An ai large model-based device energy-saving control system and method

Abstract

The application relates to the field of device control, in particular to a device energy-saving control system and method based on an AI large model, which comprises a device sensing module, a load prediction module, a fuzzy control module, a wind-water joint debugging module and an energy efficiency optimization module, the device sensing module is used for determining the heat gain in a region, the load prediction module is used for generating a pre-control decision, the fuzzy control module is used for calculating static pressure and air supply, the wind-water joint debugging module is used for correcting the air supply of each region, and the energy efficiency optimization module is used for planning the refrigerating capacity supply efficiency of an air conditioner at different time periods.The application can realize more stable temperature and humidity control, improve indoor environment comfort and device operation stability, reduce the operation time of a water chiller unit, reduce energy consumption cost, guarantee the smooth operation of air conditioner compressors, water pumps, fans and other devices, enhance the reliability of an energy-saving control system, reduce building operation energy consumption cost, and improve the comprehensive energy efficiency of an air conditioning system.

Description

Technical Field

[0001] This invention relates to the field of equipment control, specifically to an energy-saving control system and method for equipment based on an AI large model. Background Technology

[0002] Energy-saving equipment control is a technical means of monitoring and managing the operating status, operating time, and energy consumption of electrical equipment to reduce unnecessary energy consumption. Since air conditioning equipment typically accounts for 60%-70% of the total energy consumption in a park, energy-saving control of central air conditioning is crucial for limiting energy efficiency. Precise and effective energy-saving control can accurately match the actual temperature control needs of a building, reducing the energy consumption of central air conditioning equipment by 20%-30%.

[0003] Because central air conditioning systems within buildings involve multiple heat exchange stages and have complex water circulation processes, achieving precise load control is challenging. Traditional PID control algorithms are easily affected by nonlinear and time-delayed interference factors such as equipment operating status, personnel movement, and temperature changes. This results in low control accuracy and weak anti-interference capabilities for the cold source system, leading to prolonged high-load operation of air conditioning equipment and impacting energy-saving control performance.

[0004] Furthermore, the heating and cooling loads inside buildings are in a dynamic and complex state, with many factors influencing temperature, complex load control methods, and insufficient coordination between airflow and cooling water flow, resulting in inefficient energy utilization. For scenarios such as server rooms and indoor construction sites, inefficient central air conditioning equipment may lead to temperature runaway, interfering with the normal operation of equipment, and failing to achieve the goal of energy saving. Summary of the Invention

[0005] The purpose of this invention is to provide an energy-saving control system and method for equipment based on an AI large model, so as to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an equipment energy-saving control system based on an AI large model, comprising: an equipment sensing module, a load prediction module, a fuzzy control module, a wind and water joint adjustment module, and an energy efficiency optimization module;

[0007] The device sensing module is used to monitor the real-time power of the device through the power monitoring socket. It constructs a multi-dimensional feature dataset using the ratio of regional lighting power consumption to equipment operation power consumption, regional functions, and power consumption time. The multi-dimensional feature dataset is input into a pre-trained machine learning model. The clustering algorithm is used to identify regional power change patterns, estimate real-time personnel distribution density, and obtain a personnel density distribution heat map. Passive infrared sensors and contact temperature sensors are deployed in the building to detect the surface temperature of the operating equipment. Based on the personnel density distribution heat map and equipment temperature, the thermal gain in the area is determined.

[0008] The load prediction module is used to establish a heat load distribution model based on ambient temperature, set temperature and heat gain, use a genetic algorithm to optimize the backpropagation neural network to learn the heat load change pattern, identify typical heat load cycles, and output the predicted heat load in the current cycle. The heat load prediction model is established with the regional temperature and predicted heat load as the state space and the energy reduction amount as the reward function. The DF-DQN algorithm is used to reduce the actions in the temperature change space and determine the pre-control decision of the air conditioner in the cycle.

[0009] The fuzzy control module is used to install a return pipe at the air conditioning terminal, detect the inlet and outlet temperatures of chilled water, mix the return pipe with the supply water pipe, and adjust the mixing ratio through a three-way valve. Based on historical detection data and heat exchanger parameters, it fits a nonlinear heat transfer model of the surface heat exchanger, establishes the relationship between the heat exchanger's heat exchange performance and the inlet and outlet water temperatures, air volume, and static pressure, calculates the static pressure and air volume based on the heat load prediction model, and ensures that the output heat exchange meets the set temperature range of equipment and personnel in the area. Based on the PMV model, it optimizes control decisions to ensure that the cooling capacity utilization rate of each area is synchronized.

[0010] The air-water joint adjustment module is used to adjust the three-way valve and variable frequency fan according to the planned static pressure setpoint and air supply setpoint, so that the adjustment amount is consistent with the planned amount. The difference between the actual heat exchange at the end of each water tank and the estimated cooling amount is input into the fuzzy PID controller in real time, and the correction amount of the air supply setpoint of each area is output to redistribute the cooling amount among different ends so that the temperature of all areas reaches the set value at the same time.

[0011] The energy efficiency optimization module is used to introduce an energy consumption contribution rate index to evaluate the contribution rate of each central air conditioning air supply area, evaluate the contribution rate of temperature regulation in different areas at different times, use the contribution rate as the power allocation weight, take maximizing the weighted power load efficiency as the objective function, and take meeting the heat load demand as the constraint condition to plan the cooling capacity supply efficiency of the air conditioner at different times, and correct the set temperature range of each area according to the inertia, response speed and time delay of the air conditioning system.

[0012] Furthermore, the device sensing module includes: a personnel distribution unit and a temperature sensing unit;

[0013] The personnel distribution unit is used to determine the sensible heat load of personnel based on power monitoring and infrared scanning, and generate a thermal map of personnel density on the interior plane of the building.

[0014] The temperature sensing unit is used to accumulate the thermal gain of each area based on the sensible heat of personnel and the sensible heat of equipment operation. The thermal gain is the additional heat generated per unit time.

[0015] Furthermore, the load forecasting module includes: a large model unit and a space reduction unit;

[0016] The large model unit is used to establish a heat load distribution model using AI, predict fluctuations in ambient temperature and heat gain, and output the overall heating load.

[0017] The space reduction unit is used to model heat load prediction as a sequence decision, reduce the dimensionality of temperature changes, and generate an air conditioning cooling capacity execution sequence.

[0018] Furthermore, the fuzzy control module includes: a water supply loop unit, a setting optimization unit, and a global planning unit;

[0019] The water supply circuit unit is used to install a return pipe in the air conditioning terminal fan coil unit or air handling unit, and calculate the actual heat exchange by combining the water supply pipe temperature;

[0020] The setting optimization unit is used to fit the performance of the heat exchanger and generate a joint influence function for real-time heat exchange.

[0021] The global planning unit is used to calculate the duct static pressure setpoint based on the predicted cooling load and personnel distribution, using a model predictive control algorithm, and optimizing the air conditioning static pressure supply by taking these as inputs.

[0022] Furthermore, the air-water joint adjustment module includes: an air-cooled distribution unit and a static pressure control unit;

[0023] The air-cooled distribution unit is used to calculate the cooling capacity distribution using an alternating direction multiplier algorithm, so that the cooling capacity of each area is output synchronously.

[0024] The static pressure control unit is used to adjust the opening degree of branch valves and the speed of variable frequency pumps according to the heat load forecast of each area, while using pipeline resistance and equipment capacity limits as adjustment constraints.

[0025] Furthermore, the energy efficiency optimization module includes: a contribution evaluation unit, a spatiotemporal quantization unit, and an intelligent decision-making unit;

[0026] The contribution evaluation unit is used to assess the regional thermal comfort level using the PMV-PPD model, obtain the thermal impact level based on equipment operating parameters, determine the regional weight based on regional functions, and calculate the regional contribution rate by time period.

[0027] The spatiotemporal quantization unit is used to maximize the overall contribution rate and minimize the total energy consumption while meeting the cooling load requirements, and to determine the temperature control priority.

[0028] The intelligent decision-making unit is used to test the inertia, response speed, and time delay of the air conditioning system through step response testing, establish a transfer function model, dynamically adjust the set temperature range of the region, and perform time delay compensation control.

[0029] An energy-saving control method for equipment based on an AI large model includes the following steps:

[0030] Step S1. Construct a multidimensional feature dataset based on the proportion of regional lighting electricity consumption, functional settings, and electricity consumption time. Input the dataset into a machine learning model to estimate the population distribution density and output a population density heat map. Based on the population density distribution heat map and the surface temperature of the operating equipment, determine the thermal gain distribution of the building space.

[0031] Step S2. Establish a heat load distribution model based on ambient temperature, set temperature and heat gain, predict heat gain fluctuations, output the predicted heat load in the current cycle, reduce the dimension of predicted heat load temperature change, and generate an air conditioning cooling capacity execution sequence.

[0032] Step S3. Install a return pipe at the air conditioning terminal, detect the inlet and outlet temperatures of chilled water, establish the relationship between the heat exchanger's heat exchange performance and the inlet and outlet water temperatures, air volume, and static pressure, calculate the static pressure and air volume according to the execution sequence, so that the output cooling capacity controls the area temperature within the set temperature range for equipment and personnel.

[0033] Step S4. Adjust the static pressure and air volume of the air conditioner, determine the actual heat exchange based on the inlet and outlet water temperature and flow rate, input the difference between the actual heat exchange and the estimated cooling capacity into the fuzzy PID controller in real time, correct the air volume set value of each area, and redistribute the cooling capacity.

[0034] Step S5. Based on personnel distribution, equipment operating parameters, and regional functions, calculate the regional contribution rate in different time periods, plan the cooling capacity supply efficiency for different time periods, maximize the average contribution rate of power load while meeting the cooling capacity demand, and adjust the set temperature range of each region according to the inertia, response speed, and time delay of the air conditioning system.

[0035] Furthermore, step S1 includes:

[0036] Step S11. Monitor the real-time power of the equipment through the power monitoring socket, construct a multi-dimensional feature dataset using the ratio of regional lighting power consumption to equipment operation power consumption, regional function settings and power consumption time, and input the multi-dimensional feature dataset into the pre-trained machine learning model;

[0037] In a multidimensional feature dataset, select K time series as initial cluster centers within each feature dimension, assign features to these cluster centers, and minimize the objective function:

[0038] ;

[0039] Where J is the intra-cluster squared error and K is the number of clusters, determined by the data silhouette coefficient. x is the center of the k-th cluster. iIt is the i-th sample point. C is the Euclidean distance from the sample point to the cluster center. k Let J represent the k-th cluster. Determine the cluster where J is the smallest. Output the cluster center of each cluster as a power change pattern. Estimate the real-time population distribution density based on the regional power change pattern to obtain a population density distribution heat map.

[0040] Step S12. Deploy passive infrared sensors and contact temperature sensors inside the building to detect the surface temperature of the operating equipment. Based on the sensible heat of personnel and the sensible heat of equipment operation, accumulate the thermal gain of each area, where the thermal gain is the additional heat generated by the area per unit time.

[0041] Furthermore, step S2 includes:

[0042] Step S21. Based on the ambient temperature, set temperature and heat gain, use AI to establish a heat load distribution model, use a genetic algorithm to optimize the backpropagation neural network to learn the heat load change pattern, identify typical heat load cycles, and output the predicted heat load in the current cycle.

[0043] Step S22. Establish a heat load prediction model with regional temperature and predicted heat load as the state space and energy reduction as the reward function. The model function table is as follows:

[0044] ;

[0045] Where C is the indoor heat capacity, R is the building thermal resistance, and T is the thermal resistance. out (t) and T in (t) represents the outdoor temperature and indoor temperature, respectively, and Q is a function of the indoor temperature. load (t) represents the internal heat load, Q cool (t) represents the air conditioning cooling capacity. Let t be the differential of indoor temperature with respect to time;

[0046] The DF-DQN algorithm is used to reduce the actions within the temperature change space, determine the pre-control decisions for the air conditioner within the period, and generate a cooling capacity execution sequence, which includes cooling capacity and cooling time.

[0047] Furthermore, step S3 includes:

[0048] Step S31. Install a return pipe in the fan coil unit or air handling unit of the air conditioner terminal, detect the inlet and outlet water temperatures of chilled water, mix the return pipe with the supply water pipe, and adjust the mixing ratio through a three-way valve. Based on historical detection data and heat exchanger parameters, fit a nonlinear heat transfer model of the surface heat exchanger, and establish the relationship between the heat exchanger heat transfer performance and the inlet and outlet water temperatures, air volume and static pressure.

[0049] Step S32. Using the predicted cooling load and personnel distribution as input, the model predictive control algorithm is used to plan the duct static pressure setpoint based on the duct resistance characteristics and fan performance curve, so that the area temperature is within the set temperature range of equipment and personnel, and the control decision is optimized based on the PMV model to ensure that the cooling capacity utilization rate of each area is synchronized.

[0050] Furthermore, step S4 includes:

[0051] Step S41. Based on the planned static pressure setpoint and air volume setpoint, use conventional PID control of the three-way valve and variable frequency fan, and use the alternating direction multiplier algorithm to calculate the cooling capacity distribution so that the cooling capacity of each area is output synchronously.

[0052] Step S42. Input the difference between the actual heat exchange at the end of each water tank and the estimated cooling capacity into the fuzzy PID controller in real time, and output the correction amount of the air supply set value of each area. Based on the heat load prediction of each area, and with pipeline resistance and equipment capacity limit as adjustment constraints, adjust the opening of branch valves and variable frequency pump speed.

[0053] Furthermore, step S5 includes:

[0054] Step S51. Use the PMV-PPD model to evaluate the regional thermal comfort level, obtain the thermal impact level based on the equipment operating parameters, determine the regional weight based on the regional function, and add the thermal comfort level and thermal impact level of each time period according to the regional weight to obtain the regional contribution rate of each time period.

[0055] Step S52. Using the contribution rate as the power allocation weight, maximizing the weighted power load efficiency as the objective function, and ensuring that the cooling capacity meets the heat load demand as the constraint, plan the cooling capacity supply efficiency of the air conditioner at different times. Test the inertia, response speed, and time delay of the air conditioning system through step response testing, establish a transfer function model, dynamically adjust the set temperature range of the area, and perform time delay compensation control.

[0056] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0057] 1. This invention identifies personnel distribution based on the ratio of regional lighting electricity consumption to equipment operation electricity consumption, uses PCR sensors to detect the temperature of mobile devices, learns the temperature change patterns of the equipment, establishes a heat load prediction model, and realizes the optimization and control of the central air conditioning system. It matches the actual needs of the building's internal air conditioning, achieves more stable temperature and humidity control, improves indoor environmental comfort and equipment operation stability, and at the same time reduces the operating time of the chiller unit, reduces energy consumption costs, and improves the energy-saving effect of the park.

[0058] 2. This invention achieves air-water coupling by installing a return pipe at the terminal of the central air conditioning system, fitting the effects of chilled water flow rate, surface heat exchanger parameters, and air supply volume on the heat exchange performance of the surface heat exchanger and the chilled water temperature, thereby balancing the cooling and heat exchange system, reducing the peak power demand of the air conditioning system, ensuring the stable operation of air conditioning compressors, water pumps, fans, and other equipment, and improving the overall energy efficiency of the air conditioning system.

[0059] 3. This invention introduces an energy consumption contribution rate index to evaluate the contribution rate of each central air conditioning air supply area. Based on the inertia, response speed, and time delay of the air conditioning system, it regulates the energy supply efficiency of the air conditioning at different times, realizes rapid disturbance response to sudden loads, reduces waste of electricity resources, enhances the reliability of the energy-saving control system, reduces building operating energy consumption costs, and enhances the value of buildings in the park. Attached Figure Description

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0061] Figure 1 This is a schematic diagram of the structure of an energy-saving control system for equipment based on an AI large model according to the present invention;

[0062] Figure 2 This is a schematic diagram illustrating the steps of an energy-saving control method for equipment based on an AI large model according to the present invention. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Please see Figure 1 The present invention provides a technical solution: an energy-saving control system for equipment based on an AI big data model, comprising: an equipment sensing module, a load prediction module, a fuzzy control module, a wind and water system adjustment module, and an energy efficiency optimization module;

[0065] The device sensing module is used to monitor the real-time power of the device through the power monitoring socket. It constructs a multi-dimensional feature dataset using the ratio of regional lighting power consumption to equipment operation power consumption, regional functions, and power consumption time. The multi-dimensional feature dataset is input into a pre-trained machine learning model. The clustering algorithm is used to identify regional power change patterns, estimate real-time personnel distribution density, and obtain a personnel density distribution heat map. Passive infrared sensors and contact temperature sensors are deployed in the building to detect the surface temperature of the operating equipment. Based on the personnel density distribution heat map and equipment temperature, the thermal gain in the area is determined.

[0066] The device sensing module includes: a personnel distribution unit and a temperature sensing unit;

[0067] The personnel distribution unit is used to determine the sensible heat load of personnel based on power monitoring and infrared scanning, and generate a thermal map of personnel density on the interior plane of the building.

[0068] The temperature sensing unit is used to accumulate the thermal gain of each area based on the sensible heat of personnel and the sensible heat of equipment operation. The thermal gain is the additional heat generated per unit time.

[0069] The load prediction module is used to establish a heat load distribution model based on ambient temperature, set temperature and heat gain, use a genetic algorithm to optimize the backpropagation neural network to learn the heat load change pattern, identify typical heat load cycles, and output the predicted heat load in the current cycle. The heat load prediction model is established with the regional temperature and predicted heat load as the state space and the energy reduction amount as the reward function. The DF-DQN algorithm is used to reduce the actions in the temperature change space and determine the pre-control decision of the air conditioner in the cycle.

[0070] The load forecasting module includes: a large model unit and a space reduction unit;

[0071] The large model unit is used to establish a heat load distribution model using AI, predict fluctuations in ambient temperature and heat gain, and output the overall heating load.

[0072] The space reduction unit is used to model heat load prediction as a sequence decision, reduce the dimensionality of temperature changes, and generate an air conditioning cooling capacity execution sequence.

[0073] The fuzzy control module is used to install a return pipe at the air conditioning terminal, detect the inlet and outlet temperatures of chilled water, mix the return pipe with the supply water pipe, and adjust the mixing ratio through a three-way valve. Based on historical detection data and heat exchanger parameters, it fits a nonlinear heat transfer model of the surface heat exchanger, establishes the relationship between the heat exchanger's heat exchange performance and the inlet and outlet water temperatures, air volume, and static pressure, calculates the static pressure and air volume based on the heat load prediction model, and ensures that the output heat exchange meets the set temperature range requirements of equipment and personnel. Based on the PMV model, it optimizes control decisions to ensure that the cooling capacity utilization rate of each area is synchronized.

[0074] The fuzzy control module includes: a water supply loop unit, a setting optimization unit, and a global planning unit;

[0075] The water supply circuit unit is used to install a return pipe in the air conditioning terminal fan coil unit or air handling unit, and calculate the actual heat exchange by combining the water supply pipe temperature;

[0076] The setting optimization unit is used to fit the performance of the heat exchanger and generate a joint influence function for real-time heat exchange.

[0077] The global planning unit is used to calculate the duct static pressure setpoint based on the predicted cooling load and personnel distribution, using a model predictive control algorithm, and optimizing the air conditioning static pressure supply by taking these as inputs.

[0078] The air-water joint adjustment module is used to adjust the three-way valve and variable frequency fan according to the planned static pressure setpoint and air supply setpoint, so that the adjustment amount is consistent with the planned amount. The difference between the actual heat exchange at the end of each water tank and the estimated cooling amount is input into the fuzzy PID controller in real time, and the correction amount of the air supply setpoint of each area is output to redistribute the cooling amount among different ends so that the temperature of all areas reaches the set value at the same time.

[0079] The air-cooled distribution module includes: an air-cooled distribution unit and a static pressure control unit;

[0080] The air-cooled distribution unit is used to calculate the cooling capacity distribution using an alternating direction multiplier algorithm, so that the cooling capacity of each area is output synchronously.

[0081] The static pressure control unit is used to adjust the opening degree of branch valves and the speed of variable frequency pumps according to the heat load forecast of each area, while using pipeline resistance and equipment capacity limits as adjustment constraints.

[0082] The energy efficiency optimization module is used to introduce an energy consumption contribution rate index to evaluate the contribution rate of each central air conditioning air supply area, evaluate the contribution rate of temperature regulation in different areas at different times, use the contribution rate as the power allocation weight, take maximizing the weighted power load efficiency as the objective function, and take meeting the heat load demand as the constraint condition to plan the cooling capacity supply efficiency of the air conditioner at different times, and correct the set temperature range of each area according to the inertia, response speed and time delay of the air conditioning system.

[0083] The energy efficiency optimization module includes: a contribution evaluation unit, a spatiotemporal quantification unit, and an intelligent decision-making unit;

[0084] The contribution evaluation unit is used to assess the regional thermal comfort level using the PMV-PPD model, obtain the thermal impact level based on equipment operating parameters, determine the regional weight based on regional functions, and calculate the regional contribution rate by time period.

[0085] The spatiotemporal quantization unit is used to maximize the overall contribution rate and minimize the total energy consumption while meeting the cooling load requirements, and to determine the temperature control priority.

[0086] The intelligent decision-making unit is used to test the inertia, response speed, and time delay of the air conditioning system through step response testing, establish a transfer function model, dynamically adjust the set temperature range of the region, and perform time delay compensation control.

[0087] like Figure 2 As shown, an energy-saving control method for equipment based on an AI large model includes the following steps:

[0088] Step S1. Construct a multidimensional feature dataset based on the proportion of regional lighting electricity consumption, functional settings, and electricity consumption time. Input the dataset into a machine learning model to estimate the population distribution density and output a population density heat map. Based on the population density distribution heat map and the surface temperature of the operating equipment, determine the thermal gain distribution of the building space.

[0089] Step S1 includes:

[0090] Step S11. Monitor the real-time power of the equipment through the power monitoring socket, construct a multi-dimensional feature dataset using the ratio of regional lighting power consumption to equipment operation power consumption, regional function settings and power consumption time, input the multi-dimensional feature dataset into a pre-trained machine learning model, identify regional power change patterns through clustering algorithms, estimate real-time personnel distribution density, and obtain a personnel density distribution heat map.

[0091] Step S12. Deploy passive infrared sensors and contact temperature sensors inside the building to detect the surface temperature of the operating equipment. Based on the sensible heat of personnel and the sensible heat of equipment operation, accumulate the thermal gain of each area, where the thermal gain is the additional heat generated by the area per unit time.

[0092] Step S2. Establish a heat load distribution model based on ambient temperature, set temperature and heat gain, predict heat gain fluctuations, output the predicted heat load in the current cycle, reduce the dimension of predicted heat load temperature change, and generate an air conditioning cooling capacity execution sequence.

[0093] Step S2 includes:

[0094] Step S21. Based on the ambient temperature, set temperature and heat gain, use AI to establish a heat load distribution model, use a genetic algorithm to optimize the backpropagation neural network to learn the heat load change pattern, identify typical heat load cycles, and output the predicted heat load in the current cycle.

[0095] Step S22. Establish a heat load prediction model with regional temperature and predicted heat load as the state space and energy reduction amount as the reward function. Use the DF-DQN algorithm to reduce the actions in the temperature change space, determine the pre-control decision of air conditioning within the cycle, and generate a cooling capacity execution sequence, which includes cooling capacity and cooling time.

[0096] Step S3. Install a return pipe at the air conditioning terminal, detect the inlet and outlet temperatures of chilled water, establish the relationship between the heat exchanger's heat exchange performance and the inlet and outlet water temperatures, air volume, and static pressure, calculate the static pressure and air volume according to the execution sequence, so that the output cooling capacity controls the area temperature within the set temperature range for equipment and personnel.

[0097] Step S3 includes:

[0098] Step S31. Install a return pipe in the fan coil unit or air handling unit of the air conditioner terminal, detect the inlet and outlet water temperatures of chilled water, mix the return pipe with the supply water pipe, and adjust the mixing ratio through a three-way valve. Based on historical detection data and heat exchanger parameters, fit a nonlinear heat transfer model of the surface heat exchanger, and establish the relationship between the heat exchanger heat transfer performance and the inlet and outlet water temperatures, air volume and static pressure.

[0099] Step S32. Using the predicted cooling load and personnel distribution as input, the model predictive control algorithm is used to plan the duct static pressure setpoint based on the duct resistance characteristics and fan performance curve, so that the area temperature is within the set temperature range of equipment and personnel, and the control decision is optimized based on the PMV model to ensure that the cooling capacity utilization rate of each area is synchronized.

[0100] Step S4. Adjust the static pressure and air volume of the air conditioner, determine the actual heat exchange based on the inlet and outlet water temperature and flow rate, input the difference between the actual heat exchange and the estimated cooling capacity into the fuzzy PID controller in real time, correct the air volume set value of each area, and redistribute the cooling capacity.

[0101] Step S4 includes:

[0102] Step S41. Based on the planned static pressure setpoint and air volume setpoint, use conventional PID control of the three-way valve and variable frequency fan, and use the alternating direction multiplier algorithm to calculate the cooling capacity distribution so that the cooling capacity of each area is output synchronously.

[0103] Step S42. Input the difference between the actual heat exchange at the end of each water tank and the estimated cooling capacity into the fuzzy PID controller in real time, and output the correction amount of the air supply set value of each area. Based on the heat load prediction of each area, and with pipeline resistance and equipment capacity limit as adjustment constraints, adjust the opening of branch valves and variable frequency pump speed.

[0104] Step S5. Based on personnel distribution, equipment operating parameters, and regional functions, calculate the regional contribution rate in different time periods, plan the cooling capacity supply efficiency for different time periods, maximize the average contribution rate of power load while meeting the cooling capacity demand, and adjust the set temperature range of each region according to the inertia, response speed, and time delay of the air conditioning system.

[0105] Step S5 includes:

[0106] Step S51. Use the PMV-PPD model to evaluate the regional thermal comfort level, obtain the thermal impact level based on the equipment operating parameters, determine the regional weight based on the regional function, and add the thermal comfort level and thermal impact level of each time period according to the regional weight to obtain the regional contribution rate of each time period.

[0107] Step S52. Using the contribution rate as the power allocation weight, maximizing the weighted power load efficiency as the objective function, and ensuring that the cooling capacity meets the heat load demand as the constraint, plan the cooling capacity supply efficiency of the air conditioner at different times. Test the inertia, response speed, and time delay of the air conditioning system through step response testing, establish a transfer function model, dynamically adjust the set temperature range of the area, and perform time delay compensation control.

[0108] Example: A single-story building has three areas: offices, a computer room, and a control room. The air conditioning system uses an AI model to predict the heat gain distribution within the area and determines that the cooling capacity distribution at the current ambient temperature is 20,000 KJ, 50,000 KJ, and 10,000 KJ. The system will then transfer heat according to the cooling capacity distribution for the next period. During the cooling process, the return water temperature is monitored, and the air supply volume is adjusted to ensure that the temperature in each area is within the temperature range determined by the contribution rate index. The static pressure distribution within the area is adjusted so that the cooling capacity in each area is exhausted simultaneously.

[0109] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0110] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI large model-based device energy saving control method, characterized by, The method comprises the following steps: Step S1. Construct a multi-dimensional feature data set according to the proportion of regional lighting electricity consumption, function setting and electricity consumption time, input the machine learning model to estimate the personnel distribution density, output the personnel density heat map, and determine the building space heat gain distribution according to the personnel density distribution heat map and the surface temperature of the running equipment; Step S2. Establish a heat load distribution model according to the environment temperature, set temperature and heat gain, predict the heat gain fluctuation, output the predicted heat load in the current period, reduce the dimension of the predicted heat load temperature change, and generate an air conditioning refrigeration quantity execution sequence; Step S3. Set a return pipe at the air conditioning terminal, detect the chilled water inlet and outlet water temperature, establish the relationship between the heat exchanger heat exchange performance and the inlet and outlet water temperature, air supply quantity and static pressure, calculate the static pressure and air supply quantity according to the execution sequence, and control the output refrigeration quantity in the equipment and personnel set temperature interval; Step S4. Adjust the air conditioning static pressure and air supply quantity, determine the actual heat exchange capacity according to the inlet and outlet water temperature and flow, input the difference between the actual heat exchange capacity and the estimated refrigeration quantity into the fuzzy PID controller in real time, correct the regional air supply quantity set value, and re-distribute the refrigeration quantity; Step S5. According to the personnel distribution, equipment operation parameters and regional function, calculate the regional contribution rate in time periods, plan the refrigeration quantity supply efficiency in different time periods, maximize the average contribution rate of the power load under the premise of meeting the refrigeration quantity demand, and correct the set temperature interval of each region according to the inertia, response speed and time lag of the air conditioning system; Step S2 comprises: Step S21. According to the environment temperature, set temperature and heat gain, an AI is used to establish a heat load distribution model, a genetic algorithm is used to optimize a back propagation neural network to learn the heat load change mode, a typical heat load cycle is identified, and a predicted heat load in the current period is output; Step S22. A heat load prediction model is established by taking the regional temperature and the predicted heat load as a state space and taking the energy reduction as a reward function, and a model function table is: ； where C is the indoor thermal capacity, R is the building thermal resistance, T out (t) and T in (t) are the outdoor temperature and indoor temperature functions, respectively, Q load (t) is the internal heat load, Q cool (t) is the air conditioning refrigeration capacity, is the differential of the indoor temperature with respect to time, and t is time. The DF-DQN algorithm is used to reduce the action in the temperature change space, a pre-control decision of the air conditioner in the period is determined, and a refrigeration quantity execution sequence is generated, wherein the refrigeration quantity execution sequence comprises refrigeration quantity and refrigeration time; Step S5 comprises: Step S51. The PMV-PPD model is used to evaluate the regional thermal comfort level, the heat influence level is obtained according to the equipment operation parameters, the regional weight is determined according to the regional function, the thermal comfort level and the heat influence level in each time period are weighted and added according to the regional weight, and the regional contribution rate in the time period is obtained; Step S52. The contribution rate is taken as the power distribution weight, the maximum weighted power load efficiency is taken as the objective function, the refrigeration quantity meets the heat load demand is taken as the restriction condition, the refrigeration quantity supply efficiency of the air conditioner in different time periods is planned, the inertia, response speed and time lag of the air conditioning system are tested through a step response, a transfer function model is established, the set temperature interval of each region is dynamically adjusted, and time lag compensation control is performed.

2. The AI large model-based device energy saving control method of claim 1, wherein: Step S1 comprises: Step S11. Monitor the real-time power of the equipment through the power monitoring socket, construct a multi-dimensional feature data set using the proportion of regional lighting power consumption and equipment operation power consumption, regional function settings, and power consumption time, and input the multi-dimensional feature data set into a pre-trained machine learning model; In each dimension of the multi-dimensional feature data set, select K time series as initial cluster centers, assign features to cluster centers, and minimize the objective function: ； Wherein, J is the within-cluster sum of squares, K is the number of clusters, determined by the data silhouette coefficient, is the center of the kth cluster, x i is the i th sample point, is the Euclidean distance from the sample point to the cluster center, C k Indicates the kth cluster, determines the cluster with the minimum J, and outputs the cluster center of each cluster as a power change mode, estimates the real-time personnel distribution density according to the regional power change mode, and obtains a personnel density distribution heat map; Step S12. Deploy passive infrared sensors and contact temperature sensors inside the building to detect the surface temperature of the running equipment, and accumulate the heat gain of each region according to the sensible heat of personnel and the sensible heat of equipment operation. The heat gain is the additional heat generated by the region per unit time.

3. The AI large model-based device energy saving control method of claim 2, characterized in that: Step S3 includes: Step S31. Set a return pipe in the air conditioning terminal fan coil or air handling unit to detect the inlet and outlet water temperature of chilled water. The return pipe is mixed with the water supply pipe, and the mixing ratio is adjusted by a three-way valve. According to historical detection data and heat exchanger parameters, a nonlinear heat transfer model of the surface heat exchanger is fitted, and a relationship between heat exchanger performance, inlet and outlet water temperature, air supply volume, and static pressure is established; Step S32. Use the model predictive control algorithm to plan the static pressure set value based on the air duct resistance characteristics and the fan performance curve, so that the region temperature is within the set temperature interval of the equipment and personnel, and optimize the control decision based on the PMV model to ensure that the cold quantity usage rate of each region is synchronized.

4. The AI large model-based device energy saving control method of claim 3, characterized in that: Step S4 includes: Step S41. According to the planned static pressure set value and air supply volume set value, use conventional PID control three-way valve and variable frequency fan, and use alternating direction multiplier algorithm to calculate cold quantity distribution, so that each region cold quantity is output synchronously; Step S42. Real-time input the difference between the actual heat exchange capacity of each water tank terminal and the estimated cold quantity into the fuzzy PID controller, output the correction amount of each region air supply volume set value, predict the heat load of each region, and adjust the branch valve opening and variable frequency pump speed based on the pipeline resistance and equipment capacity constraints.

5. An AI large model-based device energy saving control system, characterized by, The system includes the following modules: equipment perception module, load prediction module, fuzzy control module, air-water joint debugging module, and energy efficiency optimization module; The equipment perception module is used to monitor the real-time power of the equipment through the power monitoring socket, construct a multi-dimensional feature data set using the proportion of regional lighting power consumption and equipment operation power consumption, regional function and power consumption time, input the multi-dimensional feature data set into a pre-trained machine learning model, identify regional power change patterns through clustering algorithm, estimate real-time personnel distribution density, obtain personnel density distribution thermograph, deploy passive infrared sensors and contact temperature sensors inside the building to detect the surface temperature of the running equipment, and determine the heat gain in the region according to the personnel density distribution thermograph and equipment temperature. The load prediction module is configured to establish a heat load distribution model according to an ambient temperature, a set temperature and a heat gain, to learn a heat load change mode by using a genetic algorithm to optimize a back propagation neural network, to identify a typical heat load cycle, and to output a predicted heat load in a current cycle, to establish a heat load prediction model by taking a regional temperature and the predicted heat load as a state space and taking an energy reduction as a reward function, and to determine a pre-control decision for the air conditioner in the cycle by reducing actions in a temperature change space by using a DF-DQN algorithm. The fuzzy control module is configured to set a return pipe at an air conditioner terminal, to detect inlet and outlet water temperatures of chilled water, to mix the return pipe with a water supply pipe and to adjust a mixing ratio by using a three-way valve, to fit a nonlinear heat transfer model of a surface heat exchanger according to historical detection data and heat exchanger parameters, to establish a relationship between heat exchanger heat exchange performance and inlet and outlet water temperatures, air supply and static pressure, to calculate the static pressure and the air supply according to the heat load prediction model, so that an output heat exchange amount meets a set temperature interval requirement of equipment and personnel in a region, to optimize a control decision based on a PMV model, and to ensure that cold energy usage rates of the regions are synchronized. The air-water joint adjustment module is configured to set a static pressure set value and an air supply set value according to a plan, to control the three-way valve and a variable frequency fan by using a conventional PID control, to make an adjustment amount consistent with a planned amount, to input a difference between actual heat exchange amounts of water tanks at terminals and estimated cold energy into a fuzzy PID controller in real time, to output a correction amount of the air supply set value of each region, and to redistribute the cold energy among different terminals, so that temperatures of all regions reach the set values at the same time. The energy efficiency optimization module is configured to introduce an energy consumption contribution rate index, to evaluate a contribution rate of each central air conditioner air supply region, to evaluate the contribution rate of temperature adjustment of different regions in different time periods, to take the contribution rate as a power distribution weight, to maximize a weighted power load efficiency as an objective function, to plan a refrigeration amount supply efficiency of the air conditioner in different time periods under a condition that the refrigeration amount meets a heat load requirement, and to correct a set temperature interval of each region according to inertia, response speed and time lag of the air conditioning system. The energy efficiency optimization module comprises a contribution evaluation unit, a space-time quantization unit and an intelligent decision unit. The contribution evaluation unit is configured to evaluate a regional thermal comfort level by using a PMV-PPD model, to obtain a thermal influence level according to equipment operation parameters, to determine a regional weight according to a region function, and to calculate a regional contribution rate in different time periods. The space-time quantization unit is configured to maximize an overall contribution rate and minimize total energy consumption under a condition that a refrigeration load requirement is met, and to determine a temperature adjustment priority. The intelligent decision unit is configured to test inertia, response speed and time lag of the air conditioning system by using a step response, to establish a transfer function model, to dynamically adjust a set temperature interval of a region, and to perform time lag compensation control.

6. The AI large model-based device energy saving control system according to claim 5, characterized in that: The device perception module comprises a personnel distribution unit and a temperature measurement sensing unit. The personnel distribution unit is configured to determine a personnel sensible heat load according to power monitoring and infrared scanning, and to generate a personnel density thermal map of a building internal plane. The temperature measurement sensing unit is configured to accumulate a heat gain of each region according to personnel sensible heat and equipment operation sensible heat, the heat gain being an additional heat amount generated per unit time. The load prediction module comprises a large model unit and a spatial reduction unit; The large model unit is configured to establish a heat load distribution model by using AI, predict environmental temperature and heat gain fluctuation, and output overall heating load; The spatial reduction unit is configured to model the heat load prediction as a sequence decision, reduce the dimension of temperature change, and generate an air conditioning refrigeration quantity execution sequence.

7. The AI large model-based device energy saving control system according to claim 6, characterized in that: The fuzzy control module comprises a water supply circuit unit, a setting optimization unit, and a global planning unit; The water supply circuit unit is configured to set a return pipe in an air conditioner terminal fan coil or an air handling unit, combine with a water supply pipe temperature, and calculate actual heat exchange capacity; The setting optimization unit is configured to perform performance fitting on a heat exchanger, and generate a joint influence function of real-time heat exchange capacity; The global planning unit is configured to use a model predictive control algorithm, calculate a wind pipe static pressure setting value based on wind pipe resistance characteristics and fan performance curve, and optimize air conditioning static pressure air supply, with predicted cooling load and personnel distribution as input.

8. The AI large model-based device energy saving control system according to claim 7, characterized in that: The air-water joint adjustment module comprises a wind cooling distribution unit and a static pressure control unit; The wind cooling distribution unit is configured to calculate cooling capacity distribution by using an alternating direction multiplier algorithm, so that cooling capacity of each region is synchronously outputted; The static pressure control unit is configured to adjust branch valve opening degree and variable frequency pump speed according to heat load prediction of each region, while taking pipe resistance and equipment capacity limit as adjustment constraints.

Images