in theoretical applications, enhanced self-learning is widely applicable to the optimization of air-conditioning systems and energy conservation and carbon reduction in intelligent buildings. Through self-learning and optimized control, the system can respond to changes in real time under different environments, enabling the system to continuously optimize control strategies based on real-time data, provide personalized control solutions, and improve Energy utilization efficiency, prediction accuracy and control effect, reduce carbon emissions, and achieve the goals of intelligent and green buildings; it also uses machine learning and reinforcement learning technologies to predict and optimize control to achieve the goals of energy saving and improving indoor comfort. The system can collect and analyze the flow of people and environmental data in the building in real time, dynamically optimize the air-conditioning operating parameters, and realize real-time operation and management of the air-conditioning system through remote monitoring, provide fault diagnosis and optimization suggestions, and ensure the efficient operation and stability of the system. Experimental evaluation shows that the air-conditioning intelligent control system significantly reduces energy consumption in practical applications and improves the comfort of the indoor environment; this embodiment has been proved through system experiments that the cooling energy consumption of the building can be reduced by 58.5%, and the power consumption of the air-conditioning system in the lecture hall can be reduced by 36.7%; The present invention realizes the efficient management, intelligent regulation, environmental protection, energy saving and carbon reduction goals of the air-conditioning intelligent control system; through the comprehensive use of advanced artificial intelligence technology and data analysis methods, the intelligent control and dynamic regulation of the air-conditioning system are realized, which significantly improves the energy utilization efficiency and environmental comfort, and has broad application prospects and significant economic benefits; This embodiment significantly reduces energy consumption and carbon emissions while ensuring indoor comfort, supports the realization of green buildings and environmental protection goals, and can be widely used in large public places such as office buildings, shopping centers, airports and stations, providing strong support for energy conservation, emission reduction and environmental protection. .

[0116] Embodiment 2

[0117] like Figure 3 The figure shows a schematic diagram of the structure of a multi-objective intelligent optimization air-conditioning control system based on multi-dimensional machine learning of the present invention, wherein the system includes a data acquisition module, a data processing module, a model training module, an optimization control module, and a feedback dynamic adjustment module;

[0118] The data acquisition module is connected to the data processing module, and is used to collect first initial data of the building operation environment and second initial data of the weather forecast of the building in real time, and send the first initial data and the second initial data to the data processing module;

[0119] The first initial data is personnel flow data in the building and indoor and outdoor environmental parameter data;

[0120] The data collection module can collect the personnel flow data in the building in real time through cameras and personnel flow sensors. The collection method includes installing high-definition cameras and high-sensitivity sensors at key entrances and main passages to capture and record the number and actions of personnel in real time;

[0121] The data acquisition module can also monitor indoor and outdoor environmental parameter data in real time through devices such as thermometers, hygrometers, air quality sensors and carbon dioxide concentration sensors installed in various key areas of the building. The collected indoor and outdoor environmental parameter data include indoor and outdoor temperature, humidity, air quality (such as PM2.5 concentration) and carbon dioxide concentration, etc. These indoor and outdoor environmental parameter data can reflect the operating effect of the air-conditioning system and the comfort of the indoor and outdoor environment;

[0122] The second initial data is real-time weather forecast data of the location of the building, the weather forecast data is acquired in real time, and the second initial data is combined with the first initial data to optimize the air conditioning control strategy of the building;

[0123] The data acquisition module acquires the second initial data of the building in real time by integrating with the API interface of a third-party weather forecast service provider to acquire the outdoor real-time weather forecast data near the area where the building is located in real time. The weather forecast data includes current and future weather forecast information, such as temperature, humidity, rainfall, wind speed and wind direction, etc., which provides an important reference for the air conditioning control strategy of the building on the impact of external environmental changes on the internal environment of the building;

[0124] The weather forecast data generally includes current weather data and weather forecast information within the next 24 hours;

[0125] The data acquisition module uses standard HTTP requests to obtain real-time weather forecast data in JSON format through the API interface to ensure the real-time and accuracy of the weather forecast data. This part of the real-time weather forecast data is defined as unprocessed initial weather forecast data;

[0126] The data processing module is connected to the data acquisition module and the model training module, and is used to clean, preprocess, smooth and calculate the collected first initial data and second initial data in sequence, extract characteristic values, and generate a first initial data input set and a second initial data input set respectively, and randomly divide the first initial data input set and the second initial data input set into a training data set and a verification data set respectively, and send the training data set and the verification data set in the first initial data input set and the second initial data input set to the model training module respectively;

[0127] The cleaning and preprocessing of the first initial data is to remove noise and abnormal values ​​in the first initial data, such as obviously abnormal temperature and humidity, so as to ensure the accuracy of the first initial data;

[0128] The cleaning and preprocessing of the second initial data is to clean the acquired initial real-time weather forecast data to remove invalid data and abnormal values, such as missing weather data or erroneous weather values;

[0129] The smoothing calculation of the first initial data is a moving average calculation, which effectively smooths out short-term fluctuations and noise, makes the data trend clearer, reduces the random fluctuations of the first initial data, and at the same time, the extreme values ​​and abnormal fluctuations in the data are averaged, reducing data uncertainty and improving the stability of the first initial data;

[0130] The formula for moving average calculation is:

[0131]

[0132] Among them, MA trepresents the moving average of the first initial data at time t, n represents the number of moving average items, x t-i represents the first initial data at the ti-th moment; extracting feature values ​​from the first initial data refers to extracting key features such as time features, spatial features and environmental features; the time features include time periods of days, weeks and months, reflecting the periodicity of personnel flow; the spatial features include the density and distribution of personnel in different areas of the building; the environmental features include external environmental factors such as weather and temperature;

[0133] The smoothing calculation of the second initial data is an exponential smoothing algorithm, which reduces the fluctuation of the real-time weather forecast data and improves the stability of the weather forecast data; the formula of the exponential smoothing method is:

[0134] S t =αY t +(1-α)S t-1

[0135] Among them, S t represents the smoothed value of the weather forecast data at time t, α is the smoothing coefficient, and Y t represents the value of the initial weather forecast data at time t; S t-1 represents the value of the initial weather forecast data at time t-1; the second initial data extraction feature refers to extracting time features and environmental features from the initial weather forecast data; the time features include daily, weekly, and seasonal changes, reflecting the periodicity of the weather; the environmental features include current temperature, humidity, wind speed, etc., as well as historical weather data; feature engineering technology is used to extract the above feature values ​​from the initial weather forecast data to provide high-quality input data for model training;

[0136] Based on the above processing, the first initial data and the second initial data are cleaned, preprocessed, smoothed, and feature values ​​are extracted to obtain a first initial data input set and a second initial data input set;

[0137] Performing random proportion division on the first initial data input set and the second initial data input set respectively;

[0138] The random ratio adopted is 8:2, and the first initial data input set is divided into a training data set and a verification data set of the first initial data input set according to the random ratio. Similarly, the second initial data input set is divided into a training data set and a verification data set of the second initial data input set according to the random ratio; the training data set and the verification data set are used for model training and verification;

[0139] The model training module is connected to the data processing module and the optimization control module, and is used to prefabricate a first training model and a second training model based on an LSTM neural network, respectively train the first training model and the second training model to generate first prediction data and second prediction data with time series output, and send the first prediction data and the second prediction data to the optimization control module;

[0140] The training process refers to using a machine learning algorithm to perform real-time time series prediction on the first training model and the second training model based on the LSTM neural network according to the first initial data input set and the second initial data input set;

[0141] The data collection module collects personnel flow data, indoor and outdoor environmental parameter data and weather forecast data of the operating environment in the target building in real time, or / and obtains sufficient first initial data and second initial data after a period of data collection. The data processing module cleans, pre-processes, smoothes and extracts characteristic values ​​of the first initial data and the second initial data, and gradually establishes sufficient first initial data input set and second initial data input set;

[0142] The training process of the first training model by the model training module refers to optimizing the parameters of the first training model according to the training data set in the first initial data input set; the parameter optimization refers to optimizing the parameters of the first training model using cross-validation and grid search technology to improve the parameter optimization degree and prediction accuracy of the first training model; the first training model is a training model based on LSTM neural network, that is, a long short-term memory network training model (LSTM), and the formula of the long short-term memory network training model (LSTM) is:

[0143] h t1 =LSTM(x t1 ,h t1-1 )

[0144] Among them, h t1 represents the hidden state of the first initial data at time t1, x t1 represents the first initial data input value, h t1-1Represents the hidden state of the first initial data at the previous moment; the time series real-time prediction refers to inputting the training data set in the first initial data input set into the first training model that is gradually trained and improved, and outputting the first prediction data, the first prediction data is the personnel flow data and indoor and outdoor environmental parameter data in the future period, that is, the personnel flow prediction results and indoor and outdoor environmental parameter prediction results in the future period; the first prediction data is used for subsequent air-conditioning intelligent adjustment and control; finally, the personnel flow prediction results and indoor and outdoor environmental parameter prediction results are displayed in the form of visual charts, including time series charts, heat maps, etc.; the personnel flow prediction results include personnel density, flow trends and peak hours in the future period, etc.; the indoor and outdoor environmental parameter prediction results include Indoor and outdoor environmental parameter data for a period of time in the future, including indoor and outdoor temperature, humidity, air quality (such as PM2.5 concentration) and carbon dioxide concentration; through data visualization technology, users can intuitively understand and use the prediction results, which can be applied to large public places such as office buildings, shopping malls, airports and stations. By predicting the flow of people and changes in indoor and outdoor environments in different time periods, the operating parameters of the air-conditioning system can be optimized to improve energy efficiency and comfort; the theoretical basis includes time series analysis and machine learning technology, using the long short-term memory network (LSTM) training model for high-precision prediction. Through mathematical training models and algorithms, it can efficiently and accurately predict the flow of people in buildings and changes in indoor and outdoor environments, providing a solid foundation for the optimization control strategy of air-conditioning intelligent systems;

[0145] The training process of the second training model by the model training module refers to optimizing the parameters of the second training model according to the training data set in the second initial data input set; the parameter optimization refers to optimizing the parameters of the second training model using cross-validation and grid search technology to improve the parameter optimization degree and prediction accuracy of the second training model; the second training model is a training model based on LSTM neural network, that is, a long short-term memory network training model (LSTM), and the formula of the long short-term memory network training model (LSTM) is:

[0146] h t2 =LSTM(x t2 ,h t2-1 )

[0147] Among them, h t2 represents the hidden state of the second initial data at time t2, x t2 represents the second initial data input value, h t2-1Indicates the hidden state of the second initial data at the previous moment; the time series real-time prediction refers to inputting the training data set in the second initial data input set into the second training model that is gradually trained and improved, and outputting the second prediction data, the second prediction data is the weather forecast result for a period of time in the future, including the temperature, humidity, rainfall, wind speed, temperature change rate, humidity change trend and rainfall probability of the location of the building for a period of time in the future, and the weather forecast result is used for the subsequent air conditioning intelligent adjustment control strategy;

[0148] The model training module is also used for weather prediction. The real-time weather forecast data and extracted feature values ​​are input into the trained prediction training model to output the future weather change trend. The prediction process uses the optimized training model and combines the real-time data to perform high-precision weather prediction. The obtained weather prediction results can be used for the subsequent intelligent air conditioning adjustment control strategy.

[0149] The model training module is also used for result output, and the weather forecast results are displayed in the form of visual charts, including the forecast of weather changes in the next 24 hours, including temperature, humidity, rainfall probability, wind speed, etc., and the forecast results are displayed in the form of charts using data visualization technology, which is convenient for users to understand and use;

[0150] Through the above detailed description and workflow, future weather changes can be predicted efficiently and accurately, providing data support for the intelligent control strategy of the air-conditioning system, achieving the goals of energy saving, carbon reduction and comfort optimization;

[0151] The optimization control module is connected to the model training module and the feedback dynamic adjustment module, and is used to optimize the first prediction data and the second prediction data according to the multi-objective optimization control strategy and generate secondary input data, and send the secondary input data to the feedback dynamic adjustment module;

[0152] The first prediction data includes population density, flow trend and peak hours, etc.; the second prediction data includes temperature change rate, humidity change trend and rainfall probability, etc. in weather prediction data;

[0153] The optimization control module can firstly initialize the multi-objective optimization control strategy in an automatic or semi-automatic mode, wherein the multi-objective optimization control strategy refers to comprehensively considering energy efficiency and indoor comfort, achieving balance through weight adjustment, and generating secondary input data, i.e., generating operating parameters for adjusting the air conditioning system;

[0154] The goal of the optimization control strategy is to reduce energy consumption and improve comfort. The optimization process is to train and optimize parameters such as temperature setting value, wind speed setting value and running time through historical data. The calculation formula (objective function) of the multi-objective optimization control strategy is:

[0155] min(α·E+β·C)

[0156] Among them, E represents energy consumption, C represents comfort deviation, α and β are weight coefficients used to balance energy consumption and comfort;

[0157] The algorithm optimizes the operating parameters of the air-conditioning system by comprehensively considering the factors of energy consumption and comfort by establishing an objective function. The system automatically adjusts the values ​​of α and β according to the first and second prediction data obtained in the above steps, thereby dynamically optimizing the operating parameters of the air-conditioning system under different environments and needs. For example, during periods of high traffic, the system may give priority to comfort, while at night or during idle periods, the system tends to save energy.

[0158] The feedback dynamic adjustment module is connected to the optimization control module and is used to provide real-time feedback and dynamically adjust the operating parameters of the air conditioner according to the secondary input data;

[0159] This setting allows us to flexibly adjust the balance between energy consumption and comfort according to the specific needs of the building or the preferences of the users. For example, in an office building, the system may prioritize comfort to improve the work efficiency of employees, while in places such as shopping malls, the system can prioritize energy conservation. In the adaptive process, the system will continuously learn and optimize the control strategy based on historical operation data to improve overall efficiency;

[0160] After the optimization control strategy (objective function) is determined, the optimization control strategy is sent to the air-conditioning equipment through a standardized interface (such as BACnet and Modbus). The system dynamically adjusts the operating parameters of the air-conditioning in real time and enters the execution stage. The air-conditioning equipment dynamically adjusts the temperature, humidity, wind speed and other settings of the air-conditioning equipment according to the control instructions of the multi-objective optimization control strategy received (that is, the air-conditioning equipment according to the received real-time feedback parameters) to ensure that the system operates in the best state;

[0161] During the execution of the feedback dynamic adjustment module, the operating status and environmental parameters of the air conditioner are monitored in real time, and the air conditioner is fed back and dynamically adjusted through sensor feedback data. The formula for the feedback and dynamic adjustment is as follows:

[0162]

[0163] Among them, ΔP represents the adjustment amount, K p ,K i , K d are proportional, integral and differential coefficients respectively, SP represents the set value, PV represents the process variable, and dt represents the derivative at time t;

[0164] Furthermore, the system further comprises a self-enhancing module, which is connected to the feedback dynamic adjustment module and the optimization control module, and is used to strengthen self-learning, improve the multi-objective optimization control strategy, and send the multi-objective optimization control strategy to the optimization control module;

[0165] The optimization control module is also connected to the self-enhancing module, and is used to receive the multi-objective optimization control strategy sent by the self-enhancing module, and optimize the first prediction data and the second prediction data according to the multi-objective optimization control strategy and generate secondary input data, and send the secondary input data to the feedback dynamic adjustment module;

[0166] The enhanced self-learning refers to continuously improving the multi-objective optimization control strategy by analyzing historical data and current data, improving the adaptability and intelligence level of the system, and using enhanced self-learning to gradually improve the accuracy and robustness of the control model. The algorithm formula for enhanced self-learning is:

[0167] Q(s,a)←Q(s,a)+α(r+γmax a Q(s′,a′)-Q(s,a))

[0168] Where Q(s,a) represents the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ and a′ represent the next state and the next action respectively;

[0169] Strengthening self-learning also includes a feedback mechanism, which continuously optimizes and updates the control model parameters by analyzing the system's operating data and control effects. The system identifies the effectiveness and deficiencies of the control strategy based on the comparison of historical data and current data, and makes targeted adjustments and improvements. The core of the feedback mechanism is to continuously adjust the model parameters to improve the system's prediction accuracy and adaptability.

[0170] In terms of theoretical application, enhanced self-learning is widely applicable to the optimization of air conditioning systems and energy conservation and carbon reduction in intelligent buildings. Through self-learning and optimized control, the system can respond to changes in real time under different environments, provide personalized control solutions, improve energy efficiency, reduce carbon emissions, and achieve the goals of intelligent and green buildings.

[0171] Furthermore, the optimization control module is also used for real-time monitoring, remote management and maintenance management of the air conditioning system, and for data analysis and output reports;

[0172] The remote management refers to real-time monitoring of the operating status and environmental parameters of the air-conditioning system through the cloud platform and remote operation. The system uses data visualization technology to display the monitoring data in the form of charts and dashboards. Users can view the operating status of the air-conditioning system in real time through web pages or mobile applications. Commonly used visualization tools include time series graphs, heat maps and pie charts, which intuitively display temperature changes, energy consumption and equipment operating status;

[0173] Remote operation means that users and administrators can remotely operate and manually adjust the air conditioning control through the Internet or mobile devices. The system provides a remote control interface, allowing users to send control instructions through mobile applications or web clients at any location. These instructions include temperature setting, wind speed adjustment, mode switching, etc. The remote control interface uses secure encryption technology to ensure the security and reliability of data transmission. The control instructions are transmitted to the local controller through the cloud platform, and the controller performs specific operations;

[0174] The data analysis and output reports refer to the system regularly generating detailed operation reports and optimization suggestions to help users and administrators understand the performance and energy consumption of the system. The report content includes energy consumption statistics of the air-conditioning system, operation efficiency analysis, environmental parameter change trends and energy-saving optimization suggestions. By analyzing these reports, users can identify the peak hours of energy use and the reasons for inefficient operation, and take corresponding measures to optimize;

[0175] This embodiment can respond to changes in real time under different environments, dynamically adjust control parameters, ensure the best operating effect, and is widely used in large public places such as office buildings, shopping centers, airports, and stations, providing strong support for energy conservation, emission reduction, and environmental protection of buildings; through advanced artificial intelligence and data analysis technology, efficient, intelligent, and energy-saving air-conditioning control is achieved, which can effectively reduce energy consumption and carbon emissions;

[0176] This embodiment includes a data acquisition module, a data processing module, a model training module, an optimization control module, and a feedback dynamic adjustment module. The data acquisition module collects the first initial data of the building operation environment and the second initial data of the weather forecast of the building in real time to achieve an accurate control strategy for the dual goals of energy saving and comfort. The data processing module performs cleaning, preprocessing, smoothing calculation, and feature value extraction on the collected initial data in sequence to ensure the accuracy of the initial data, effectively smooth out short-term fluctuations and noise, make the data trend clearer, and reduce the random fluctuations of the initial data. At the same time, the extreme values ​​and abnormal fluctuations in the data are averaged, which reduces data uncertainty, improves the stability of the initial data, and provides high-quality input data for model training. The initial data input set is randomly divided into proportions to respectively train and generate first prediction data and second prediction data with time series output. The optimization control module optimizes the first prediction data and the second prediction data according to the multi-objective optimization control strategy and generates secondary input data. The parameter optimization improves the parameter optimization degree and prediction accuracy of the training model. The feedback dynamic adjustment module feeds back and dynamically adjusts the operating parameters of the air conditioner in real time according to the secondary input data. Through data visualization technology, users can intuitively understand and use the prediction results , can be applied to large public places such as office buildings, shopping malls, airports and stations. By predicting the flow of people and changes in indoor and outdoor environments in different time periods, the operating parameters of the air-conditioning system can be optimized to improve energy efficiency and comfort. Through time series analysis and machine learning technology, the long short-term memory network (LSTM) training model is used for high-precision prediction. Through mathematical training models and algorithms, the flow of people in buildings and changes in indoor and outdoor environments can be predicted efficiently and accurately, providing a solid foundation for the optimization control strategy of the air-conditioning intelligent system. It can efficiently and accurately predict future weather changes and provide data support for the intelligent control strategy of the air-conditioning system. The goals of energy conservation, carbon reduction and comfort optimization are achieved; through multi-objective optimization control strategies, energy efficiency and indoor comfort are comprehensively considered, and balance is achieved through weight adjustment to reduce energy consumption and improve comfort; through strengthening self-learning to analyze historical data and current data, the control strategy is continuously optimized, the adaptability and intelligence level of the system are improved, and the accuracy and robustness of the control model are gradually improved; through the feedback mechanism, the system's operating data and control effects are analyzed, and the control model parameters are continuously optimized and updated. The system identifies the effectiveness and shortcomings of the control strategy based on the comparison of historical data and current data, and makes targeted adjustments and improvements to improve the prediction accuracy and adaptability of the system;In theoretical application, reinforced self-learning is widely used in air-conditioning system optimization and energy conservation and carbon reduction in intelligent buildings. Through self-learning and optimized control, the system can respond to changes in real time under different environments, so that the system can continuously optimize the control strategy according to real-time data, provide personalized control solutions, improve energy utilization efficiency, prediction accuracy and control effect, reduce carbon emissions, and achieve the goals of intelligent and green buildings; it also uses machine learning and reinforcement learning technology to predict and optimize control to achieve the goals of energy saving and improving indoor comfort. The system can collect and analyze the flow of people and environmental data in the building in real time, dynamically optimize air-conditioning operating parameters, realize real-time operation and management of the air-conditioning system through remote monitoring, provide fault diagnosis and optimization suggestions, and ensure the efficient operation and stability of the system. Experimental evaluation shows that the air-conditioning intelligent control system is very effective in actual application. The energy consumption is significantly reduced during use, and the comfort of the indoor environment is improved; this embodiment has been proved through system experiments that the cooling energy consumption of the building can be reduced by 58.5%, and the power consumption of the air conditioning system in the lecture hall can be reduced by 36.7%; the present invention realizes the efficient management, intelligent adjustment, environmental protection, energy saving and carbon reduction goals of the air conditioning intelligent control system; through the comprehensive use of advanced artificial intelligence technology and data analysis methods, the intelligent control and dynamic adjustment of the air conditioning system are realized, and the energy utilization efficiency and environmental comfort are significantly improved, which has a wide application prospect and significant economic benefits; this embodiment significantly reduces energy consumption and carbon emissions while ensuring indoor comfort, supports the realization of green buildings and environmental protection goals, and can be widely used in large public places such as office buildings, shopping centers, airports and stations, providing strong support for energy conservation, emission reduction and environmental protection. ;

[0177] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning, the method comprising the steps of: Collecting in real time first initial data of the building operating environment and second initial data of the weather forecast of the building; The collected first initial data and second initial data are cleaned, preprocessed, smoothed, and feature values ​​are extracted in sequence to generate a first initial data input set and a second initial data input set, and the first initial data input set and the second initial data input set are randomly divided into a training data set and a validation data set, respectively; Training a first training model and a second training model based on an LSTM neural network with a time series output and outputting first prediction data and second prediction data respectively; Optimizing the first prediction data and the second prediction data according to a multi-objective optimization control strategy and generating secondary input data; The operating parameters of the air conditioner are adjusted dynamically and in real time based on the secondary input data.

2. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 1 is characterized in that: The first initial data is personnel flow data in the building and indoor and outdoor environmental parameter data; The second initial data is real-time weather forecast data for the location of the building.

3. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 1 or 2, characterized in that: The smoothing calculation formula of the first initial data is: Among them, MA t represents the moving average of the first initial data at time t, n represents the number of moving average items, x t-i Represents the first initial data at time ti.

4. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 3 is characterized in that: The smoothing calculation formula of the second initial data is: S t =αY t +(1-α)S t-1 Among them, S t represents the smoothed value of the weather forecast data at time t, α is the smoothing coefficient, and Y t Represents the value of the initial weather forecast data at time t.

5. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 3 or 4, characterized in that: The formula for the first training model based on the LSTM neural network is: h t1 =LSTM(x t1 ,h t1-1 ) Among them, h t1 represents the hidden state of the first initial data at time t1, x t1 represents the first initial data input value, h t1-1 Represents the hidden state of the first initial data at the previous moment.

6. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 5 is characterized in that: The formula for the second training model based on the LSTM neural network is: h t2 =LSTM(x t2 ,h t2-1 ) Among them, h t2 represents the hidden state of the second initial data at time t2, x t2 represents the second initial data input value, h t2-1 Represents the hidden state of the second initial data at the previous moment.

7. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 5 or 6, characterized in that: The calculation formula of the multi-objective optimization control strategy is: min(α·E+β·C) Among them, E represents energy consumption, C represents comfort deviation, and α and β are weight coefficients.

8. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 7 is characterized in that: The feedback and dynamic adjustment formula is as follows: Among them, ΔP represents the adjustment amount, K p ,K i , K d They are proportional, integral and differential coefficients respectively, SP represents the set value, and PV represents the process variable.

9. The multi-objective intelligent optimization air conditioning control method based on multi-dimensional machine learning according to claim 7 or 8, characterized in that: The step of "real-time feedback based on the secondary input data, dynamically adjusting the operating parameters of the air conditioner" also includes the following steps: Strengthen self-learning, improve the multi-objective optimization control strategy, and enter the step of "optimizing the first prediction data and the second prediction data according to the multi-objective optimization control strategy and generating secondary input data"; The algorithm formula for strengthening self-learning is: Q(s,a)←Q(s,a)+α(r+γmax a Q(s′,a′)-Q(s,a)) Among them, Q(s,a) represents the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ and a′ represent the next state and the next action respectively.

10. A multi-objective intelligent optimization air conditioning control system based on multi-dimensional machine learning, characterized in that: The system includes a data acquisition module, a data processing module, a model training module, an optimization control module, and a feedback dynamic adjustment module; The data acquisition module is connected to the data processing module, and is used to collect first initial data of the building operation environment and second initial data of the weather forecast of the building in real time, and send the first initial data and the second initial data to the data processing module; The data processing module is connected to the data acquisition module and the model training module, and is used to clean, preprocess, smooth and calculate the collected first initial data and second initial data in sequence, extract characteristic values, and generate a first initial data input set and a second initial data input set respectively, and randomly divide the first initial data input set and the second initial data input set into a training data set and a verification data set respectively, and send the training data set and the verification data set in the first initial data input set and the second initial data input set to the model training module respectively; The model training module is connected to the data processing module and the optimization control module, and is used to prefabricate a first training model and a second training model based on an LSTM neural network, respectively train the first training model and the second training model to generate first prediction data and second prediction data with time series output, and send the first prediction data and the second prediction data to the optimization control module; The optimization control module is connected to the model training module and the feedback dynamic adjustment module, and is used to optimize the first prediction data and the second prediction data according to the multi-objective optimization control strategy and generate secondary input data, and send the secondary input data to the feedback dynamic adjustment module; The feedback dynamic adjustment module is connected to the optimization control module, and is used for real-time feedback and dynamic adjustment of the operating parameters of the air conditioner according to the secondary input data.

Images