AI-based smart building HVAC energy-saving control system

Abstract

The present application relates to air conditioning energy saving control technical field, specifically for the energy saving control system of heating ventilation air conditioner in intelligent building based on artificial intelligence, including refrigeration machine modeling unit, total energy consumption modeling unit, target modeling unit, constraint modeling unit, energy consumption optimization modeling unit, model solution unit, air conditioning control unit, state prediction unit and energy saving control unit, it also includes: fan modeling unit, the fan modeling unit is used for based on intelligent building indoor cold load, intelligent building indoor set temperature and supply air temperature constructs the fan air outlet model of target heating ventilation air conditioner, the present application utilizes fan modeling unit, refrigerated water pump modeling unit and refrigeration machine modeling unit, through establishing and optimizing model to accurately predict and control the energy consumption of heating ventilation air conditioner system, so as to ensure the accurate control of the energy consumption of each component of the system, thereby maximizing energy efficiency.

Description

Technical Field

[0001] This invention relates to the field of air conditioning energy-saving control technology, specifically to an energy-saving control system for HVAC in smart buildings based on artificial intelligence. Background Technology

[0002] Smart buildings refer to a type of building that uses advanced information technologies (such as the Internet of Things, sensors, data analytics, and artificial intelligence) and automated control technologies to manage and optimize the operation and maintenance of building facilities, thereby improving the building's operational efficiency, comfort, safety, and sustainability.

[0003] Traditional systems typically lack precise energy consumption modeling and optimization capabilities, making it impossible to accurately control the energy consumption of individual system components (such as fans, chilled water pumps, and chillers). This often results in the inability to effectively minimize energy consumption during actual operation, leading to low energy efficiency and energy waste. Furthermore, traditional systems typically cannot monitor and predict the operating status of HVAC systems in real time, and lack predictive intelligent control capabilities. This makes it difficult for the system to dynamically adjust to environmental changes and fluctuations in energy demand during operation, thus affecting system stability and efficiency. In addition, traditional systems often use simple adjustment strategies and fixed control parameters for energy consumption optimization, lacking the support of complex optimization algorithms. This prevents the system from automatically adjusting to achieve optimal energy efficiency under different operating conditions, missing opportunities for energy saving and cost reduction. Summary of the Invention

[0004] The purpose of this invention is to address the problems existing in the background technology by proposing an energy-saving control system for HVAC in smart buildings based on artificial intelligence.

[0005] The technical solution of this invention: An energy-saving control system for HVAC in smart buildings based on artificial intelligence, comprising a chiller modeling unit, a total energy consumption modeling unit, a target modeling unit, a constraint modeling unit, an energy consumption optimization modeling unit, a model solving unit, an air conditioning control unit, a state prediction unit, and an energy-saving control unit, and further comprising:

[0006] The fan modeling unit is used to construct a fan outlet model of the target HVAC system based on the indoor cooling load, indoor set temperature, and supply air temperature of the smart building; obtain the supply air flow rate of the target HVAC system based on the fan outlet model of the target HVAC system; construct a fan energy consumption model of the target HVAC system based on the supply air flow rate of the target HVAC system; obtain the fan energy consumption of the target HVAC system based on the fan energy consumption model of the target HVAC system; transmit the supply air flow rate of the target HVAC system to the chilled water pump modeling unit; and transmit the fan energy consumption of the target HVAC system to the total energy consumption modeling unit.

[0007] The chilled water pump modeling unit receives the airflow rate of the target HVAC system transmitted by the fan modeling unit, and constructs a chilled water pump refrigeration model of the target HVAC system based on the airflow temperature, the inlet air temperature of the surface cooler, the temperature difference between the chilled water supply and return water, and the airflow rate of the target HVAC system. Based on the chilled water pump refrigeration model, it obtains the chilled water flow rate of the target HVAC system, constructs a chilled water pump energy consumption model of the target HVAC system based on the chilled water flow rate, obtains the chilled water pump energy consumption of the target HVAC system based on the chilled water pump energy consumption model, transmits the chilled water flow rate of the target HVAC system to the chiller modeling unit, and transmits the chilled water pump energy consumption of the target HVAC system to the total energy consumption modeling unit.

[0008] Preferably, the chiller modeling unit receives the chilled water flow rate of the target HVAC system transmitted by the chilled water pump modeling unit, and uses it to construct a chiller performance model of the target HVAC system based on the chilled water outlet temperature, cooling water return temperature, the chilled water supply and return temperature difference, and the chilled water flow rate of the target HVAC system. Based on the chiller performance model of the target HVAC system, it obtains the chiller performance curve of the target HVAC system, constructs a chiller energy consumption model of the target HVAC system based on the chiller performance curve, obtains the chiller energy consumption of the target HVAC system based on the chiller energy consumption model of the target HVAC system, and transmits the chiller energy consumption of the target HVAC system to the total energy consumption modeling unit.

[0009] Preferably, the total energy consumption modeling unit receives the fan energy consumption of the target HVAC system transmitted by the fan modeling unit, the chilled water pump energy consumption of the target HVAC system transmitted by the chilled water pump modeling unit, and the chiller energy consumption of the target HVAC system transmitted by the chiller modeling unit. It then constructs a total power model of the target HVAC system based on the fan energy consumption, chilled water pump energy consumption, and chiller energy consumption, obtains the total power of the target HVAC system based on the total power model, and transmits the total power of the target HVAC system to the target modeling unit.

[0010] Preferably, the target modeling unit receives the total power of the target HVAC system transmitted by the total energy consumption modeling unit, and uses it to construct a target optimization function to minimize the total power of the target HVAC system, and transmits the target optimization function to the energy consumption optimization modeling unit.

[0011] Preferably, the constraint modeling unit is used to construct the constraint conditions of the target HVAC based on the operating parameters of the target HVAC, and transmit the constraint conditions of the target HVAC to the energy consumption optimization modeling unit.

[0012] Preferably, the energy consumption optimization modeling unit receives the target optimization function transmitted by the target modeling unit and the constraint conditions of the target HVAC system transmitted by the constraint modeling unit, and uses it to construct an energy consumption optimization model based on the target optimization function and the constraint conditions of the target HVAC system, and transmits the energy consumption optimization model to the model solving unit.

[0013] Preferably, the model solving unit receives the energy consumption optimization model transmitted by the energy consumption optimization modeling unit, and uses it to solve the energy consumption optimization model through an optimization algorithm to obtain the first control parameters of the target HVAC system, and transmits the first control parameters of the target HVAC system to the air conditioning control unit.

[0014] Preferably, the air conditioning control unit receives the first control parameters of the target HVAC system transmitted by the model solving unit, and controls the target HVAC system based on the first control parameters. It also monitors the operating status data of the target HVAC system in real time to obtain the operating status data, which includes compressor high-pressure pressure, compressor low-pressure pressure, actual frequency of the inverter equipment, outdoor unit fan output speed, opening degree of the two expansion valves of the indoor unit, current indoor temperature of the smart building, and the current energy efficiency ratio of the target HVAC system. The operating status data of the target HVAC system is then transmitted to the status prediction unit.

[0015] Preferably, the state prediction unit receives the operating status data of the target HVAC system transmitted by the air conditioning control unit, and inputs the operating status data of the target HVAC system into the trained state prediction model. The trained state prediction model outputs the operating status data of the target HVAC system at the next preset time. The state prediction model adopts a neural network model and transmits the operating status data of the target HVAC system at the next preset time to the energy-saving control unit.

[0016] Preferably, the energy-saving control unit receives the indoor temperature and energy efficiency ratio of the target HVAC system at the next preset time transmitted by the state prediction unit, and constructs a first intelligent agent of the target HVAC outdoor unit and a second intelligent agent of the target HVAC indoor unit through reinforcement learning based on the indoor temperature and energy efficiency ratio of the target HVAC system at the next preset time. The first intelligent agent and the second intelligent agent are solved to obtain the second control parameters of the target HVAC system, and the target HVAC system is controlled based on the second control parameters.

[0017] Preferably, the fan outlet model of the target HVAC system is as follows:

[0018]

[0019] Where, m a Q represents the actual airflow rate delivered by the fan. S Indicates the indoor cooling load of a smart building, T N Indicates the set indoor temperature of a smart building, T S Indicates the supply air temperature;

[0020] The fan energy consumption model for the target HVAC system is as follows:

[0021]

[0022] Among them, P fan The actual power of the wind turbine is represented by α, a1, b1, c1, d1, and e1 represent the wind turbine characteristic fitting coefficients, and α represents the actual power of the wind turbine. fan ρ represents the overall efficiency of the fan. air β represents air density. a This indicates the main load rate of the wind turbine, and m a-L This indicates the rated airflow of the fan;

[0023] The chilled water pump chilling model of the target HVAC system is as follows:

[0024]

[0025] Where, m e The temperature represents the actual chilled water flow rate of the chilled water pump, T1 represents the inlet air temperature of the surface cooler, and T2 represents the outlet air temperature of the surface cooler. p ΔT represents the specific heat of water. e Indicates the temperature difference between chilled water supply and return;

[0026] The energy consumption model of the chilled water pump of the target HVAC system is as follows:

[0027]

[0028] Among them, P e The value represents the actual power of the chilled water pump, and a2, b2, c2, and d2 represent the characteristic fitting coefficients of the chilled water pump. β e This indicates the main load rate of the chilled water pump, and m e-L P represents the rated chilled water flow rate of the chilled water pump. e-L This indicates the rated power of the chilled water pump.

[0029] Preferably, the performance model of the chiller of the target HVAC system is as follows:

[0030]

[0031] Among them, Q e Indicates the cooling capacity of the refrigeration system, ηch1 The curve representing the relationship between cooling capacity and temperature, η ch2 The curve representing the reciprocal of the energy efficiency ratio as a function of chilled water outlet temperature and cooling water return temperature, η ch3 The curve represents the relationship between the reciprocal of the energy efficiency ratio and the main load rate of the chiller. a3, b3, c3, d3, e3, and f3 represent the fitting coefficients for the cooling capacity characteristics; a4, b4, c4, d4, e4, and f4 represent the fitting coefficients for the chilled water characteristics; a5, b5, and c5 represent the fitting coefficients for the main load rate characteristics of the chiller; T eo T represents the chilled water outlet temperature of the chiller. ci Indicates the cooling water return temperature, β ch This indicates the main load rate of the refrigeration unit, and Q avail Indicates the available cooling capacity of the refrigeration unit;

[0032] The energy consumption model of the target HVAC system is as follows:

[0033]

[0034] Among them, P ch C represents the actual power of the refrigeration unit. COP-L This represents the rated energy efficiency ratio of the refrigeration unit, and Q... avail =Q N η ch1 Q N This indicates the rated cooling capacity of the refrigeration unit;

[0035] The energy consumption optimization model is as follows:

[0036]

[0037] Where min P represents the objective function, st represents the constraints, and m a-min and m a-max T represents the minimum and maximum actual airflow rates of the fan, respectively. S-min and T S-max These represent the minimum and maximum values ​​of the supply air temperature, respectively, m. e-min and m e-max ΔT represents the minimum and maximum actual chilled water flow rates of the chilled water pump, respectively. e-min and ΔT e-max T represents the minimum and maximum temperature difference between the chilled water supply and return water, respectively. eo-min and T eo-max T represents the minimum and maximum outlet water temperature of the chiller, respectively. ci-min and T ci-max These represent the minimum and maximum values ​​of the cooling water return temperature, respectively.

[0038] Preferably, the energy consumption optimization model is solved using an optimization algorithm to obtain the first control parameters of the target HVAC system, including the following steps:

[0039] A1. Set the proportion of discoverers (PD), the proportion of scouts (SD), and the alert value (R2), and initialize the population. The population initialization formula is as follows:

[0040]

[0041] Where, x i and x i+1 Let i represent the i-th and (i+1)-th individuals in the population, and mod represents the remainder function;

[0042] A2. Calculate the fitness value of each individual in the population, sort them according to their fitness values, select the discoverer from the individuals whose fitness values ​​are greater than a preset fitness value threshold, and update the position of the discoverer. The position update formula for the discoverer is as follows:

[0043]

[0044] in, and Let r1 represent the position of the discoverer with coordinates (i,j) in the population during the t-th and t+1-th iterations in the d-dimensional solution space, and let r2 represent the random number that determines the distance the discoverer moves. Let y1 and y2 represent the optimal position of the current discoverer in the t-th iteration, and let ST represent the safety threshold.

[0045] A3. After selecting the discoverer from the population, the remaining individuals in the population are designated as new members, and the positions of these new members are updated using the following formula:

[0046]

[0047] in, and Let represent the position of the joiner with coordinates (i,j) in the d-dimensional solution space during the t-th and t+1-th iterations. This represents the worst-case global position of the current joiner in the t-th iteration. Let A represent the optimal position of the current joiner in the (t+1)th iteration, let L represent the identity matrix, and let n represent the total number of individuals in the population.

[0048] A4. Randomly select scouts from the population and update their positions using the following formula:

[0049]

[0050] in, and Let represent the position of the scout with coordinates (i,j) in the population during the t-th and t+1-th iterations in the d-dimensional solution space. This represents the worst-case global position of the current scout in the t-th iteration. Let f represent the optimal position of the scout in the (t+1)th iteration, β represent a random number that follows a Gaussian distribution with a mean of 0 and a variance of 1, γ be 10E⁻⁸, K represent the random coefficients for direction and step size, and f i f represents the fitness value of the investigator. g and f worst This represents the investigator's best and worst fitness values;

[0051] A5. Perturb the current optimal solution, recalculate the population fitness, and find the optimal solution after perturbation. The perturbation formula for the current optimal solution is as follows:

[0052]

[0053] in, X represents the position of the individual with coordinates (i,j) in the population in the (t+1)th iteration in the d-dimensional solution space. best (t) represents the current optimal solution, and cauchy(0,1) represents the standard Cauchy distribution function;

[0054] A6. Determine if the termination condition has been met. If the termination condition has been met, output the optimal individual position and its fitness value. Otherwise, repeat step A2.

[0055] Preferably, the first intelligent agent of the target HVAC outdoor unit includes a first state, a first action, and a first reward function, and the expression for the first state is as follows:

[0056] S0 = [T] set ,T out ,EER,T in ];

[0057] Where S0 represents the first state, T set This indicates the set temperature of the indoor unit, T. out The outdoor temperature is represented by EER, and the energy efficiency ratio for the next preset time is represented by T. in Indicates the indoor temperature at the next preset time;

[0058] The first action expression is as follows:

[0059] A0 = [f com,PD high ,PD low ,f fan ];

[0060] Where A0 represents the first action, f com PD represents the actual frequency of the frequency converter. high PD indicates the high pressure of the compressor. low f represents the low-pressure value of the compressor. fan Indicates the output fan speed of the outdoor unit;

[0061] The expression for the first reward function is as follows:

[0062] R0 = EER;

[0063] Where R0 represents the first reward function;

[0064] The second intelligent agent of the target HVAC indoor unit includes a second state, a second action, and a second reward function. The expression for the second state is as follows:

[0065]

[0066] Among them, S i Indicates the second state. This represents the set temperature of the i-th indoor unit. This indicates the indoor temperature at the next preset time for the i-th indoor unit;

[0067] The second action expression is as follows:

[0068]

[0069] Among them, A i Indicates the second action. and This indicates the opening degree of the two expansion valves of the indoor unit;

[0070] The expression for the second reward function is as follows:

[0071]

[0072] Among them, R i This represents the second reward function.

[0073] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects:

[0074] 1. This invention utilizes a fan modeling unit, a chilled water pump modeling unit, and a chiller modeling unit to accurately predict and control the energy consumption of the HVAC system by establishing and optimizing models. This ensures precise control over the energy consumption of each component of the system (fan, chilled water pump, chiller), thereby maximizing energy efficiency. Furthermore, through an energy consumption optimization modeling unit and a model solving unit, optimization algorithms are used to solve the energy consumption model to obtain the optimal control parameters. This enables the HVAC system to provide the required cooling effect with the lowest energy consumption under different operating conditions, thereby saving energy costs.

[0075] 2. This invention monitors the operating status data of HVAC in real time through a state prediction unit, predicts the operating status in future time periods using a trained state prediction model, and optimizes the system control using reinforcement learning technology. The optimal control parameters are obtained through the solution of intelligent agents (outdoor and indoor units), further improving the system's energy efficiency and performance. The air conditioning control unit is responsible for real-time control of the HVAC system and monitoring the operating status data, thereby ensuring that the system can dynamically adjust during operation to adapt to environmental changes and fluctuations in energy consumption demand, thus ensuring system stability and efficiency. Attached Figure Description

[0076] Figure 1 This is a flowchart of the overall system in one embodiment of the present invention.

[0077] Figure labels: 1. Fan modeling unit; 2. Chilled water pump modeling unit; 3. Chiller modeling unit; 4. Total energy consumption modeling unit; 5. Target modeling unit; 6. Constraint modeling unit; 7. Energy consumption optimization modeling unit; 8. Model solving unit; 9. Air conditioning control unit; 10. State prediction unit; 11. Energy saving control unit. Detailed Implementation

[0078] Example 1, as Figure 1 As shown, the energy-saving control system for HVAC in smart buildings based on artificial intelligence proposed in this invention includes a chiller modeling unit 3, a total energy consumption modeling unit 4, a target modeling unit 5, a constraint modeling unit 6, an energy consumption optimization modeling unit 7, a model solving unit 8, an air conditioning control unit 9, a state prediction unit 10, and an energy-saving control unit 11, and further includes:

[0079] Fan Modeling Unit 1 is used to construct the fan outlet model of the target HVAC system based on the indoor cooling load, indoor set temperature and supply air temperature of the smart building; obtain the supply air flow of the target HVAC system based on the fan outlet model of the target HVAC system; construct the fan energy consumption model of the target HVAC system based on the supply air flow of the target HVAC system; obtain the fan energy consumption of the target HVAC system based on the fan energy consumption model of the target HVAC system; transmit the supply air flow of the target HVAC system to the chilled water pump modeling unit 2; and transmit the fan energy consumption of the target HVAC system to the total energy consumption modeling unit 4.

[0080] Chilled water pump modeling unit 2 receives the air flow rate of the target HVAC system transmitted by fan modeling unit 1, and uses it to construct a chilled water pump chilling model of the target HVAC system based on the air supply temperature, surface cooler inlet air temperature, chilled water supply and return water temperature difference, and the air flow rate of the target HVAC system. Based on the chilled water pump chilling model of the target HVAC system, it obtains the chilled water flow rate of the target HVAC system, constructs a chilled water pump energy consumption model of the target HVAC system based on the chilled water flow rate of the target HVAC system, obtains the chilled water pump energy consumption of the target HVAC system based on the chilled water pump energy consumption model of the target HVAC system, transmits the chilled water flow rate of the target HVAC system to chiller modeling unit 3, and transmits the chilled water pump energy consumption of the target HVAC system to total energy consumption modeling unit 4.

[0081] In this invention, indoor cooling load refers to the total amount of heat inside a building that needs to be cooled by the air conditioning system, which is usually determined by indoor heat sources, external environmental conditions, and building structure; indoor set temperature refers to the indoor comfort temperature set by the user, and the air conditioning system controls the indoor environment according to this setting; supply air temperature refers to the temperature of the air supplied to the room from the air conditioning system; supply air flow rate refers to the volumetric flow rate of air delivered to the room from the air outlet by the air conditioning system; chilled water flow rate refers to the flow rate of water circulating through the cooling equipment, which is used to control and maintain the operation of the refrigeration system.

[0082] In an optional embodiment, the chiller modeling unit 3 receives the chilled water flow rate of the target HVAC system transmitted by the chilled water pump modeling unit 2, and uses it to construct a chiller performance model of the target HVAC system based on the chilled water outlet temperature, cooling water return temperature, chilled water supply and return temperature difference, and the chilled water flow rate of the target HVAC system. Based on the chiller performance model of the target HVAC system, the chiller performance curve of the target HVAC system is obtained. Based on the chiller performance curve of the target HVAC system, the chiller energy consumption model of the target HVAC system is constructed. Based on the chiller energy consumption model of the target HVAC system, the chiller energy consumption of the target HVAC system is obtained. The chiller energy consumption of the target HVAC system is then transmitted to the total energy consumption modeling unit 4.

[0083] It should be noted that chilled water outlet temperature refers to the temperature of chilled water exiting the chiller unit in the refrigeration system; cooling water return temperature refers to the temperature of cooling water returning to the chiller unit from the cooling equipment in the refrigeration system; and chilled water supply and return temperature difference refers to the temperature difference between the chilled water entering and leaving the chiller unit.

[0084] In an optional embodiment, the total energy consumption modeling unit 4 receives the fan energy consumption of the target HVAC system transmitted by the fan modeling unit 1, the chilled water pump energy consumption of the target HVAC system transmitted by the chilled water pump modeling unit 2, and the chiller energy consumption of the target HVAC system transmitted by the chiller modeling unit 3. It then uses this data to construct a total power model of the target HVAC system based on the fan energy consumption, chilled water pump energy consumption, and chiller energy consumption. Based on the total power model of the target HVAC system, it obtains the total power of the target HVAC system and transmits the total power of the target HVAC system to the target modeling unit 5.

[0085] In an optional embodiment, the target modeling unit 5 receives the total power of the target HVAC system transmitted by the total energy consumption modeling unit 4, and uses it to construct a target optimization function to minimize the total power of the target HVAC system, and transmits the target optimization function to the energy consumption optimization modeling unit 7.

[0086] In an optional embodiment, the constraint modeling unit 6 is used to construct the constraint conditions of the target HVAC based on the operating parameters of the target HVAC, and transmit the constraint conditions of the target HVAC to the energy consumption optimization modeling unit 7.

[0087] In an optional embodiment, the energy consumption optimization modeling unit 7 receives the target optimization function transmitted by the target modeling unit 5 and the target HVAC constraint conditions transmitted by the constraint modeling unit 6, and uses it to construct an energy consumption optimization model based on the target optimization function and the target HVAC constraint conditions, and transmits the energy consumption optimization model to the model solving unit 8.

[0088] In an optional embodiment, the model solving unit 8 receives the energy consumption optimization model transmitted by the energy consumption optimization modeling unit 7, and uses it to solve the energy consumption optimization model through an optimization algorithm to obtain the first control parameters of the target HVAC system, and transmits the first control parameters of the target HVAC system to the air conditioning control unit 9.

[0089] It should be noted that optimization algorithms are used to solve the energy consumption model in order to obtain the first control parameters of the target HVAC system. Optimization algorithms refer to a class of mathematical methods and computational techniques that aim to find the optimal or near-optimal solution to a problem. They are commonly used when facing complex multivariable, multi-condition, and multi-constraint problems.

[0090] In an optional embodiment, the air conditioning control unit 9 receives the first control parameters of the target HVAC system transmitted by the model solving unit 8, and uses them to control the target HVAC system based on the first control parameters. It also monitors the operating status data of the target HVAC system in real time to obtain the operating status data of the target HVAC system. The operating status data includes the compressor high pressure, compressor low pressure, actual frequency of the inverter equipment, outdoor unit fan output speed, opening degree of the two expansion valves of the indoor unit, current indoor temperature of the smart building, and current energy efficiency ratio of the target HVAC system. The operating status data of the target HVAC system is then transmitted to the status prediction unit 10.

[0091] It should be noted that a compressor refers to the equipment in a heating, ventilation, and air conditioning (HVAC) system responsible for compressing and pumping refrigerant; high pressure and low pressure refer to two key parameters of the compressor in the refrigeration cycle. The pressure values ​​at the high-pressure end and low-pressure end directly affect the compression and evaporation process of the refrigerant, and are crucial to the system efficiency and performance; a variable frequency drive (VFD) refers to a device that can adjust the output frequency; an expansion valve refers to a device used to control the flow and pressure of refrigerant, located in the indoor part of the HVAC system; the opening degree of the expansion valve refers to the degree to which the expansion valve is open, affecting the flow of refrigerant and the cooling effect of the system.

[0092] In an optional embodiment, the state prediction unit 10 receives the target HVAC operating status data transmitted by the air conditioning control unit 9, and inputs the target HVAC operating status data into the trained state prediction model. The trained state prediction model outputs the target HVAC operating status data for the next preset time. The state prediction model adopts a neural network model and transmits the target HVAC operating status data for the next preset time to the energy-saving control unit 11.

[0093] It should be noted that state prediction models include, but are not limited to, recurrent neural networks, long short-term memory networks, convolutional neural networks, and deep neural networks.

[0094] In an optional embodiment, the energy-saving control unit 11 receives the indoor temperature and energy efficiency ratio of the target HVAC system at the next preset time transmitted by the state prediction unit 10, and uses reinforcement learning to construct a first intelligent agent of the target HVAC outdoor unit and a second intelligent agent of the target HVAC indoor unit based on the indoor temperature and energy efficiency ratio of the target HVAC system at the next preset time. The first intelligent agent and the second intelligent agent are solved to obtain the second control parameters of the target HVAC system, and the target HVAC system is controlled based on the second control parameters.

[0095] It should be noted that reinforcement learning is a machine learning method used to train agents to learn optimal behavior in interactions with the environment. In reinforcement learning, the first agent and the second agent refer to the learner or controller. Here, they refer to two different agents that control the outdoor unit and the indoor unit, respectively. They learn to improve their control strategies to optimize the performance of the overall HVAC system.

[0096] Example 2: The energy-saving control system for HVAC in smart buildings based on artificial intelligence proposed in this invention, compared with Example 1, further includes the following: the air outlet model of the target HVAC fan is as follows:

[0097]

[0098] Where, m a Q represents the actual airflow rate delivered by the fan. S Indicates the indoor cooling load of a smart building, T N Indicates the set indoor temperature of a smart building, T S Indicates the supply air temperature;

[0099] The target HVAC fan energy consumption model is as follows:

[0100]

[0101] Among them, P fan The actual power of the wind turbine is represented by α, a1, b1, c1, d1, and e1 represent the wind turbine characteristic fitting coefficients, and α represents the actual power of the wind turbine. fan ρ represents the overall efficiency of the fan. air β represents air density. a This indicates the main load rate of the wind turbine, and m a-L This indicates the rated airflow of the fan;

[0102] The chilled water pump chilling model for the target HVAC system is as follows:

[0103]

[0104] Where, m e The temperature represents the actual chilled water flow rate of the chilled water pump, T1 represents the inlet air temperature of the surface cooler, and T2 represents the outlet air temperature of the surface cooler. p ΔT represents the specific heat of water. e Indicates the temperature difference between chilled water supply and return;

[0105] The energy consumption model for the chilled water pumps of the target HVAC system is as follows:

[0106]

[0107] Among them, P eThe value represents the actual power of the chilled water pump, and a2, b2, c2, and d2 represent the characteristic fitting coefficients of the chilled water pump. β e This indicates the main load rate of the chilled water pump, and m e-L P represents the rated chilled water flow rate of the chilled water pump. e-L This indicates the rated power of the chilled water pump.

[0108] In an optional embodiment, the performance model of the chiller for the target HVAC system is as follows:

[0109]

[0110] Among them, Q e Indicates the cooling capacity of the refrigeration system, η ch1 The curve representing the relationship between cooling capacity and temperature, η ch2 The curve representing the reciprocal of the energy efficiency ratio as a function of chilled water outlet temperature and cooling water return temperature, η ch3 The curve represents the relationship between the reciprocal of the energy efficiency ratio and the main load rate of the chiller. a3, b3, c3, d3, e3, and f3 represent the fitting coefficients for the cooling capacity characteristics; a4, b4, c4, d4, e4, and f4 represent the fitting coefficients for the chilled water characteristics; a5, b5, and c5 represent the fitting coefficients for the main load rate characteristics of the chiller; T eo T represents the chilled water outlet temperature of the chiller. ci Indicates the cooling water return temperature, β ch This indicates the main load rate of the refrigeration unit, and Q avail Indicates the available cooling capacity of the refrigeration unit;

[0111] The energy consumption model for the target HVAC system's chiller is as follows:

[0112]

[0113] Among them, P ch C represents the actual power of the refrigeration unit. COP-L This represents the rated energy efficiency ratio of the refrigeration unit, and Q... avail =Q N η ch1 Q N This indicates the rated cooling capacity of the refrigeration unit;

[0114] The energy consumption optimization model is as follows:

[0115]

[0116] Where min P represents the objective function, st represents the constraints, and m a-min and m a-max T represents the minimum and maximum actual airflow rates of the fan, respectively.S-min and T S-max These represent the minimum and maximum values ​​of the supply air temperature, respectively, m. e-min and m e-max ΔT represents the minimum and maximum actual chilled water flow rates of the chilled water pump, respectively. e-min and ΔT e-max T represents the minimum and maximum temperature difference between the chilled water supply and return water, respectively. eo-min and T eo-max T represents the minimum and maximum outlet water temperature of the chiller, respectively. ci-min and T ci-max These represent the minimum and maximum values ​​of the cooling water return temperature, respectively.

[0117] In an optional embodiment, the energy consumption optimization model is solved using an optimization algorithm to obtain the first control parameters of the target HVAC system, including the following steps:

[0118] A1. Set the percentage of discoverers (PD), the percentage of scouts (SD), and the alert value (R2), and initialize the population. The population initialization formula is as follows:

[0119]

[0120] Where, x i and x i+1 Let i represent the i-th and (i+1)-th individuals in the population, and mod represents the remainder function;

[0121] A2. Calculate the fitness value of each individual in the population and sort them according to their fitness values. Select the discoverer from among the individuals whose fitness values ​​are greater than a preset fitness threshold, and update the position of the discoverer. The formula for updating the position of the discoverer is as follows:

[0122]

[0123] in, and Let r1 represent the position of the discoverer with coordinates (i,j) in the population during the t-th and t+1-th iterations in the d-dimensional solution space, and let r2 represent the random number that determines the distance the discoverer moves. Let y1 and y2 represent the optimal position of the current discoverer in the t-th iteration, and let ST represent the safety threshold.

[0124] A3. After selecting the discoverer from the population, the remaining individuals in the population are considered as new members, and the positions of the new members are updated. The formula for updating the positions of new members is as follows:

[0125]

[0126] in, and Let represent the position of the joiner with coordinates (i,j) in the d-dimensional solution space during the t-th and t+1-th iterations. This represents the worst-case global position of the current joiner in the t-th iteration. Let A represent the optimal position of the current joiner in the (t+1)th iteration, let L represent the identity matrix, and let n represent the total number of individuals in the population.

[0127] A4. Randomly select scouts from the population and update their positions. The formula for updating scout positions is as follows:

[0128]

[0129] in, and Let represent the position of the scout with coordinates (i,j) in the population during the t-th and t+1-th iterations in the d-dimensional solution space. This represents the worst-case global position of the current scout in the t-th iteration. Let f represent the optimal position of the scout in the (t+1)th iteration, β represent a random number that follows a Gaussian distribution with a mean of 0 and a variance of 1, γ be 10E⁻⁸, K represent the random coefficients for direction and step size, and f i f represents the fitness value of the investigator. g and f worst This represents the scout's best and worst fitness values;

[0130] A5. Perturb the current optimal solution, recalculate the population fitness, and find the optimal solution after perturbation. The perturbation formula for the current optimal solution is as follows:

[0131]

[0132] in, X represents the position of the individual with coordinates (i,j) in the population in the (t+1)th iteration in the d-dimensional solution space. best (t) represents the current optimal solution, and cauchy(0,1) represents the standard Cauchy distribution function;

[0133] A6. Determine if the termination condition has been met. If the termination condition has been met, output the optimal individual position and its fitness value. Otherwise, repeat step A2.

[0134] In an optional embodiment, the first intelligent agent of the target HVAC outdoor unit includes a first state, a first action, and a first reward function, the expression of which is as follows:

[0135] S0 = [T]set ,T out ,EER,T in ];

[0136] Where S0 represents the first state, T set This indicates the set temperature of the indoor unit, T. out The outdoor temperature is represented by EER, and the energy efficiency ratio for the next preset time is represented by T. in Indicates the indoor temperature at the next preset time;

[0137] The first action expression is as follows:

[0138] A0 = [f com ,PD high ,PD low ,f fan ];

[0139] Where A0 represents the first action, f com PD represents the actual frequency of the frequency converter. high PD indicates the high pressure of the compressor. low f represents the low-pressure value of the compressor. fan Indicates the output fan speed of the outdoor unit;

[0140] The expression for the first reward function is as follows:

[0141] R0 = EER;

[0142] Where R0 represents the first reward function;

[0143] The second agent of the target HVAC indoor unit includes a second state, a second action, and a second reward function. The expression for the second state is as follows:

[0144]

[0145] Among them, S i Indicates the second state. This represents the set temperature of the i-th indoor unit. This indicates the indoor temperature at the next preset time for the i-th indoor unit;

[0146] The second action expression is as follows:

[0147]

[0148] Among them, A i Indicates the second action. and This indicates the opening degree of the two expansion valves of the indoor unit;

[0149] The expression for the second reward function is as follows:

[0150]

[0151] Among them, R i This represents the second reward function.

[0152] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. An HVAC energy-saving control system based on artificial intelligence in a smart building, comprising a chiller modeling unit (3), a total energy consumption modeling unit (4), a target modeling unit (5), a constraint modeling unit (6), an energy consumption optimization modeling unit (7), a model solving unit (8), an air conditioning control unit (9), a state prediction unit (10), and an energy-saving control unit (11), characterized in that: a fan modeling unit (1) is configured to construct a fan air outlet model of a target HVAC based on indoor cooling load, indoor set temperature, and supply air temperature of the smart building, obtain supply air flow of the target HVAC based on the fan air outlet model of the target HVAC, construct a fan energy consumption model of the target HVAC based on the supply air flow of the target HVAC, obtain fan energy consumption of the target HVAC based on the fan energy consumption model of the target HVAC, transmit the supply air flow of the target HVAC to the chilled water pump modeling unit (2), and transmit the fan energy consumption of the target HVAC to the total energy consumption modeling unit (4); a chilled water pump modeling unit (2) is configured to receive the supply air flow of the target HVAC transmitted by the fan modeling unit (1), and construct a chilled water pump cooling model of the target HVAC based on the supply air temperature, the cooling coil inlet air temperature, the chilled water supply and return temperature difference, and the supply air flow of the target HVAC, obtain chilled water flow of the target HVAC based on the chilled water pump cooling model of the target HVAC, construct a chilled water pump energy consumption model of the target HVAC based on the chilled water flow of the target HVAC, obtain chilled water pump energy consumption of the target HVAC based on the chilled water pump energy consumption model of the target HVAC, transmit the chilled water flow of the target HVAC to the chiller modeling unit (3), and transmit the chilled water pump energy consumption of the target HVAC to the total energy consumption modeling unit (4); the chiller modeling unit (3) is configured to receive the chilled water flow of the target HVAC transmitted by the chilled water pump modeling unit (2), construct a chiller performance model of the target HVAC based on chilled water outlet temperature, cooling water return temperature, the chilled water supply and return temperature difference, and the chilled water flow of the target HVAC, obtain a chiller performance curve of the target HVAC based on the chiller performance model of the target HVAC, construct a chiller energy consumption model of the target HVAC based on the chiller performance curve of the target HVAC, obtain chiller energy consumption of the target HVAC based on the chiller energy consumption model of the target HVAC, and transmit the chiller energy consumption of the target HVAC to the total energy consumption modeling unit (4). The total energy consumption modeling unit (4) receives the total power of the target HVAC transmitted by the total energy consumption modeling unit (4), and is configured to construct a total power model of the target HVAC based on the total power of the target HVAC, obtain the total power of the target HVAC based on the total power model of the target HVAC, and transmit the total power of the target HVAC to the target modeling unit (5); The target modeling unit (5) receives the total power of the target HVAC transmitted by the total energy consumption modeling unit (4), and is configured to construct a target optimization function with the target of minimizing the total power of the target HVAC, and transmit the target optimization function to the energy consumption optimization modeling unit (7); The constraint modeling unit (6) is configured to construct constraint conditions of the target HVAC based on the operating parameters of the target HVAC, and transmit the constraint conditions of the target HVAC to the energy consumption optimization modeling unit (7); The energy consumption optimization modeling unit (7) receives the target optimization function transmitted by the target modeling unit (5) and the constraint conditions of the target HVAC transmitted by the constraint modeling unit (6), and is configured to construct an energy consumption optimization model based on the target optimization function and the constraint conditions of the target HVAC, and transmit the energy consumption optimization model to the model solving unit (8); The model solving unit (8) receives the energy consumption optimization model transmitted by the energy consumption optimization modeling unit (7), and is configured to solve the energy consumption optimization model by an optimization algorithm to obtain first control parameters of the target HVAC, and transmit the first control parameters of the target HVAC to the air conditioner control unit (9); The air conditioner control unit (9) receives the first control parameters of the target HVAC transmitted by the model solving unit (8), and is configured to control the target HVAC based on the first control parameters of the target HVAC, and monitor the operating state data of the target HVAC in real time to obtain the operating state data of the target HVAC, wherein the operating state data includes compressor high-pressure pressure, compressor low-pressure pressure, actual frequency of the frequency conversion device, output air speed of the outdoor fan, opening degree of two expansion valves of the indoor unit, current indoor temperature of the intelligent building, and energy efficiency ratio of the current target HVAC, and transmit the operating state data of the target HVAC to the state prediction unit (10); The state prediction unit (10) receives the operation state data of the target HVAC transmitted by the HVAC control unit (9), and inputs the operation state data of the target HVAC into a trained state prediction model, and outputs the operation state data of the target HVAC at the next preset time through the trained state prediction model, wherein the state prediction model adopts a neural network model, and the operation state data of the target HVAC at the next preset time is transmitted to the energy-saving control unit (11); The energy-saving control unit (11) receives the indoor temperature and the energy efficiency ratio of the target HVAC at the next preset time transmitted by the state prediction unit (10), and constructs a first agent of the outdoor unit of the target HVAC and a second agent of the indoor unit of the target HVAC based on the indoor temperature and the energy efficiency ratio of the target HVAC at the next preset time through reinforcement learning, solves the first agent and the second agent to obtain the second control parameter of the target HVAC, and controls the target HVAC based on the second control parameter of the target HVAC; The first agent of the outdoor unit of the target HVAC includes a first state, a first action and a first reward function, and the first state expression is as follows: ； wherein, represents a first state, represents a set temperature of the indoor unit, represents an outdoor temperature, represents an energy efficiency ratio at a next preset time, represents an indoor temperature at a next preset time; The first action expression is as follows: ； wherein, represents a first action, represents an actual frequency of the variable frequency device, represents a high pressure of the compressor, represents a low pressure of the compressor, represents an outdoor fan output air speed; The first reward function expression is as follows: ； wherein, represents the first reward function; The second agent of the indoor unit of the target HVAC includes a second state, a second action and a second reward function, and the second state expression is as follows: ； wherein, represents the second state, represents the set temperature of the i-th indoor unit, represents the indoor temperature of the i-th indoor unit at the next preset time. The second action expression is as follows: ； wherein, represents a second action, and represents two expansion valve openings of the indoor unit; The second reward function expression is as follows: ； wherein, represents a second reward function; The energy consumption optimization model is solved by an optimization algorithm to obtain the first control parameter of the target HVAC, including the following steps: A1, set the proportion of discoverers , the proportion of investigators and the alert value , and initialize the population, the population initialization formula is as follows: ； wherein and denote the and the individuals in the population, denotes the modulo function; A2, calculate the fitness value of each individual in the population, and sort according to the fitness value, select the discoverer from the individual whose fitness value is greater than the preset fitness value threshold, and update the position of the discoverer, the position update formula of the discoverer is as follows: ； wherein, and denotes the position of the discoverer in the d-dimensional solution space in the th and th iteration of the population, denotes the position of the discoverer in the d-dimensional solution space in the denotes a random number that determines the distance of the discoverer's movement, denotes a random number that determines the direction of the discoverer's movement, denotes the current discoverer's best position of the th iteration, and denotes a split factor, denotes a security threshold; A3, after selecting the discoverer from the population, the remaining individuals in the population are used as joiners, and the position of the joiner is updated, the joiner position update formula is as follows: ； wherein, and denotes the position of the i-th subscriber in the d-dimensional solution space at the j-th iteration, denotes the current subscriber global worst position at the j-th iteration, denotes the current subscriber best position at the j-th iteration, denotes a matrix of random numbers 1 or -1, denotes the identity matrix, denotes the total number of individuals in the population;​​​​​ A4, a scout is randomly selected from the population, and the position of the scout is updated, the scout position update formula is as follows: ； wherein, and denotes the position of the investigator in the d-dimensional solution space at iteration and iteration , , denotes the current investigator global worst position at iteration , denotes the current investigator best position at iteration , denotes a random number, which follows a Gaussian distribution with an expectation of 0 and a variance of 1, is taken as 10E-8, denotes a random coefficient for direction and step length, denotes the fitness value of the investigator, and denotes the best and worst fitness values of the investigator; A5, disturb the current optimal solution, recalculate the population fitness, find the optimal solution after disturbance, and the current optimal solution disturbance formula is as follows: ； wherein, denotes the position of an individual in the d-dimensional solution space at iteration denotes the current best solution, denotes the standard Cauchy distribution function;​​ A6, judge whether the termination condition is reached, if the termination condition is reached, output the optimal individual position and its fitness value, otherwise, repeat step A2.

2. The artificial intelligence-based intelligent building HVAC energy consumption saving control system according to claim 1, characterized in that: The fan outflow model of the target HVAC is as follows: ； wherein, represents the actual supply air flow rate of the fan, represents the indoor cooling load of the smart building, represents the indoor set temperature of the smart building, represents the supply air temperature; The fan energy consumption model of the target HVAC is as follows: ； wherein, represents the actual power of the fan, , , , and represents the fan characteristic fitting coefficient, represents the total efficiency of the fan, represents the air density, represents the main load rate of the fan, and , represents the rated supply air flow of the fan; The refrigeration water pump refrigeration model of the target HVAC is as follows: ； wherein, represents the actual chilled water flow rate of the chilled water pump, represents the inlet air temperature of the surface cooler, represents the outlet air temperature of the surface cooler, represents the specific heat of water, represents the chilled water supply and return water temperature difference; The refrigeration water pump energy consumption model of the target HVAC is as follows: ； wherein, represents the actual power of the chilled water pump, , , and represents the chilled water pump characteristic fitting coefficient, represents the main load rate of the chilled water pump, and , represents the rated chilled water flow rate of the chilled water pump, represents the rated power of the chilled water pump. 3.The AI-based intelligent building HVAC energy consumption saving control system according to claim 2, characterized in that: The refrigeration machine performance model of the target HVAC is as follows: ； wherein, represents a refrigeration capacity of the chiller, represents a curve showing a relationship of the refrigeration capacity with respect to temperature, represents a curve showing a relationship of an inverse of the energy efficiency ratio with respect to chilled water outlet temperature and cooling water return temperature, represents a curve showing a relationship of an inverse of the energy efficiency ratio with respect to the chiller main load ratio, , , , , and represents a refrigeration capacity characteristic fitting coefficient, , , , , and represents a chilled water characteristic fitting coefficient, , and represents a chiller main load ratio characteristic fitting coefficient, represents a chiller chilled water outlet temperature, represents a cooling water return temperature, represents a chiller main load ratio, and , represents a chiller available cooling capacity; The refrigeration machine energy consumption model of the target HVAC is as follows: ； wherein, represents the actual power of the chiller, represents the rated energy efficiency ratio of the chiller, and , represents the rated refrigeration capacity of the chiller; The energy consumption optimization model is as follows: ； wherein, represents a target optimization function, represents a constraint condition, and respectively represent a minimum value and a maximum value of an actual supply air flow of a fan, and respectively represent a minimum value and a maximum value of a supply air temperature, and respectively represent a minimum value and a maximum value of an actual chilled water flow of a chilled water pump, and respectively represent a minimum value and a maximum value of a chilled water supply and return water temperature difference, and respectively represent a minimum value and a maximum value of a chilled water outlet temperature of a chiller, and respectively represent a minimum value and a maximum value of a cooling water return water temperature.

Images