An AI-based load control method for water-cooled central air conditioners

Abstract

The present invention discloses a load control method for water-cooled central air conditioners based on artificial intelligence, which relates to the technical field of load control. In the present invention, in the load prediction link of the chiller, a long short-term memory time series model and a physics-informed neural network are combined to perform load prediction using historical data, and the error is corrected through the energy balance equation of the chiller to improve the accuracy of the prediction; multi-objective optimization control is adopted, with maximizing the chiller COP, minimizing the energy consumption of the cooling tower fan, and the stability of the cooling water supply temperature as the optimization objectives, and combined with the prediction of environmental temperature and humidity, the cooling water supply temperature set point and the start-stop strategy of the cooling tower fan are dynamically calculated; based on the cooling water supply temperature set point and the fan start-stop strategy, the chiller, the cooling tower fan and the water pump are dynamically adjusted, and the high-frequency operation phenomenon is reduced through the optimization of the fan start-stop.

Description

Technical Field

[0001] The present invention relates to the technical field of load control, and in particular to a water-cooled central air-conditioning load control method based on artificial intelligence. Background Art

[0002] The water-cooled central air-conditioning system undertakes the key temperature control task, and its operating efficiency directly affects the overall energy consumption level; with the application of artificial intelligence technology, more and more central air-conditioning systems begin to adopt intelligent load control methods.

[0003] In scenarios with high temperature control requirements such as data centers and hospitals, the operating states of the chiller and the cooling tower are closely related. However, the current load control methods usually optimize them separately, lacking a coordinated scheduling strategy, resulting in insufficient energy efficiency optimization of the cooling water system; currently, most of the artificial intelligence load control methods focus on the optimal control of the chiller, such as using deep learning or time series prediction models to improve the load prediction accuracy and accordingly optimize the operating state of the chiller; however, the accuracy of load prediction does not directly equal the maximization of system energy efficiency; in actual operation, the optimal scheduling of the chiller often operates independently of the operating strategy of the cooling tower, and the cooling tower fan and pump still rely on fixed rules or simple temperature difference thresholds for control, making it difficult to adapt to the changing load demand of the chiller in a timely manner, resulting in unstable cooling water temperature regulation, which in turn affects the coefficient of performance (COP) of the chiller.

[0004] In the humid and hot southern regions or environments with large changes in external temperature and humidity, this problem is particularly prominent; the existing hierarchical fan control strategy only starts and stops based on a fixed supply and return water temperature difference, rather than dynamic optimization, resulting in frequent changes in the operation of the cooling tower fan, further increasing the burden on the chiller. At the same time, the coordinated operation among the pump, chiller, and cooling tower is not fully considered, and situations such as "frequent start and stop of the chiller and high-frequency operation of the cooling tower fan" are likely to occur, causing additional energy consumption losses; although some optimization schemes introduce an adaptive adjustment mechanism, such as adjusting the cooling water temperature set point in combination with the chiller load prediction model or using fuzzy control to optimize the operating strategy of the cooling tower fan, the non-linear coupling relationship between the chiller and the cooling tower has still not been completely solved, and the dynamic matching ability of the system under different load conditions is still limited.

[0005] It can be seen that although the existing artificial intelligence load control methods have partially improved the operating efficiency of water-cooled central air conditioners, due to the insufficient coordinated optimization of the chiller and the cooling tower, problems such as system response lag, energy efficiency loss, and frequent equipment start and stop still exist; therefore, on the basis of the current load control methods, further improving the coordinated optimization ability of the cooling tower and the chiller and constructing an intelligent control strategy based on global energy efficiency optimization will become the key direction for energy conservation and optimization of water-cooled central air conditioners. Summary of the Invention

[0006] In view of the above existing problems, the present invention is proposed.

[0007] The present invention provides a method for controlling the load of a water-cooled central air conditioner based on artificial intelligence to solve the problems that the traditional load control method fails to achieve the coordinated optimization of the chiller and the cooling tower, resulting in lag in the regulation of the cooling water temperature, reduction of COP, and serious energy efficiency loss.

[0008] To solve the above technical problems, the present invention provides the following technical solutions:

[0009] An embodiment of the present invention provides a method for controlling the load of a water-cooled central air conditioner based on artificial intelligence, which includes:

[0010] Step S1, obtaining the operation data of the water-cooled central air conditioner, cleaning, denoising and feature extraction of the operation data, and constructing an operation data set of the water-cooled central air conditioner;

[0011] Step S2, predicting the load demand of the chiller at a future time based on the operation data set of the water-cooled central air conditioner in step S1;

[0012] Step S3, calculating the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan according to the load prediction result of the chiller in step S2;

[0013] Step S4, dynamically adjusting the operation of the chiller, the cooling tower fan and the water pump based on the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan;

[0014] Step S5, based on the operation data of the water-cooled central air conditioner monitored in step S1, combined with the adjustment result in step S4, introducing a reinforcement learning algorithm to adaptively optimize the cooling water temperature control logic and the start-stop strategy of the cooling tower fan, and adjusting according to the actual operation effect.

[0015] As a preferred scheme of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, wherein: the operation data includes chiller load data, cooling water supply and return temperatures, ambient temperature and humidity, and building internal load demand.

[0016] As a preferred scheme of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, wherein: in step S2, a chiller load prediction model is constructed by using a long short-term memory (LSTM) time series model combined with a physics-informed neural network (PINN) to predict the load demand of the chiller at a future time.

[0017] As a preferred solution of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, wherein: the step of using the long short-term memory (LSTM) time series model combined with the physics-informed neural network (PINN) to construct a load prediction model for the chiller and predict the load demand of the chiller at a future time is as follows.

[0018] Construct an LSTM time series model, and the model formula is:

[0019] ,

[0020] ,

[0021] wherein, represents the LSTM hidden layer state at time step , represents the LSTM hidden layer state at the previous time step, represents the input feature vector at time step , and are respectively the weight matrices of the hidden layer state and the input layer, and are bias terms, is a non-linear activation function, is the load prediction output at time step , is the weight matrix of the output layer;

[0022] On the basis of the LSTM prediction, use the physics-informed neural network (PINN) to provide the physical constraint conditions of the chiller load, and the physical relationship is based on the energy balance equation of the chiller. The equation is:

[0023] ,

[0024] wherein, represents the refrigeration load of the chiller at time step , is the cold water flow rate, is the specific heat capacity of water, and respectively represent the inlet water temperature and outlet water temperature of the chiller at time step ;

[0025] Define the loss function combined with the PINN constraint, including:

[0026] ,

[0027] ,

[0028] ,

[0029] Among them, is the total loss function, is the LSTM prediction error loss, is the physical constraint error loss, is the weight parameter of the physical loss, is the predicted time series length, is the time step LSTM predicted load value of

[0030] The final optimization goal is:

[0031] ,

[0032] Among them, represents the set of trainable parameters of the LSTMPINN model.

[0033] As a preferred solution of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, wherein: in step S3, the multi-objective optimization control MPC method is used to construct an energy efficiency optimization model for the cooling water system.

[0034] As a preferred solution of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, wherein: in step S3, using the energy efficiency optimization model of the cooling water system, with the maximization of the COP of the chiller, the minimization of the energy consumption of the cooling tower fan, and the stability of the cooling water supply temperature as the optimization goals, combined with the predicted data of the environmental temperature and humidity, calculate the cooling water supply temperature set point and the start-stop strategy of the cooling tower fan.

[0035] As a preferred solution of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, wherein: the steps of using the multi-objective optimization control MPC method to construct an energy efficiency optimization model for the cooling water system are,

[0036] Establish an optimization control model for the cooling water system, and the energy efficiency optimization goal is defined as :

[0037] ,

[0038] Among them, represents the optimization objective function of the cooling water system, represents the performance coefficient of the chiller, represents the total energy consumption of the cooling tower fan, represents the stability index of the cooling water supply temperature,

[0039] is the weight coefficient,

[0040] The calculation formula for the coefficient of performance of the chiller is:

[0041] ,

[0042] where, represents the refrigeration load of the chiller, represents the input power of the chiller,

[0043] The calculation formula for the total energy consumption of the cooling tower fan is:

[0044] ,

[0045] where, represents the time step of the cooling tower fan power, represents the time step of the optimization time domain,

[0046] The calculation formula for the stability index of the cooling water supply temperature is:

[0047] ,

[0048] where, represents the time step of the cooling water supply temperature, represents the target water supply temperature set point,

[0049] The final optimization goal is defined as:

[0050] ,

[0051] The constraint conditions are defined as:

[0052] ,

[0053] ,

[0054] ,

[0055] where, represents the dynamic change function of the cooling water supply temperature, represents the time step of the ambient temperature, represents the maximum refrigeration load of the chiller, represents the maximum power of the cooling tower fan.

[0056] As a preferred solution of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, wherein: the steps of calculating the cooling water supply temperature set point and the start-stop strategy of the cooling tower fan are,

[0057] Under the MPC calculation framework, calculate the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan,

[0058] Calculate the set point of the cooling water supply temperature , and the calculation formula is:

[0059] ,

[0060] Combined with the predicted data of environmental temperature and humidity, use the model predictive control MPC method to update , and the update formula is:

[0061] ,

[0062] Among them, represents the learning rate parameter,

[0063] Define the operating state of the cooling tower fan :

[0064] If , then ,

[0065] If , then ,

[0066] Among them, represents the time step The start-stop state of the cooling tower fan, 1 means running, 0 means off, represents the start-stop power threshold of the fan,

[0067] Finally solve :

[0068] .

[0069] As a preferred solution of the water-cooled central air-conditioning load control method based on artificial intelligence described in the present invention, wherein: the steps of dynamically adjusting the operation of the chiller, cooling tower fan and water pump based on the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan are,

[0070] Based on the calculated set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan , dynamically adjust the chiller, cooling tower fan and water pump,

[0071] Based on the optimization goal, adjust the operating load of the chiller , and the adjustment formula is:

[0072] ,

[0073] Among them, represents the time step the refrigeration load of the chiller, represents the time step of the refrigeration load,

[0074] is the load adjustment gain parameter,

[0075] According to the start-stop strategy control the operating state of the fan,

[0076] If , then ,

[0077] If , then ,

[0078] Among them, represents the start-stop state of the cooling tower fan at time step , 1 means running, 0 means off, is the start-stop power threshold of the cooling tower fan,

[0079] Adjust the operating speed of the water pump according to the deviation of the cooling water supply temperature , and the adjustment formula is:

[0080] ,

[0081] Among them, represents the operating speed of the cooling water pump at time step , represents the operating speed of the cooling water pump at time step , is the pump adjustment gain coefficient.

[0082] As a preferred solution of the method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to the present invention, among them: the step of introducing a reinforcement learning algorithm to adaptively optimize the cooling water temperature control logic and the start-stop strategy of the cooling tower fan based on the operating data of the water-cooled central air conditioner monitored in step S1 and combining the adjustment results of step S4, and adjusting according to the actual operating effect is,

[0083] Define the state variable of the reinforcement learning model as:

[0084] ,

[0085] Among them, represents the system state at time step , represents the cooling water supply temperature at time step , Represents the time step of the cooling tower fan power, Represents the time step of the chiller refrigeration load, Represents the time step of the ambient temperature,

[0086] The reinforcement learning decision-making method is as follows:

[0087] ,

[0088] Among them, Represents the control action at the time step , including chiller load adjustment, cooling tower fan start / stop, and pump frequency conversion adjustment, Is the time step In the state Taking the action The reward function after that,

[0089] Define the reward function as :

[0090] ,

[0091] Among them, Represents the reward value after the reinforcement learning takes the action at the time step Taking the action After that, Respectively represent the loss values of the COP performance coefficient, fan energy consumption, and temperature stability target,

[0092] Are the corresponding weight parameters,

[0093] The reinforcement learning algorithm continuously optimizes the control strategy to maximize the long-term reward. The optimization formula is:

[0094] ,

[0095] Among them, Is the reinforcement learning strategy, Represents the expected value calculation, Is the reward discount factor, Is the time step of the optimization time domain.

[0096] The beneficial effects of the present invention are as follows: In the load prediction link of the chiller, the long short-term memory time series model and the physics-constrained neural network are combined to perform load prediction using historical data, and the error is corrected through the energy balance equation of the chiller to improve the prediction accuracy; multi-objective optimization control is adopted, maximizing the chiller COP, minimizing the energy consumption of the cooling tower fan, and the stability of the cooling water supply temperature are used as optimization objectives, and combined with the prediction of environmental temperature and humidity, the cooling water supply temperature set point and the start-stop strategy of the cooling tower fan are dynamically calculated.

[0097] Based on the cooling water supply temperature set point and the fan start-stop strategy, the present invention dynamically adjusts the chiller, the cooling tower fan and the water pump, and reduces the high-frequency operation phenomenon through the optimization of the fan start-stop; and introduces a reinforcement learning algorithm, continuously adjusts the cooling water temperature control logic and the fan start-stop strategy by combining the operation data and the optimization objectives, through reinforcement learning, accumulates experience in the long term, and continuously optimizes the control strategy according to the dynamic changes of the external environment and the load demand, enhances the adaptability to complex environments, and at the same time improves the overall energy efficiency and reduces unnecessary energy consumption.

[0098] In summary, the present invention can perform adaptive adjustment under different load conditions, effectively improve the cooperative control ability of the chiller and the cooling tower, reduce the ineffective operation time of the cooling tower fan, and reduce the additional energy consumption caused by the frequent start and stop of the chiller. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0100] Figure 1 It is a schematic flow chart of the load control method for water-cooled central air-conditioning based on artificial intelligence in this aspect. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0101] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings in the specification.

[0102] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.

[0103] Second, the "one embodiment" or "embodiment" referred to herein means a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0104] Embodiment 1, referring to Figure 1 , this embodiment provides an artificial intelligence-based load control method for water-cooled central air conditioners, including the following steps:

[0105] Step S1, obtain the operation data of the water-cooled central air conditioner, clean, denoise, and extract features from the operation data, and construct an operation data set for the water-cooled central air conditioner;

[0106] The operation data includes chiller load data, cooling water supply and return temperatures, ambient temperature and humidity, and building internal load demands;

[0107] Step S2, based on the operation data set of the water-cooled central air conditioner in step S1, predict the chiller load demand at a future time;

[0108] In step S2, use a long short-term memory (LSTM) time series model combined with a physics-informed neural network (PINN) to construct a chiller load prediction model to predict the chiller load demand at a future time;

[0109] The steps of using a long short-term memory (LSTM) time series model combined with a physics-informed neural network (PINN) to construct a chiller load prediction model to predict the chiller load demand at a future time are as follows:

[0110] Construct an LSTM time series model, and the model formula is:

[0111] ,

[0112] ,

[0113] where represents the LSTM hidden layer state at time step , represents the LSTM hidden layer state at the previous time step, represents the input feature vector at time step , and are the weight matrices of the hidden layer state and the input layer respectively, and are the bias terms, is the non-linear activation function, is the load prediction output at time step , is the weight matrix of the output layer;

[0114] Based on the LSTM prediction, the Physics-Informed Neural Network (PINN) is used to provide the physical constraint conditions for the chiller load. The physical relationship is based on the energy balance equation of the chiller, and the equation is:

[0115] ,

[0116] where, represents the time step of the chiller cooling load, is the chilled water flow rate, is the specific heat capacity of water, and respectively represent the inlet temperature and outlet temperature of the chiller at time step ;

[0117] Define the loss function by combining the PINN constraints, including:

[0118] ,

[0119] ,

[0120] ,

[0121] where, is the total loss function, is the LSTM prediction error loss, is the physical constraint error loss, is the weight parameter of the physical loss, is the length of the prediction time series, is the time step of the LSTM predicted load value;

[0122] The final optimization objective is:

[0123] ,

[0124] where, represents the set of trainable parameters of the LSTMPINN model.

[0125] Specifically, LSTM is used to process time series data to predict the chiller load, and capture long-term dependencies through the propagation of hidden states; PINN corrects the LSTM prediction error through the physical loss term, and then optimizes the LSTMPINN joint model by minimizing the loss function, so as to predict the chiller load at future moments;

[0126] Step S3: Calculate the set point of the cooling water supply temperature and the start / stop strategy of the cooling tower fan according to the predicted result of the chiller load in Step S2.

[0127] In Step S3, the multi-objective predictive control (MPC) method is used to establish an energy efficiency optimization model for the cooling water system.

[0128] In Step S3, with the energy efficiency optimization model of the cooling water system, taking the maximization of the COP of the chiller, the minimization of the energy consumption of the cooling tower fan, and the stability of the cooling water supply temperature as the optimization objectives, combined with the predicted data of the environmental temperature and humidity, calculate the set point of the cooling water supply temperature and the start / stop strategy of the cooling tower fan.

[0129] The steps of using the multi-objective predictive control (MPC) method to establish an energy efficiency optimization model for the cooling water system are as follows:

[0130] Establish an optimization control model for the cooling water system, and define the energy efficiency optimization objective as :

[0131] ,

[0132] where represents the optimization objective function of the cooling water system, represents the coefficient of performance of the chiller, represents the total energy consumption of the cooling tower fan, represents the stability index of the cooling water supply temperature,

[0133] are the weight coefficients,

[0134] The calculation formula for the coefficient of performance of the chiller is:

[0135] ,

[0136] where represents the refrigeration load of the chiller, represents the input power of the chiller,

[0137] The calculation formula for the total energy consumption of the cooling tower fan is:

[0138] ,

[0139] where represents the power of the cooling tower fan at time step , represents the time step of the optimization horizon,

[0140] The calculation formula for the stability index of the cooling water supply temperature is:

[0141] ,

[0142] Among them, represents the cooling water supply temperature at the time step of, represents the target water supply temperature set point,

[0143] The final optimization goal is defined as:

[0144] ,

[0145] The constraint conditions are defined as:

[0146] ,

[0147] ,

[0148] ,

[0149] Among them, represents the dynamic change function of the cooling water supply temperature, represents the time step of the ambient temperature, represents the maximum cooling load of the chiller, represents the maximum power of the cooling tower fan;

[0150] The steps to calculate the cooling water supply temperature set point and the start / stop strategy of the cooling tower fan are as follows,

[0151] Under the MPC calculation framework, calculate the cooling water supply temperature set point and the start / stop strategy of the cooling tower fan,

[0152] Calculate the cooling water supply temperature set point , and the calculation formula is:

[0153] ,

[0154] Combined with the ambient temperature and humidity prediction data, use the model predictive control MPC method to update , and the update formula is:

[0155] ,

[0156] Among them, represents the learning rate parameter,

[0157] Define the operating state of the cooling tower fan :

[0158] If , then ,

[0159] If , then ,

[0160] wherein, represents the time step the start / stop state of the cooling tower fan, 1 represents running, 0 represents shutdown, represents the start / stop power threshold of the fan,

[0161] Final solution :

[0162] ;

[0163] Specifically, based on the multi-objective optimization control MPC method here, an energy efficiency optimization model of the cooling water system is constructed. Specifically:

[0164] Taking the maximization of the chiller COP, the minimization of the cooling tower fan energy consumption, and the stability of the cooling water supply temperature as the optimization directions, balancing the influence of different objectives through weight parameters, combining the energy efficiency characteristics of the chiller, calculating the COP using the energy balance equation, constructing a fan energy consumption loss function; and establishing a cooling water temperature stability index, solving the optimization problem to calculate the cooling water supply temperature set point, and adjusting it in real time in combination with the ambient temperature and humidity data;

[0165] Based on the optimization objectives, determine the start / stop strategy of the cooling tower fan, taking the fan power consumption threshold as the criterion, calculate the start / stop state, and dynamically adjust the system operation through the optimal control strategy, improving the overall energy efficiency while ensuring the system stability and effectively reducing unnecessary fan energy consumption;

[0166] Step S4, based on the cooling water supply temperature set point and the cooling tower fan start / stop strategy, dynamically adjust the operation of the chiller, cooling tower fan and water pump;

[0167] The steps of dynamically adjusting the operation of the chiller, cooling tower fan and water pump based on the cooling water supply temperature set point and the cooling tower fan start / stop strategy are as follows

[0168] Based on the calculated cooling water supply temperature set point and the cooling tower fan start / stop strategy , dynamically adjust the chiller, cooling tower fan and water pump,

[0169] Based on the optimization objectives, adjust the operation load of the chiller , and the adjustment formula is:

[0170] ,

[0171] wherein, represents the time step the refrigeration load of the chiller, Indicates the time step of the cooling load,

[0172] is the load adjustment gain parameter,

[0173] According to the start-stop strategy control the operating state of the fan,

[0174] If , then ,

[0175] If , then ,

[0176] Among them, Indicates the start-stop state of the cooling tower fan at time step , 1 means running, 0 means off, is the start-stop power threshold of the cooling tower fan,

[0177] Adjust the operating speed of the water pump according to the deviation of the cooling water supply temperature , and the adjustment formula is:

[0178] ,

[0179] Among them, Indicates the time step The operating speed of the cooling water pump, Indicates the time step of the operating speed of the cooling water pump, is the pump adjustment gain coefficient;

[0180] Specifically, based on the cooling water supply temperature set point and the start-stop strategy of the cooling tower fan, dynamically adjust the operation of the chiller, cooling tower fan and water pump to achieve better energy-saving control;

[0181] Adjust the load response of the chiller to the deviation of the supply water temperature and the change of the fan power, and calculate the load adjustment amount of the chiller; According to the fan power consumption threshold and the optimal strategy, dynamically control the start-stop state of the fan to reduce the ineffective operation time of the fan; Finally, for the variable frequency control strategy of the water pump, adopt an incremental adjustment method based on the temperature deviation to make the operating speed of the water pump adaptively adjust with the change of the cooling water temperature; Improve energy utilization rate and reduce equipment operation cost;

[0182] Step S5, based on the operating data of the water-cooled central air conditioner monitored in step S1, combined with the adjustment results of step S4, introduce a reinforcement learning algorithm to adaptively optimize the cooling water temperature control logic and the start-stop strategy of the cooling tower fan, and adjust according to the actual operation effect;

[0183] Based on the operating data of the water-cooled central air conditioner monitored in step S1 and combined with the adjustment results of step S4, the step of introducing a reinforcement learning algorithm to adaptively optimize the cooling water temperature control logic and the start-stop strategy of the cooling tower fan and adjusting according to the actual operation effect is as follows:

[0184] Define the state variables of the reinforcement learning model as:

[0185] ,

[0186] Among them, represents the system state at time step represents the supply water temperature of the cooling water at time step represents the power of the cooling tower fan at time step represents the refrigeration load of the chiller at time step represents the ambient temperature at time step

[0187] The reinforcement learning decision-making method is:

[0188] ,

[0189] Among them, represents the control action at time step , including the adjustment of the chiller load, the start-stop of the cooling tower fan, and the variable frequency adjustment of the water pump, is the reward function after taking action at time step in state

[0190] Define the reward function as :

[0191] ,

[0192] Among them, represents the reward value after taking action at time step in the reinforcement learning, respectively represent the loss values of the COP performance coefficient, the fan energy consumption, and the temperature stability target,

[0193] are the corresponding weight parameters,

[0194] The reinforcement learning algorithm continuously optimizes the control strategy to maximize the long-term reward. The optimization formula is:

[0195] , ​​​​​​

[0196] Among them, is the reinforcement learning policy, represents the expected value calculation, is the reward discount factor, is the time step of the optimization time domain;

[0197] Specifically, a reinforcement learning algorithm is introduced to adaptively optimize the cooling water temperature control logic and the start-stop strategy of the cooling tower fan. Based on the reinforcement learning policy, key decision variables such as the set point of the cooling water supply temperature, the start-stop of the fan, and the regulation of the water pump are dynamically optimized, and the energy efficiency performance of different decisions is quantified through the reward function;

[0198] The reinforcement learning algorithm accumulates experience through long-term operation and continuously adjusts the control strategy, enabling the system to have stronger adaptability in complex environments, improving the operation efficiency, and reducing unnecessary energy consumption at the same time.

[0199] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based load control method for water-cooled central air conditioners, characterized in that: Including, Step S1: Obtain the operation data of the water-cooled central air conditioner, clean, denoise, and extract features from the operation data, and construct an operation dataset for the water-cooled central air conditioner; Step S2: Based on the operation dataset of the water-cooled central air conditioner in Step S1, predict the load demand of the chiller at future times; Step S3: According to the chiller load prediction result in Step S2, calculate the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan; Step S4: Based on the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan, dynamically adjust the operation of the chiller, cooling tower fan, and water pump; Step S5: Based on the operation data of the water-cooled central air conditioner monitored in Step S1, combined with the adjustment result in Step S4, introduce a reinforcement learning algorithm to adaptively optimize the cooling water temperature control logic and the start-stop strategy of the cooling tower fan, and make adjustments according to the actual operation effect; In Step S3, adopt the multi-objective optimization control MPC method to construct an energy efficiency optimization model for the cooling water system, including: Establish an optimized control model for the cooling water system, and define the energy efficiency optimization goal as : ， Among them, represents the optimization objective function of the cooling water system, represents the performance coefficient of the chiller, represents the total energy consumption of the cooling tower fan, represents the stability index of the cooling water supply temperature, is the weight coefficient, The calculation formula for the coefficient of performance of the chiller is: ， Among them, represents the refrigeration load of the chiller, represents the input power of the chiller, The calculation formula for the total energy consumption of the cooling tower fan is: ， Among them, represents the cooling tower fan power at the time step , and represents the time step of the optimization time domain. The calculation formula for the stability index of the cooling water supply temperature is: ， Among them, represents the cooling water supply temperature at the time step , and represents the target supply temperature set point The final optimization objective is defined as: ， The constraint conditions are defined as: ， ， ， Among them, represents the dynamic change function of the cooling water supply temperature, represents the time step of the ambient temperature, represents the maximum refrigeration load of the chiller, represents the maximum power of the cooling tower fan; The step of calculating the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan is Under the MPC calculation framework, calculate the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan, Calculating the set point of the cooling water supply temperature , and the calculation formula is as follows: ， Update using the model predictive control (MPC) method in combination with the predicted ambient temperature and humidity data , and the update formula is: ， Among them, represents the learning rate parameter, Define time step Start / stop status of the cooling tower fan : If , then , If , then , Among them, represents the time step the start / stop status of the cooling tower fan, 1 indicates running, 0 indicates shutdown, represents the start / stop power threshold of the fan Final solution : 。 2. The water-cooled central air-conditioning load control method based on artificial intelligence according to claim 1, wherein: The operation data includes chiller load data, cooling water supply and return temperatures, ambient temperature and humidity, and building internal load demand.

3. The method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to claim 2, wherein: In Step S2, use the long short-term memory LSTM time series model combined with the physics-informed neural network PINN to construct a chiller load prediction model to predict the load demand of the chiller at future times.

4. The method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to claim 3, wherein: The step of using the long short-term memory LSTM time series model combined with the physics-informed neural network PINN to construct a chiller load prediction model to predict the load demand of the chiller at future times is Construct an LSTM time series model, and the model formula is: ， ， Among them, represents the LSTM hidden layer state at time step , represents the LSTM hidden layer state at the previous time step, represents the input feature vector at time step , and are the weight matrices of the hidden layer state and the input layer respectively, and are the bias terms, is the non-linear activation function, is the load prediction output at time step , is the output layer weight matrix; Based on the LSTM prediction, use the physics-informed neural network PINN to provide the physical constraint conditions of the chiller load. The physical relationship is based on the energy balance equation of the chiller, and the equation is: ， Among them, represents the refrigeration load of the chiller at the time step . is the chilled water flow rate, is the specific heat capacity of water, and respectively represent the inlet temperature and outlet temperature of the chiller at the time step . Define the loss function combined with the PINN constraint, including: ， ， ， Among them, is the total loss function, is the LSTM prediction error loss, is the physical constraint error loss, is the weight parameter of the physical loss, is the predicted time series length, is the time step is the LSTM predicted load value at The final optimization objective is: ， Among them, represents the set of trainable parameters of the LSTMPINN model.

5. The method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to claim 4, characterized in that: In Step S3, adopt the energy efficiency optimization model of the cooling water system, with the maximization of the COP of the chiller, the minimization of the energy consumption of the cooling tower fan, and the stability of the cooling water supply temperature as the optimization objectives, and combine the predicted data of the ambient temperature and humidity to calculate the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan.

6. The method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to claim 5, characterized in that: The step of dynamically adjusting the operation of the chiller, cooling tower fan, and water pump based on the set point of the cooling water supply temperature and the start-stop strategy of the cooling tower fan is At the calculated cooling water supply temperature set point and time step Based on the start / stop status of the cooling tower fan dynamic adjustment is performed on the chiller, cooling tower fan, and water pump Adjust the operating load of the chiller based on the optimization objective , and the adjustment formula is: ， Among them, represents the time step the refrigeration load of the chiller, represents the time step of the refrigeration load is the load regulation gain parameter, According to the time step Start / stop status of the cooling tower fan Control the operating status of the fan If , then , If , then , Among them, represents the start / stop state of the cooling tower fan at time step , where 1 indicates running and 0 indicates shutdown, is the start / stop power threshold of the cooling tower fan, Adjust the operating speed of the water pump according to the deviation of the cooling water supply temperature , and the adjustment formula is: ， Among them, represents the time step operating speed of the cooling water pump, represents the time step operating speed of the cooling water pump at that time step, is the pump adjustment gain coefficient.

7. The method for controlling the load of a water-cooled central air conditioner based on artificial intelligence according to claim 6, characterized in that: The step of introducing a reinforcement learning algorithm to adaptively optimize the cooling water temperature control logic and the start-stop strategy of the cooling tower fan based on the operation data of the water-cooled central air conditioner monitored in Step S1, combined with the adjustment result in Step S4, and making adjustments according to the actual operation effect is Define the state variables of the reinforcement learning model as: ， Among them, represents the system state at time step . represents the cooling water supply temperature at time step . represents the cooling tower fan power at time step . represents the refrigeration load of the chiller at time step . represents the ambient temperature at time step . The reinforcement learning decision-making method is: ， Among them, represents the control actions at the time step , including the load adjustment of the chiller, the start / stop of the cooling tower fan, and the variable frequency adjustment of the pump. is the time step at the state when the action is taken, and it is the reward function. Define the reward function as :[[]]END]] ， Among them, represents the reward value obtained by the reinforcement learning at time step after taking action ; represents the coefficient of performance of the chiller, represents the total energy consumption of the cooling tower fan, represents the stability index of the cooling water supply temperature, is the weight coefficient. The reinforcement learning algorithm continuously optimizes the control strategy to maximize the long-term reward, and the optimization formula is: ， Among them, is the reinforcement learning policy, represents the expected value calculation, is the reward discount factor, is the time step of the optimization time domain.

Images