An intelligent control method, system, device and medium

Abstract

The application relates to the technical field of intelligent control, and discloses an intelligent control method, system, device and medium, which is characterized in that real-time state data and corresponding real-time refrigeration power data of a heating ventilation air conditioner are input into a regression model based on a gradient boosting tree for multi-scene prediction, the temperature variation corresponding to each wind speed gear is obtained, and the real-time prediction error is quantified according to the temperature variation; a first residual error of an active wind speed gear is strongly updated at a first update rate, and a second residual error of an inactive wind speed gear is weakly updated at a second update rate which is smaller than the first update rate; the temperature variation obtained through prediction is corrected according to the strong update result and the weak update result obtained through updating; the final predicted temperature corresponding to each wind speed gear is obtained; a target function is constructed and solved; a target control instruction is generated and used to control the heating ventilation air conditioner to execute; and the heating ventilation air conditioner is accurately, energy-efficiently and stably controlled.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to an intelligent control method, system, device and medium. Background Technology

[0002] Heating, ventilation, and air conditioning (HVAC) systems are a core component of building energy consumption, and their control performance directly affects energy costs and indoor comfort. Currently, most existing HVAC control methods employ reactive control strategies based on fixed setpoints (such as PID control). However, these methods exhibit significant lag and struggle to cope with dynamically changing external environments, resulting in low accuracy in indoor temperature control. Although some advanced control strategies based on predictive models have emerged in existing technologies, these predictive models accumulate errors over long-term operation, making it difficult to adapt to system performance degradation or long-term changes in environmental characteristics.

[0003] Therefore, improving the control accuracy of existing HVAC control methods has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] This invention provides an intelligent control method, system, device, and medium to address the problem of improving the control accuracy of existing HVAC control methods.

[0005] To address the aforementioned technical problems, the first aspect of this invention provides an intelligent control method, comprising:

[0006] Acquire real-time status data of HVAC system, and quantify the effective cooling potential of HVAC system based on the real-time status data to obtain real-time cooling power data;

[0007] The real-time status data and the real-time cooling power data are input into a gradient boosting tree-based regression model for multi-scenario prediction to obtain the temperature change corresponding to each wind speed level.

[0008] The real-time prediction error of each wind speed level is calculated, and a strong update is performed on the first residual corresponding to the active wind speed level at a first update rate to obtain a strong update result. A weak penetration update is performed on the second residual corresponding to the inactive wind speed level at a second update rate to obtain a weak penetration update result. The temperature change is then corrected by combining the strong update result to obtain the final predicted temperature corresponding to each wind speed level. The first update rate is greater than the second update rate.

[0009] Based on the final predicted temperatures, an objective function is constructed and solved to obtain the target control command to control the HVAC system.

[0010] As one preferred embodiment, the real-time status data includes equipment operation data and building structure data; wherein,

[0011] The process of quantifying the effective cooling potential of the HVAC system based on the real-time status data to obtain real-time cooling power data includes:

[0012] Based on the equipment operation data, the fan speed efficiency coefficient and water supply temperature efficiency coefficient of the HVAC system are determined through a multi-parameter dynamic coupling correction mechanism, and the maximum sensible heat cooling capacity data of the HVAC system is extracted from the building structure data.

[0013] Based on the fan speed efficiency coefficient, the water supply temperature efficiency coefficient, and the maximum sensible heat cooling capacity data, the effective cooling potential of the HVAC system is quantified to obtain real-time cooling power data.

[0014] As one preferred embodiment, the step of inputting the real-time status data and the real-time cooling power data into a gradient boosting tree-based regression model for multi-scenario prediction to obtain the temperature change corresponding to each wind speed setting includes:

[0015] State features and power features are extracted from the real-time status data and the real-time cooling power data, respectively, to construct a feature input set;

[0016] Choose any wind speed setting as the target wind speed setting, construct a target virtual scenario corresponding to the target wind speed setting based on digital twin technology, and input the feature input set into the gradient boosting tree-based regression model in the target virtual scenario to make a prediction and obtain the target temperature change corresponding to the target wind speed setting.

[0017] The target wind speed setting is updated to iteratively execute the construction process of the target virtual scenario and the temperature prediction process in the target virtual scenario based on the updated target wind speed setting, until all wind speed settings are iterated to obtain the temperature change corresponding to each wind speed setting.

[0018] As one preferred embodiment, the real-time status data further includes real-time meteorological data; wherein,

[0019] The step of performing a strong update on the first residual corresponding to the active wind speed level at a first update rate to obtain a strong update result includes:

[0020] A feature mapping function is constructed based on the real-time meteorological data and the historical meteorological data of the HVAC system, and a first update rate is determined according to the feature mapping function;

[0021] The currently used wind speed setting is taken as the active wind speed setting, and based on the first update rate, the first residual corresponding to the active wind speed setting and the real-time prediction error are updated using the exponential mean index method to obtain a strong update result.

[0022] As one preferred embodiment, the construction of the feature mapping function based on the real-time meteorological data and the historical meteorological data of the HVAC system includes:

[0023] Real-time environmental characteristics are constructed based on the real-time meteorological data, historical environmental characteristics are constructed based on the historical meteorological data, and the real-time fluctuation index of the HVAC system is determined based on the historical environmental characteristics and the real-time environmental characteristics.

[0024] The historical meteorological data are clustered according to time segments and weather types to obtain multiple environmental scenarios. The characteristic mean and standard deviation of each environmental scenario are calculated as the historical normal baseline of the environmental scenario, and the normal deviation index of the real-time environmental characteristics is determined based on the historical normal baseline.

[0025] Calculate the environmental mutation amount of the real-time environmental feature relative to the historical environmental feature, and determine the feature mutation index based on the environmental mutation amount;

[0026] The real-time fluctuation index, the normal deviation index, and the feature mutation index are fused to obtain the feature complexity, and the feature complexity is nonlinearly mapped to obtain the feature mapping function.

[0027] As one preferred embodiment, the step of performing a weak-penetration update on the second residual corresponding to the inactive wind speed level at a second update rate, obtaining a weak-penetration update result, and then combining the strong update result to correct each of the temperature changes to obtain the final predicted temperature corresponding to each of the wind speed levels includes:

[0028] Other wind speed levels besides the active wind speed level are taken as the inactive wind speed level, and an error penetration mechanism is introduced to update the second residual and the real-time prediction error corresponding to the inactive wind speed level based on the second update rate and using the exponential mean index method, so as to obtain a weak penetration update result.

[0029] Based on the correspondence between the active wind speed level, the inactive wind speed level, and each wind speed level, the strong update result and the weak penetration update result are respectively incorporated into the corresponding temperature change amount to obtain the final predicted temperature corresponding to each wind speed level.

[0030] As one preferred embodiment, the step of constructing and solving the objective function based on each of the final predicted temperatures to obtain the target control command includes:

[0031] The objective function is constructed by minimizing the sum of the absolute differences between the final predicted temperature and the target temperature, and the energy required to reach the target temperature. The optimal wind speed setting is obtained by solving the objective function.

[0032] The optimal wind speed setting is compared with the current wind speed setting, and a target control command is generated based on the comparison result.

[0033] A second aspect of the present invention provides an intelligent control system, comprising:

[0034] The data acquisition module is used to acquire real-time status data of the HVAC system and quantify the effective cooling potential of the HVAC system based on the real-time status data to obtain real-time cooling power data.

[0035] The temperature prediction module is used to input the real-time status data and the real-time cooling power data into a regression model based on gradient boosting tree to perform multi-scenario prediction and obtain the temperature change corresponding to each wind speed level.

[0036] The temperature correction module is used to calculate the real-time prediction error of each wind speed level, and to perform a strong update on the first residual corresponding to the active wind speed level at a first update rate to obtain a strong update result, and to perform a weak penetration update on the second residual corresponding to the inactive wind speed level at a second update rate to obtain a weak penetration update result. The weak penetration update result is then combined with the strong update result to correct the temperature change of each wind speed level, thereby obtaining the final predicted temperature corresponding to each wind speed level; the first update rate is greater than the second update rate.

[0037] The instruction generation module is used to construct and solve objective functions based on the final predicted temperatures to obtain target control instructions to control the HVAC system.

[0038] A third aspect of the present invention provides an electronic device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the intelligent control method as described above.

[0039] A fourth aspect of the present invention provides a computer-readable storage medium comprising a stored computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the intelligent control method as described above.

[0040] Compared with the prior art, the beneficial effects of the embodiments of the present invention are as follows:

[0041] By acquiring real-time status data of HVAC systems and quantifying their effective cooling potential, the system directly addresses the blindness of traditional control based on experience, laying the foundation for precise control. Multi-scenario prediction using a gradient boosting tree regression model can anticipate temperature changes at different wind speeds, shifting control from "passive response" to "active prediction," thus enhancing the forward-looking nature of temperature regulation. High-rate strong updates are used for active wind speed settings to quickly correct prediction errors and ensure the accuracy of current temperature control, while low-rate weak penetration updates are used for inactive settings to prevent residuals from becoming invalid due to prolonged lack of updates, thereby improving the overall stability of control. Through real-time data-driven and dynamic model correction—namely, using digital twins (machine learning + physical correlation) and multi-level residual correction—the system significantly improves the intelligence level of air conditioning operation, achieving precise, energy-saving, and stable control of HVAC systems. Attached Figure Description

[0042] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of an intelligent control method provided in a certain embodiment of the present invention;

[0044] Figure 2 This is a structural diagram of an intelligent control system provided in a certain embodiment of the present invention;

[0045] Figure 3 This is a structural diagram of an electronic device provided in a certain embodiment of the present invention;

[0046] Figure label:

[0047] Among them, 10 is the data acquisition module; 20 is the temperature prediction module; 30 is the temperature correction module; 40 is the instruction generation module; 5000 is the electronic device; 5001 is the processor; 5002 is the bus; 5003 is the memory; and 5004 is the transceiver. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings and examples. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0049] In this invention description, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In this invention description, unless otherwise stated, "a plurality of" means two or more. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0050] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is merely for describing specific embodiments and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0051] In one embodiment, such as Figure 1 As shown, the first aspect of the present invention provides an intelligent control method, comprising:

[0052] S1. Obtain real-time status data of the HVAC system, and quantify the effective cooling potential of the HVAC system based on the real-time status data to obtain real-time cooling power data; the real-time status data includes equipment operation data, building structure data and real-time meteorological data;

[0053] Specifically, this invention collects real-time status data of the indoor environment where the HVAC system is located through internal IoT sensor data and external API data. This includes equipment operation data (data that changes during air conditioner operation), such as current fan speed, air conditioner on / off status, supply water temperature, return water temperature, and electric valve status; building structure data (data that remains constant during air conditioner operation), such as room area, building envelope characteristics, indoor set temperature, rated power data of the air conditioning equipment, and sensible heat cooling capacity data; and real-time meteorological data (i.e., environmental data), such as indoor and outdoor temperature, indoor and outdoor humidity, outdoor light intensity, and outdoor wind speed. It should be noted that the "collected data" listed here is merely an example and not an exhaustive list; the actual types of data collected are much more diverse.

[0054] The acquired data from different sources are then synchronized and aligned according to a unified timestamp, and data cleaning and other processes are performed to improve data accuracy.

[0055] In one embodiment, quantifying the effective cooling potential of the HVAC system based on the real-time status data to obtain real-time cooling power data includes:

[0056] Based on the equipment operation data, the fan speed efficiency coefficient and water supply temperature efficiency coefficient of the HVAC system are determined through a multi-parameter dynamic coupling correction mechanism, and the maximum sensible heat cooling capacity data of the HVAC system is extracted from the building structure data.

[0057] Based on the fan speed efficiency coefficient, the water supply temperature efficiency coefficient, and the maximum sensible heat cooling capacity data, the effective cooling potential of the HVAC system is quantified to obtain real-time cooling power data.

[0058] Specifically, this invention uses the fan speed efficiency coefficient and the water supply temperature efficiency coefficient to reflect the impact of fan speed and water supply temperature on the cooling capacity of air conditioning, respectively. These two coefficients are combined with the maximum sensible heat cooling capacity data of the fan coil unit (FCU) in the air conditioner. This breaks through the limitations of traditional methods that only use the rated power of the equipment to estimate the cooling capacity. By integrating the inherent performance of the hardware and the influence of operating parameters in real time, the theoretical cooling upper limit of the fan coil unit is accurately quantified, providing reliable basic parameters for subsequent temperature prediction, thereby ensuring the accuracy of temperature change prediction.

[0059] This invention breaks through the simple logic of traditional linear mapping or fixed gear correspondence, and introduces a multi-parameter dynamic coupling correction mechanism to determine the fan speed efficiency coefficient and water supply temperature efficiency coefficient by combining equipment operating status, environmental interference, and historical best efficiency data.

[0060] This invention uses the rated water supply temperature of the fan coil unit (equipment factory parameter) as a benchmark and employs an exponential square form to represent the accelerated decay characteristic of the fan coil unit's cooling efficiency as the water supply temperature increases. In other words, it determines the basic efficiency coefficient of the fan coil unit based on equipment operating data. This process is expressed by the following formula:

[0061] η b =e [-k1×(Ts-Tsr)²]

[0062] In the formula, η b k1 is the basic efficiency coefficient; k1 is the attenuation coefficient (pre-configured, preferably 0.05). (T) can also be determined by fitting historical data; s The real-time water supply temperature of the fan coil unit; T sr This refers to the rated water supply temperature of the fan coil unit.

[0063] The supply-return water temperature difference is introduced to reflect the actual heat exchange efficiency of the FCU. Based on the principle that the larger the temperature difference, the more complete the heat exchange, and the weakening of the actual influence of the supply water temperature, a heat exchange correction coefficient is determined, which is expressed by the following formula:

[0064] η d =1-k2×e (-k3×ΔTw)

[0065] In the formula, η d k1 is the heat exchange correction coefficient; k2 is the correction weight, preferably 0.2; k3 is the temperature difference sensitivity coefficient, preferably 0.3. ); ΔTw is the difference between the supply water temperature and the return water temperature.

[0066] The difference between outdoor and indoor set temperatures reflects the interference of environmental heat load on water supply temperature efficiency. Based on the principle that a higher environmental load requires a lower water supply temperature to ensure efficiency, an environmental load correction coefficient is determined, expressed by the following formula:

[0067] η e =1 / [1+k4×max(ΔTe,0)]

[0068] In the formula, η e is the environmental load correction factor; k4 is the environmental sensitivity factor, preferably 0.02. ΔTe is the difference between the outdoor temperature and the indoor set temperature.

[0069] The water supply temperature efficiency coefficient can be obtained by multiplying the basic efficiency coefficient, the heat exchange correction coefficient, and the environmental load correction coefficient. The calculation process innovatively integrates three dimensions: heat exchange temperature difference, environmental load, and historical best efficiency. Through a multi-parameter dynamic coupling correction mechanism, it solves the problem of the traditional method relying solely on the linear deviation between the indoor set temperature and the rated temperature, making the quantification of the effective cooling potential of the air conditioner more accurate.

[0070] This invention is based on the ratio of wind speed to the rated maximum wind speed. Based on the principle that the higher the wind speed, the smaller the efficiency gain per unit increase in wind speed, this invention calculates the wind speed gradient coefficient by introducing a diminishing marginal utility factor. This process is expressed by the following formula:

[0071] η w =[1-e (-k5×Fcur / Fmax) ] / [1-e (-k5) ]

[0072] In the formula, η w is the wind speed gradient coefficient; k5 is the marginal benefit coefficient, preferably 3.0; F cur For wind speed; F max This is the rated maximum wind speed.

[0073] Based on the principle that wind speed's promoting effect on heat exchange is weakened under high humidity, and that water vapor condensation hinders heat transfer, an air disturbance correction coefficient is obtained by introducing changes in indoor humidity to reflect the actual disturbance effect of airflow. This process is expressed by the following formula:

[0074] η db =1-k6×(ΔH in / Δt)²

[0075] In the formula, η db ΔH is the air disturbance correction factor; k6 is the humidity disturbance factor, preferably 0.01 min² / %²; in Δt represents the change in indoor humidity between the current cycle and the previous cycle; Δt is the time interval.

[0076] By combining the energy consumption growth rate corresponding to wind speed, and avoiding a surge in energy consumption due to blindly increasing wind speed in pursuit of efficiency, the energy consumption balance coefficient is calculated, which is expressed by the following formula:

[0077] η ey =1 / [1+k7×(ΔE / ΔF)×F cur ]

[0078] In the formula, η ey k is the energy consumption balance coefficient; k7 is the energy consumption constraint coefficient, preferably 0.001W. -1 ΔE / ΔF represents the energy consumption growth rate corresponding to wind speed, which is fitted using FCU measured data; ΔE represents the change in energy consumption; ΔF represents the change in wind speed.

[0079] By multiplying the wind speed gradient coefficient, the air disturbance correction coefficient, and the energy consumption balance coefficient, the wind turbine speed efficiency coefficient can be obtained. The calculation process innovatively constructs a three-dimensional calculation model of continuous wind speed gradient + air disturbance correction + energy consumption balance constraint. Through a multi-parameter dynamic coupling correction mechanism, it solves the problem of the discrete mapping of "wind speed level → fixed coefficient" used in traditional methods.

[0080] Subsequently, the maximum sensible heat cooling capacity data of the fan coil unit is extracted from the building structure data (this value can be determined by the building load calculation, reflecting the maximum sensible heat cooling capacity that the air conditioner can output in the space), and multiplied by the fan speed efficiency coefficient and the water supply temperature efficiency coefficient to obtain the real-time cooling power data used to characterize the effective cooling potential of the fan coil unit.

[0081] This invention quantifies the impact of fan speed and water supply temperature on the cooling capacity of air conditioners through a multi-parameter dynamic coupling correction mechanism, making the obtained real-time cooling power more consistent with the actual capacity of the equipment. This provides a reliable data foundation for subsequent temperature prediction and control based on cooling potential, thereby avoiding energy waste caused by selecting excessively high fan speeds due to underestimating the cooling potential.

[0082] S2. Input the real-time status data and the real-time cooling power data into the regression model based on gradient boosting tree to perform multi-scenario prediction and obtain the temperature change corresponding to each wind speed level.

[0083] In one embodiment, step S2 includes:

[0084] State features and power features are extracted from the real-time status data and the real-time cooling power data, respectively, to construct a feature input set;

[0085] Choose any wind speed setting as the target wind speed setting, construct a target virtual scenario corresponding to the target wind speed setting based on digital twin technology, and input the feature input set into the gradient boosting tree-based regression model in the target virtual scenario to make a prediction and obtain the target temperature change corresponding to the target wind speed setting.

[0086] The target wind speed setting is updated to iteratively execute the construction process of the target virtual scenario and the temperature prediction process in the target virtual scenario based on the updated target wind speed setting, until all wind speed settings are iterated to obtain the temperature change corresponding to each wind speed setting.

[0087] Specifically, this invention extracts time, room area, indoor temperature, set temperature, current wind speed, switch status, water supply / return temperature, indoor humidity, valve status, building characteristics, outdoor temperature, outdoor humidity, light intensity, outdoor wind speed, and building insulation coefficient from real-time status data as status features, and combines these with real-time cooling power data as power features. These two features are then concatenated to obtain a feature input set.

[0088] Select any fan speed setting from the air conditioner as the target fan speed. Based on digital twin technology, and grounded in physical laws and real-time state data, including building thermodynamic models (such as heat balance equations), air conditioning system models (such as fan coil unit heat exchange efficiency), and environmental parameters (such as indoor and outdoor temperature and humidity), construct a digital twin to replicate the physical space. Specifically, the physical entity's state (such as fan speed and water supply temperature) is synchronized to the virtual model in real time using sensor data. Historical data is used to calibrate model parameters to ensure consistency between the virtual scenario and the real environment. The specific implementation of digital twin technology can be... Referring to existing standard methods (such as modeling based on FMI or AMEsim); the air conditioning fan is set to the target fan speed in the space, and the relevant parameters corresponding to the target fan speed are synchronized to form a target virtual scenario; the initial ambient temperature and humidity of the target virtual scenario are set according to historical data to simulate the heat dissipation load of the real indoor environment, so as to realize the real-time loading of the environment of the target virtual scenario; the equipment status is synchronized in the scenario, that is, the parameters such as the heat exchanger efficiency and fan energy consumption of the air conditioner in the target virtual scenario are consistent with the current status of the real equipment (synchronized to the digital twin in real time through sensor data).

[0089] For the target virtual scenario, the spatial temperature uniformity coefficient (range 0-1, the more uniform the room, the closer to 1) is calculated based on the temperature standard deviation of each monitoring point in the room based on the digital twin simulation. The dynamic heat load coefficient (unit W, e.g., 1.2 when the room is full and 0.8 when the room is empty) is obtained based on the equipment running time determined by the digital twin (1.0 for new equipment, and the ratio of its actual running time to its rated running time for aged equipment). These are used as digital twin scenario features and feature input sets and input into a gradient boosting tree-based regression model, namely the LightGBM model, to simulate the operation of various wind speed levels in the target virtual scenario. This allows for the prediction of the temperature at a preset time in the future, outputting the target temperature change at the target wind speed level at a preset future time (e.g., 1 hour later). Alternatively, the predicted target temperature value at a preset future time (e.g., 1 hour later) can be obtained by adding the predicted temperature change to the current temperature.

[0090] The LightGBM model has been pre-trained using historical data, including: a "feature input set" at different wind speeds over the past three months, digital twin scenario features corresponding to the virtual scenario, and "actual temperature changes." The model learns the mapping relationship between features and temperature changes to achieve predictive capabilities. The historical real-time cooling power in this input set can be obtained using methods for calculating real-time cooling power. The LightGBM model learns the mapping relationship between input features and temperature changes through training on historical data. During training, feature importance assessment (such as gain analysis) is used to select key features and avoid multicollinearity. Real-time cooling power data, as a power feature, along with state features (such as indoor and outdoor temperatures) and corresponding digital twin scenario features, reflects system dynamics, ensuring logical consistency between input and output. The initial parameters during training are optimally set as follows: learning rate 0.01, tree depth 8, number of leaf nodes 200, and regularization coefficient 0.1 (to avoid overfitting). The training objective uses the mean squared error (MSE) between the predicted temperature change and the label as the loss function. A decision tree ensemble model is generated iteratively through gradient boosting—each iteration focuses on the sample with the largest prediction error in the previous round, generating a new decision tree to correct the error. Bayesian optimization is used to search for the optimal parameter combination (such as tree depth and number of leaf nodes) with the goal of minimizing the mean absolute error (MAE) of the validation set. A physical penalty term is added to the loss function—if the predicted temperature change violates the heat exchange law (e.g., the temperature rises when the sensible cooling is positive), the penalty value is increased to force the model output to conform to physical logic. Furthermore, the specific training and application process of the LightGBM model can refer to existing technologies, and adaptive adjustments can be made based on the input / output and loss function of this invention. Meanwhile, other models, such as XGBoost, can also be used for temperature prediction.

[0091] The target fan speed in the virtual scenario is then updated, while other parameters (such as ambient temperature and humidity, water supply temperature, heat dissipation load, etc.) remain consistent with the initial target fan speed. The process of constructing the target virtual scenario and predicting the temperature in the target virtual scenario is then iterated based on the updated target fan speed until all fan speeds are iterated and the temperature change corresponding to each fan speed is obtained.

[0092] This invention constructs a "virtual mirror" of an HVAC system using digital twin technology, enabling real-time interaction between physical entities (fan coil units, room space, indoor and outdoor environment) and the virtual model. It integrates physical laws with real-time data, providing a basic framework for virtual-real mapping for multi-scenario prediction and avoiding deviations caused by ignoring actual scene details in pure model prediction. By combining digital twin virtual scenarios with LightGBM model prediction, it solves the problems of incomplete scenario coverage and predictions deviating from reality in traditional air conditioning control, making the prediction results more consistent with the actual operating state of the air conditioner. Compared with traditional linear models, the prediction error is significantly reduced.

[0093] S3. Calculate the real-time prediction error of each wind speed level, and perform a strong update on the first residual corresponding to the active wind speed level at a first update rate to obtain a strong update result. Perform a weak penetration update on the second residual corresponding to the inactive wind speed level at a second update rate to obtain a weak penetration update result. Combine the strong update result with the result to correct the temperature change of each wind speed level to obtain the final predicted temperature corresponding to each wind speed level. The first update rate is greater than the second update rate. Specifically, this invention adds the temperature change predicted for each wind speed level in the previous cycle to the current temperature of the previous cycle, and then uses the difference between the sum and the current temperature as the real-time prediction error of that wind speed level (i.e., the difference between the predicted temperature value and the actual temperature value at the current time), thereby obtaining the real-time prediction error of all wind speed levels.

[0094] In one embodiment, the step of performing a strong update on the first residual corresponding to the active wind speed level at a first update rate to obtain a strong update result includes:

[0095] A feature mapping function is constructed based on the real-time meteorological data and the historical meteorological data of the HVAC system, and a first update rate is determined according to the feature mapping function;

[0096] The currently used wind speed setting is taken as the active wind speed setting, and based on the first update rate, the first residual corresponding to the active wind speed setting and the real-time prediction error are updated using the exponential mean index method to obtain a strong update result.

[0097] Specifically, this invention constructs a feature mapping function based on the degree of difference (intensity of environmental change) between real-time meteorological data and historical meteorological data. The feature mapping function value is then multiplied by an initial rate (preferably 0.3) to obtain the first update rate. Subsequently, the currently used wind speed setting is taken as the active wind speed setting, and the first residual corresponding to the active wind speed setting and the real-time prediction error corresponding to the active wind speed setting are updated using the Exponential Moving Average (EMA) method according to the first update rate, resulting in a strong update result. This process is expressed by the following formula:

[0098] R F,t =α t ×e t +(1-α t )×R F,t-1

[0099] In the formula, R F,t For strong update results; α t The first update rate is 0.3-0.6; e t For real-time prediction error; R F,t-1 The first residual is the residual left over from the previous control cycle at the active wind speed setting.

[0100] This invention addresses the cumulative correction error at active fan speed settings by dynamically iterating through a strong update mechanism that weights and fuses new errors with historical residuals. This reflects the long-term prediction deviation trend at that setting and ensures that the residuals quickly adapt to the actual deviation by correcting the current control in real time. The strong update mechanism (with high weighting for real-time errors) ensures that the prediction deviation at that setting can be promptly offset, directly improving the accuracy of the current air conditioning control. The exponential mean index method retains the trend of historical residuals, avoiding interference from sudden abnormal errors, while giving higher weight to real-time prediction errors, enabling rapid response to current deviations and balancing stability and timeliness.

[0101] In one embodiment, constructing a feature mapping function based on the real-time meteorological data and the historical meteorological data of the HVAC system includes:

[0102] Real-time environmental characteristics are constructed based on the real-time meteorological data, historical environmental characteristics are constructed based on the historical meteorological data, and the real-time fluctuation index of the HVAC system is determined based on the historical environmental characteristics and the real-time environmental characteristics.

[0103] The historical meteorological data are clustered according to time segments and weather types to obtain multiple environmental scenarios. The characteristic mean and standard deviation of each environmental scenario are calculated as the historical normal baseline of the environmental scenario, and the normal deviation index of the real-time environmental characteristics is determined based on the historical normal baseline.

[0104] Calculate the environmental mutation amount of the real-time environmental feature relative to the historical environmental feature, and determine the feature mutation index based on the environmental mutation amount;

[0105] The real-time fluctuation index, the normal deviation index, and the feature mutation index are fused to obtain the feature complexity, and the feature complexity is nonlinearly mapped to obtain the feature mapping function.

[0106] Specifically, this invention selects several core features (such as outdoor temperature, light intensity, indoor-outdoor temperature difference, water supply temperature, and indoor humidity change rate) that have the greatest impact on environmental complexity from real-time meteorological data to form real-time environmental features that include several environmental characteristics. Based on this principle, historical environmental features are constructed using historical meteorological data from the past three months. For each type of feature, the deviation rate between its current value and the average of the most recent three periods is calculated to obtain a real-time fluctuation index. This process is expressed by the following formula:

[0107]

[0108] In the formula, F f,i For real-time volatility index; x i (t) represents the value of the i-th environmental feature in the current period; mean() is the average value of environmental feature i over the last 3 periods (smoothing short-term noise), calculated using historical meteorological data; x i,max,range Let be the historical maximum fluctuation data of feature i in the characteristic environment, and be the difference between the historical maximum and minimum values.

[0109] For each feature, based on historical meteorological data, clustering is performed by time segment + weather type (e.g., 14:00-15:00 on summer weekdays + sunny days constitute one scenario). The mean and standard deviation of each scenario are calculated as the historical baseline for that scenario. Then, for the current scenario (matched by time and weather type), the Z-score standardized deviation of each feature is calculated to obtain the normal deviation index, which quantifies the deviation of the current feature from historical data of the same scenario. This process is expressed by the following formula:

[0110] F d,i =min(|x i (t)-μ i | / (2σ i ),1.0)

[0111] In the formula, F d,i The normal deviation index; μ i σ i Let be the mean and standard deviation of the i-th environmental feature.

[0112] This invention classifies a change in a feature that exceeds 70% of its historical maximum instantaneous change within the current period as a mutation, and then obtains a feature mutation index to capture short-term drastic changes. This process is expressed by the following formula:

[0113] F s,i =0.5×[1+tanh(k×(|Δx i -0.7×Δx i,max ))]

[0114] In the formula, F s,i Δx is the characteristic mutation index; tanh is the hyperbolic tangent function used to map the result to [0, 0.5]; k is the sensitivity coefficient, preferably 10; Δx i Let Δx be the change of the i-th environmental feature in the current period compared to the previous period; i,max The maximum instantaneous change in the history of the i-th environmental characteristic can be determined based on historical data or expert experience, such as Δx for outdoor temperature. i,max The preferred temperature is 5℃ / 5min. It should be noted that the hyperbolic tangent function and its sensitivity coefficient used in the above formula are an exemplary form for realizing feature mutation quantification; the specific form and sensitivity parameters of the above function can be determined in practice based on data-driven methods or expert rules, and are not limited to a single function form or specific value.

[0115] The three dimensions mentioned above are fused through dynamic weights and non-linear coupling to obtain the final feature mapping function, ensuring that the contribution of each dimension changes dynamically with the scene:

[0116] In this invention, the weight of each environmental feature is not fixed, but is positively correlated with its contribution to historical prediction errors. The calculation process is as follows:

[0117] W i =w i,base ×(1+0.5×F d,i )

[0118] In the formula, W i w represents the weight of the i-th environmental feature. i,base The basic weight of the i-th environmental feature can be determined by calculating the feature importance using a random forest model, or it can be determined directly based on experience. For example, the basic weights for outdoor temperature, light intensity, indoor-outdoor temperature difference, water supply temperature, and indoor humidity change rate are 0.25, 0.3, 0.15, 0.2, and 0.1, respectively.

[0119] For each type of feature, by fusing its fluctuation, deviation, and mutation information, the single-feature complexity index can be obtained, which is expressed by the following formula:

[0120] C i =W i×[0.4×F f,i +0.4×F d,i +0.2×F s,i ]

[0121] In the formula, C i Let be the single feature complexity index corresponding to the i-th environmental feature.

[0122] The global complexity index is obtained by summing the complexity indices of all environmental features, and then mapped to [1.0, 2.0] using this global complexity index to obtain the feature mapping function. This process employs an exponential growth function to enhance the discriminative power of high-complexity scenes, which is expressed by the following formula:

[0123]

[0124] In the formula, λ is the feature mapping function; λ is an adjustable amplification coefficient, the specific value of which and the mapping strategy can be determined in practice based on application requirements or data statistical characteristics, preferably 3.0; C total This is the global complexity index.

[0125] This invention quantifies high-dimensional environmental and equipment features into a feature mapping function that characterizes the environmental complexity index. This allows the residual correction of active wind speed levels to closely follow the dynamics of the environment and equipment, avoiding the correction lag or excessive fluctuations caused by traditional fixed-rate updates. This achieves accurate adaptation of the first update rate to the dynamics of the scene. At the same time, this function breaks through the logic of traditional single-feature linear mapping or fixed-weight fusion, and innovatively constructs a three-dimensional nonlinear fusion model of multi-feature fluctuation-historical deviation-mutation detection, achieving comprehensive quantification of environmental complexity and accurate matching of update rates.

[0126] In one embodiment, the step of performing a weak-penetration update on the second residual corresponding to the inactive wind speed level at a second update rate, obtaining a weak-penetration update result, and then combining the strong update result to correct each of the temperature changes to obtain the final predicted temperature corresponding to each of the wind speed levels, includes:

[0127] Other wind speed levels besides the active wind speed level are taken as the inactive wind speed level, and an error penetration mechanism is introduced to update the second residual and the real-time prediction error corresponding to the inactive wind speed level based on the second update rate and using the exponential mean index method, so as to obtain a weak penetration update result.

[0128] Based on the correspondence between the active wind speed level, the inactive wind speed level, and each wind speed level, the strong update result and the weak penetration update result are respectively incorporated into the corresponding temperature change amount to obtain the final predicted temperature corresponding to each wind speed level.

[0129] Specifically, this invention treats wind speed levels other than active wind speed levels as inactive wind speed levels and introduces an error penetration mechanism. To avoid over-updating of inactive levels, the following rules are set: the residual change in a single update does not exceed 0.05°C (to prevent sudden error interference); the update direction of inactive levels must be consistent with the error trend of active levels (e.g., if the error of the active level is positive, and the error of the inactive level is negative, the update amplitude must be reduced); and under these rules, the second residual and real-time prediction error corresponding to the inactive wind speed levels are updated using the exponential mean index method based on the second update rate, resulting in a weakly penetrated update result. This process is expressed by the following formula:

[0130] R nF,t =α nt ×e t +(1-α nt )×R nF,t-1

[0131] In the formula, R nF,t This is a weak penetration update result; α nt The second update rate is adjustable, and its specific value can be determined during implementation based on application requirements or data statistical characteristics, preferably 0.01; R nF,t-1 The second residual is the residual left over from the previous control cycle for the inactive wind speed setting.

[0132] Calculate the absolute value of the difference between the weak update result and the second residual of the previous period. If the absolute value of the difference is ≤0.05℃, it means that the initial result meets the single update amplitude limit, and the initial result is retained.

[0133] If the absolute value of this difference is greater than 0.05℃, then based on the second residual of the previous period, and according to the positive or negative direction of the real-time prediction error, 0.05℃ is taken as the initial result of the maximum single change—when e t When positive, the adjusted result of the weak penetration update is R. nF,t The sum with 0.05℃; when e t When negative, the adjusted result R of the weak penetration update result is... nF,t The difference of 0.05℃ ensures that a single update will not cause abrupt changes in the residual due to error fluctuations. Then, the result adjusted for single-change limits is compared with the real-time prediction error e of the active wind speed level. t Whether the signs are consistent (i.e., both are positive or both are negative): If the signs are consistent, it means that the update direction of the inactive gear residual matches the current environmental error trend, and the adjustment result is directly used as the final weak penetration update result; if the signs are inconsistent, it means that the update direction of the inactive gear residual is opposite to the current error trend, and the reverse update amplitude needs to be weakened—halve the difference between the adjustment result and the second residual of the previous period, and then add it to the second residual of the previous period to obtain the final weak penetration update result, i.e., R. nF,t=R nF,t-1 +0.5×(adjustment result - R) nF,t-1 This is to avoid the residual of inactive gears deviating from the actual environmental change pattern.

[0134] Although inactive gears are not used, environmental changes (such as outdoor temperature and light) have the same impact on their predictions as active gears. Therefore, this invention uses a smaller second update rate to permeate some information of the real-time prediction error into the slowly accumulating residuals of inactive gears, thus avoiding residual errors.

[0135] Based on the correspondence between active and inactive wind speed levels and each wind speed level, if the wind speed levels are divided into high, medium, and low, and the current wind speed level is high, then the strong update result is added to the current real-time temperature and the temperature prediction value for the high wind speed level predicted by the LightGBM model to obtain the final predicted temperature for the high wind speed level at a preset time in the future. For the medium wind speed level, the weak penetration update result for the medium wind speed level is added to the current real-time temperature and the temperature prediction value for the medium wind speed level predicted by the LightGBM model to obtain the final predicted temperature for the medium wind speed level at a preset time in the future. The calculation process for the final predicted temperature of the low wind speed level is the same as that for the medium wind speed level. Based on this principle, the final predicted temperature for all wind speed levels at a preset time in the future can be obtained.

[0136] This invention employs a low-rate and error permeation mechanism to allow the prediction error of inactive speeds to slowly adapt to environmental changes without interfering with the stability of the current active speed control. This prevents the residuals of inactive speeds from fluctuating due to frequent updates, thus ensuring the stability of the current active speed control. By slowly absorbing real-time prediction errors, the second residual of inactive speeds can gradually adapt to long-term environmental changes. Strong updates focus on real-time accuracy for active speeds, while weak updates ensure long-term adaptation for inactive speeds. The combination of these two approaches ensures that the final predicted temperature for all wind speed speeds reflects the actual operating patterns, providing a reliable basis for selecting the optimal speed for subsequent targets. Furthermore, the synergy between strong and weak updates achieves a dynamic balance in prediction accuracy across all speed speeds.

[0137] S4. Construct and solve an objective function based on the final predicted temperatures to obtain the target control command to control the HVAC system.

[0138] In one embodiment, the step of constructing and solving the objective function based on each of the final predicted temperatures to obtain the target control command includes:

[0139] The objective function is constructed by minimizing the sum of the absolute differences between the final predicted temperature and the target temperature, and the energy required to reach the target temperature. The optimal wind speed setting is obtained by solving the objective function.

[0140] The optimal wind speed setting is compared with the current wind speed setting, and a target control command is generated based on the comparison result.

[0141] Specifically, this invention uses the absolute difference between the final predicted temperature and the target temperature to characterize comfort, and constructs an objective function with the goal of maximizing comfort and minimizing energy consumption, which is expressed by the following formula:

[0142] J = arg min(C × |T) set -T final,Fi |+λE(Fi)×P)

[0143] In the formula, J is the objective function value; C is the temperature deviation comfort cost coefficient, preferably 1 yuan / ℃; T set The target temperature; T final,Fi Fi represents the final predicted temperature for the wind speed setting; λ is the comfort / energy saving trade-off coefficient (which can be configured as needed, such as λ=0.8 during peak summer electricity consumption, emphasizing energy saving; and λ=0.2 during nighttime rest, emphasizing comfort); E(Fi) is the estimated energy consumption value corresponding to the wind speed setting Fi (obtained by multiplying the wind speed-energy consumption curve provided by the FCU manufacturer by the operating time); P is the electricity price, preferably 0.5 yuan / kWh.

[0144] The wind speed setting with the minimum objective function value is taken as the optimal wind speed setting. The optimal wind speed setting is compared with the current wind speed setting. If they are the same, the setting is not switched. Otherwise, the current wind speed setting is switched to the optimal wind speed setting, and the target control command is generated to control the HVAC system.

[0145] In addition to the above embodiment, which calculates the objective function value using a formula, it can also be achieved through a corresponding AI (Artificial Intelligence) algorithm model: A sample dataset is constructed containing input content: target temperature, the final predicted temperature at wind speed setting Fi, the comfort / energy saving trade-off coefficient, the energy consumption estimate corresponding to wind speed setting Fi, and the output content, the objective function value. The input content is then labeled with the output content. Based on a learning algorithm, the AI ​​algorithm model is trained using the aforementioned sample dataset. During training, existing methods can be consulted to retrain or fine-tune the convolutional neural network model to improve its generalization ability. Finally, the trained convolutional neural network model is obtained. In practical applications, the current input content is input into the convolutional neural network model, which analyzes and processes it to obtain the output content. This is the processing procedure of the above algorithm model. It should be noted that the above training method is merely an example. Those skilled in the art can choose other suitable methods, such as reinforcement learning, federated learning, or transfer learning, depending on the scenario. These are not specifically limited in this embodiment of the invention. In addition, other data calculated using formulas or models can also be implemented using this AI algorithm model, without making specific limitations here.

[0146] This invention breaks through the traditional single control logic of only focusing on temperature or only focusing on energy saving by constructing a target function with dual objectives of comfort and energy economy corresponding to temperature. It ensures that the final temperature is close to the target value (satisfying the comfort of personnel) and controls the energy consumption to reach the target temperature (reducing operating costs). It avoids energy waste caused by selecting a high setting for cooling or temperature failure caused by selecting a low setting for energy saving.

[0147] This invention addresses the issue of improving the control accuracy of existing HVAC control methods by designing an intelligent control method. This method acquires real-time status data of the HVAC system and quantifies its effective cooling potential, directly solving the blindness of traditional experience-based adjustments and laying the foundation for precise control. Based on a gradient boosting tree regression model, multi-scenario prediction can anticipate temperature changes at different fan speeds, shifting control from "passive response" to "active prediction," thus enhancing the forward-looking nature of temperature regulation. High-rate strong updates are used for active fan speeds to quickly correct prediction errors and ensure the accuracy of current temperature control, while low-rate weak penetration updates are used for inactive speeds to prevent residuals from becoming invalid due to prolonged lack of updates, thereby improving the overall stability of control. Through real-time data-driven and dynamic model correction, the intelligent level of air conditioning operation is significantly improved, achieving precise, energy-saving, and stable control of the HVAC system.

[0148] It should be noted that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders.

[0149] In another embodiment, such as Figure 2 As shown, a second aspect of the present invention provides an intelligent control system, comprising:

[0150] The data acquisition module 10 is used to acquire real-time status data of the HVAC system and quantify the effective cooling potential of the HVAC system based on the real-time status data to obtain real-time cooling power data.

[0151] The temperature prediction module 20 is used to input the real-time status data and the real-time cooling power data into a regression model based on gradient boosting tree to perform multi-scenario prediction and obtain the temperature change corresponding to each wind speed level.

[0152] The temperature correction module 30 is used to calculate the real-time prediction error of each wind speed level, and to perform a strong update on the first residual corresponding to the active wind speed level at a first update rate to obtain a strong update result, and to perform a weak penetration update on the second residual corresponding to the inactive wind speed level at a second update rate to obtain a weak penetration update result. The weak update result is then combined with the strong update result to correct the temperature change of each wind speed level, thereby obtaining the final predicted temperature corresponding to each wind speed level; the first update rate is greater than the second update rate.

[0153] The instruction generation module 40 is used to construct and solve an objective function based on the final predicted temperature to obtain a target control instruction to control the HVAC system.

[0154] It should be noted that each module in the aforementioned intelligent control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module. For specific limitations regarding an intelligent control system, please refer to the limitations regarding an intelligent control method above; both have the same function and role, and will not be repeated here.

[0155] A third aspect of the present invention provides an electronic device comprising:

[0156] Processor, memory, and bus;

[0157] The bus is used to connect the processor and the memory;

[0158] The memory is used to store operation instructions;

[0159] The processor is configured to execute instructions by calling the operation instructions, causing the processor to perform operations corresponding to an intelligent control method as shown in the first aspect of the present invention.

[0160] In one alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The illustrated electronic device 5000 includes a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, for example, via a bus 5002. Optionally, the electronic device 5000 may also include a transceiver 5004. It should be noted that in practical applications, the transceiver 5004 is not limited to one type, and the structure of this electronic device 5000 does not constitute a limitation on the embodiments of the present invention.

[0161] Processor 5001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 5001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0162] Bus 5002 may include a path for transmitting information between the aforementioned components. Bus 5002 may be a PCI bus or an EISA bus, etc. Bus 5002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0163] The memory 5003 may be a ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or it may be an EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0164] The memory 5003 is used to store application code that executes the present invention, and its execution is controlled by the processor 5001. The processor 5001 is used to execute the application code stored in the memory 5003 to implement the content shown in any of the foregoing method embodiments.

[0165] Among them, electronic devices include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers.

[0166] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements an intelligent control method as described in the first aspect of the present invention.

[0167] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program that, when run on a computer, enables the computer to execute the corresponding content in the foregoing method embodiments.

[0168] Furthermore, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0169] In summary, this invention relates to the field of intelligent control technology, and discloses an intelligent control method, system, device, and medium. The method inputs real-time status data of a heating, ventilation, and air conditioning (HVAC) system and its corresponding real-time cooling power data into a gradient boosting tree-based regression model for multi-scenario prediction. This yields the temperature change corresponding to each fan speed setting and quantifies the real-time prediction error. A strong update is performed on the first residual of the active fan speed setting at a first update rate, and a weak penetration update is performed on the second residual of the inactive fan speed setting at a second update rate less than the first update rate. The predicted temperature change is corrected using the strong and weak update results, resulting in the final predicted temperature for each fan speed setting. This constructs and solves the objective function, generating a target control command and controlling the HVAC system to execute it. This achieves precise, energy-saving, and stable control of the HVAC system.

[0170] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0171] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. An intelligent control method, characterized in that, include: Acquire real-time status data of HVAC system, and quantify the effective cooling potential of HVAC system based on the real-time status data to obtain real-time cooling power data; The real-time status data and the real-time cooling power data are input into a gradient boosting tree-based regression model for multi-scenario prediction to obtain the temperature change corresponding to each wind speed level. The real-time prediction error for each wind speed setting is calculated, and a strong update is performed on the first residual corresponding to the active wind speed setting at a first update rate to obtain a strong update result. A weak penetration update is performed on the second residual corresponding to the inactive wind speed setting at a second update rate to obtain a weak penetration update result. This weak penetration update result is then combined with the strong update result to correct the temperature change for each wind speed setting, resulting in the final predicted temperature for each wind speed setting. The first update rate is greater than the second update rate. The active wind speed setting is the currently used wind speed setting. The inactive wind speed setting is any wind speed setting other than the active wind speed setting. The first residual is the prediction residual left over from the previous control cycle for the active wind speed setting. The second residual is the prediction residual left over from the previous control cycle for the inactive wind speed setting. Based on the final predicted temperatures, an objective function is constructed and solved to obtain the target control command to control the HVAC system to execute. The step of performing a weak-penetration update on the second residual corresponding to the inactive wind speed level at the second update rate, obtaining a weak-penetration update result, and then combining the strong update result to correct each of the temperature changes, to obtain the final predicted temperature corresponding to each of the wind speed levels, includes: An error penetration mechanism is introduced to update the second residual corresponding to the inactive wind speed level and the real-time prediction error based on the second update rate and using the exponential average index method, so as to obtain a weak penetration update result. Based on the correspondence between the active wind speed level, the inactive wind speed level and each wind speed level, the strong update result and the weak penetration update result are respectively incorporated into the corresponding temperature change amount to obtain the final predicted temperature corresponding to each wind speed level. The step of constructing and solving the objective function based on each of the final predicted temperatures to obtain the target control command includes: The objective function is constructed by minimizing the sum of the absolute differences between the final predicted temperature and the target temperature, and the energy required to reach the target temperature. The optimal wind speed setting is obtained by solving the objective function. The optimal wind speed setting is compared with the current wind speed setting, and a target control command is generated based on the comparison result.

2. The intelligent control method according to claim 1, characterized in that, The real-time status data includes equipment operation data and building structure data; wherein... The process of quantifying the effective cooling potential of the HVAC system based on the real-time status data to obtain real-time cooling power data includes: Based on the equipment operation data, the fan speed efficiency coefficient and water supply temperature efficiency coefficient of the HVAC system are determined through a multi-parameter dynamic coupling correction mechanism, and the maximum sensible heat cooling capacity data of the HVAC system is extracted from the building structure data. Based on the fan speed efficiency coefficient, the water supply temperature efficiency coefficient, and the maximum sensible heat cooling capacity data, the effective cooling potential of the HVAC system is quantified to obtain real-time cooling power data, including: multiplying the maximum sensible heat cooling capacity data by the fan speed efficiency coefficient and the water supply temperature efficiency coefficient to obtain real-time cooling power data used to characterize the effective cooling potential of the fan coil unit. The water supply temperature efficiency coefficient is used to reflect the impact of water supply temperature on the air conditioning cooling capacity. It is obtained by multiplying the basic efficiency coefficient, the heat exchange correction coefficient, and the environmental load correction coefficient, where: or b =e [-k1×(Ts-Tsr)²] or d =1-k2×e (-k3×ΔTw) or e =1 / [1+k4×max(ΔTe,0)] In the formula, η b k is the basic efficiency coefficient; k1 is the attenuation coefficient, specifically 0.

05. ;T s The real-time water supply temperature of the fan coil unit; T sr η is the rated water supply temperature of the fan coil unit. d k1 is the heat exchange correction coefficient; k2 is the correction weight, specifically 0.2; k3 is the temperature difference sensitivity coefficient, specifically 0.

3. ΔTw is the difference between the supply water temperature and the return water temperature; η e is the environmental load correction factor; k4 is the environmental sensitivity factor, specifically 0.

02. ΔTe is the difference between the outdoor temperature and the indoor set temperature. The fan speed efficiency coefficient is used to reflect the impact of fan speed on the air conditioning cooling capacity. It is obtained by multiplying the fan speed gradient coefficient, the air disturbance correction coefficient, and the energy consumption balance coefficient, where: η w =[1-e (-k5×Fcur / Fmax) ] / [1-e (-k5) ] or db =1-k6×(ΔH in / Δt)² or ey =1 / [1+k7×(ΔE / ΔF)×F cur ] In the formula, η w is the wind speed gradient coefficient; k5 is the marginal benefit coefficient, specifically 3.0; F cur For wind speed; F max The rated maximum wind speed; η db ΔH is the air disturbance correction factor; k6 is the humidity disturbance factor, specifically 0.01 min² / %²; in η represents the change in indoor humidity between the current and previous cycles; Δt represents the time interval; η represents the change in indoor humidity between the current and previous cycles. ey k is the energy consumption balance coefficient; k7 is the energy consumption constraint coefficient, specifically 0.001W. -1 ΔE / ΔF represents the energy consumption growth rate corresponding to wind speed; ΔE represents the change in energy consumption; ΔF represents the change in wind speed.

3. The intelligent control method according to claim 1, characterized in that, The step of inputting the real-time status data and the real-time cooling power data into a gradient boosting tree-based regression model for multi-scenario prediction to obtain the temperature change corresponding to each wind speed setting includes: State features and power features are extracted from the real-time status data and the real-time cooling power data, respectively, to construct a feature input set; Choose any wind speed setting as the target wind speed setting, construct a target virtual scenario corresponding to the target wind speed setting based on digital twin technology, and input the feature input set into the gradient boosting tree-based regression model in the target virtual scenario to make a prediction and obtain the target temperature change corresponding to the target wind speed setting. The target wind speed setting is updated to iteratively execute the construction process of the target virtual scenario and the temperature prediction process in the target virtual scenario based on the updated target wind speed setting, until all wind speed settings are iterated to obtain the temperature change corresponding to each wind speed setting.

4. The intelligent control method according to claim 2, characterized in that, The real-time status data also includes real-time meteorological data; wherein... The step of performing a strong update on the first residual corresponding to the active wind speed level at a first update rate to obtain a strong update result includes: A feature mapping function is constructed based on the real-time meteorological data and the historical meteorological data of the HVAC system, and a first update rate is determined according to the feature mapping function; Based on the first update rate, the first residual corresponding to the active wind speed level and the real-time prediction error are updated using the exponential average index method to obtain a strong update result.

5. The intelligent control method according to claim 4, characterized in that, The construction of the feature mapping function based on the real-time meteorological data and the historical meteorological data of the HVAC system includes: Real-time environmental characteristics are constructed based on the real-time meteorological data, historical environmental characteristics are constructed based on the historical meteorological data, and the real-time fluctuation index of the HVAC system is determined based on the historical environmental characteristics and the real-time environmental characteristics. The historical meteorological data are clustered according to time segments and weather types to obtain multiple environmental scenarios. The characteristic mean and standard deviation of each environmental scenario are calculated as the historical normal baseline of the environmental scenario, and the normal deviation index of the real-time environmental characteristics is determined based on the historical normal baseline. Calculate the environmental mutation amount of the real-time environmental feature relative to the historical environmental feature, and determine the feature mutation index based on the environmental mutation amount; The real-time fluctuation index, the normal deviation index, and the feature mutation index are fused to obtain the feature complexity, and the feature complexity is nonlinearly mapped to obtain the feature mapping function.

6. An intelligent control system, characterized in that, include: The data acquisition module is used to acquire real-time status data of the HVAC system and quantify the effective cooling potential of the HVAC system based on the real-time status data to obtain real-time cooling power data. The temperature prediction module is used to input the real-time status data and the real-time cooling power data into a regression model based on gradient boosting tree to perform multi-scenario prediction and obtain the temperature change corresponding to each wind speed level. A temperature correction module is used to calculate the real-time prediction error of each wind speed setting, and to perform a strong update on the first residual corresponding to the active wind speed setting at a first update rate to obtain a strong update result. Then, it performs a weak penetration update on the second residual corresponding to the inactive wind speed setting at a second update rate to obtain a weak penetration update result. This weak penetration update result is then combined with the strong update result to correct the temperature change for each wind speed setting, resulting in the final predicted temperature for each wind speed setting. The first update rate is greater than the second update rate. The active wind speed setting is the currently used wind speed setting. The inactive wind speed setting is any wind speed setting other than the active wind speed setting. The first residual is the prediction residual of the active wind speed setting from the previous control cycle. The second residual is the prediction residual of the inactive wind speed setting from the previous control cycle. The instruction generation module is used to construct and solve an objective function based on each of the final predicted temperatures to obtain target control instructions to control the HVAC system to execute. The step of performing a weak-penetration update on the second residual corresponding to the inactive wind speed level at the second update rate, obtaining a weak-penetration update result, and then combining the strong update result to correct each of the temperature changes, to obtain the final predicted temperature corresponding to each of the wind speed levels, includes: An error penetration mechanism is introduced to update the second residual corresponding to the inactive wind speed level and the real-time prediction error based on the second update rate and using the exponential average index method, so as to obtain a weak penetration update result. Based on the correspondence between the active wind speed level, the inactive wind speed level and each wind speed level, the strong update result and the weak penetration update result are respectively incorporated into the corresponding temperature change amount to obtain the final predicted temperature corresponding to each wind speed level. The step of constructing and solving the objective function based on each of the final predicted temperatures to obtain the target control command includes: The objective function is constructed by minimizing the sum of the absolute differences between the final predicted temperature and the target temperature, and the energy required to reach the target temperature. The optimal wind speed setting is obtained by solving the objective function. The optimal wind speed setting is compared with the current wind speed setting, and a target control command is generated based on the comparison result.

7. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the intelligent control method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the intelligent control method as described in any one of claims 1 to 5.

Images