A heat exchange station circulating pump characteristic adaptive optimization control method and system

Abstract

The present invention discloses a method and system for adaptive optimization control of the characteristics of a heat exchange station circulating pump, relating to the technical field of circulating pump control. The method comprises: establishing a building terminal heat transfer model and a building energy balance model based on the temperature distribution of the radiator surface, and then establishing a building load prediction model; numerically modeling the performance curve of the heat exchange station circulating pump to establish a circulating pump performance model; establishing a neural network control model for adaptive control of the heat exchange station circulating pump based on a deep deterministic policy gradient algorithm; establishing the state, action, and reward functions of the intelligent agent of the neural network control model; training the neural network control model to obtain a trained control model, and performing real-time control of the heat exchange station circulating pump based on the trained control model. The present invention alleviates the technical problems of low operating efficiency and severe energy waste in traditional control methods for circulating pumps.

Description

Technical Field

[0001] The present invention relates to the technical field of circulating pump control, and in particular to a method and system for adaptively optimizing the characteristics of a circulating pump in a heat exchange station. Background Art

[0002] In a centralized heating system, the heat exchange station serves as a key node connecting the primary heating network with the user end. Its operation and control directly impact the energy efficiency and operational stability of the heating system. The secondary-side circulation pump, the primary power source within the heat exchange station, is responsible for circulating hot water between user buildings to meet the heat load requirements of each building's terminal radiators or floor heating systems. Adjusting its operating frequency not only affects the comfort of the building's room temperature but also directly determines the system's energy consumption and operational efficiency.

[0003] In current engineering practice, the control method for secondary-side circulation pumps is mainly fixed-frequency operation or simple PID control. Some systems perform rough adjustments based on the secondary-side supply and return water temperature difference or water supply pressure. This traditional control method has the advantages of simple implementation and rapid response, but it generally has the following limitations: lack of perception of dynamic changes in the system's thermal load, unable to accurately adjust the pump speed according to the actual needs of the terminal, prone to "oversupply" or "undersupply", affecting heating comfort; pump frequency setting relies on manual experience or on-site trial adjustments, lacks a real-time adaptive mechanism, and the adjustment strategy is fixed, making it difficult to cope with the complex and changing characteristics of the system's operating state; the control target is single, usually focusing only on the water supply temperature or pressure difference, and fails to incorporate the pump's energy efficiency characteristics into the optimization target, resulting in the circulation pump operating in an inefficient state for a long time, resulting in serious energy waste; traditional control cannot uniformly take into account the two goals of "load matching" and "optimal energy efficiency", and it is difficult to achieve a balance between system regulation and energy conservation. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for adaptively optimizing the characteristics of a heat exchange station circulation pump in order to solve at least one of the above technical problems.

[0005] In a first aspect, an embodiment of the present invention provides a method for adaptive optimization control of the characteristics of a heat exchange station circulation pump, comprising: establishing a building terminal heat transfer model and a building energy balance model based on the temperature distribution on the radiator surface; establishing a building load prediction model based on the building terminal heat transfer model and the building energy balance model; numerically modeling the performance curve of the heat exchange station circulation pump to establish a circulation pump performance model; establishing a neural network control model for adaptive control of the heat exchange station circulation pump based on a deep deterministic policy gradient algorithm; establishing the state, action and reward function of the intelligent agent of the neural network control model based on the building load prediction model and the circulation pump performance model; training the neural network control model based on the state, action and reward function of the intelligent agent to obtain a trained control model, and performing real-time control of the heat exchange station circulation pump based on the trained control model.

[0006] Furthermore, the building terminal heat transfer model includes: ;

[0007] Where, is the temperature distribution of the radiator in the vertical direction, For time, is the vertical position of the radiator, is the height of the radiator, is the volume of the radiator, for The volume flow rate of water flowing through the radiator at any moment, is the heat transfer coefficient of the radiator, is the heat transfer area of ​​the radiator, is the specific heat capacity of water, is the density of water, is the indoor temperature of the building room.

[0008] Furthermore, the building energy balance model includes an exterior wall energy balance equation, an interior wall energy balance equation, and an indoor air energy balance equation; wherein,

[0009] The energy balance equation of the exterior wall includes: ;

[0010] The energy balance equation of the interior wall includes: ;

[0011] The indoor air energy balance equation includes: ;

[0012] Where, is the specific heat capacity of the exterior wall, For the quality of the exterior wall, for The temperature of the exterior wall at all times, is the exterior wall area, is the convection heat transfer coefficient between the inner surface of the exterior wall and the indoor air, for Indoor temperature at all times, is the convective heat transfer coefficient between the exterior wall surface and the outdoor air, for Outdoor temperature at any time, is the absorption rate of solar radiation by the exterior wall surface, for Solar radiation intensity at any moment, is the specific heat capacity of the inner wall, For the quality of the interior wall, for The temperature of the inner wall at all times, is the interior wall area, is the specific heat capacity of indoor air, For indoor air quality, is the heat transfer coefficient of the exterior window, The area of ​​the exterior windows.

[0013] Furthermore, the building load prediction model includes: ;

[0014] Where, for The predicted heat load of the building at any given moment, for The actual heat load of the building at any moment, for The indoor temperature of the building at all times, for Outdoor temperature at any time, for Solar radiation intensity at any moment, for The relative humidity outside at any time, is the model coefficient to be estimated, is the residual term;

[0015] The actual heat load calculation formula for the building includes: ;

[0016] Where, for The actual heat load of the building at any moment, is the specific heat capacity of water, is the density of water, is the circulation pump flow rate, for Secondary side water supply temperature at this moment, for Secondary side return water temperature at the moment;

[0017] The model coefficients to be estimated are determined through training with historical operating data.

[0018] Furthermore, the circulating pump performance model includes: ;

[0019] Where, is the circulation pump efficiency, is the circulation pump flow rate, is the circulating pump head, is the circulation pump head coefficient, is the effective power of the circulation pump, is the density of water, is the acceleration due to gravity, is the shaft power of the circulation pump, is the motor efficiency of the circulation pump, is the electrical power of the circulation pump, It is the ratio of the operating frequency of the circulating pump to the rated frequency.

[0020] Furthermore, the neural network control model includes a policy network and a value network; wherein the policy network includes two layers of fully connected networks and one layer of rectified linear unit activation network; and the value network includes a value function network.

[0021] Furthermore, the state of the agent of the neural network control model includes: ;

[0022] Where, For the agent The state vector at the moment, for Circulation pump flow at all times, for The pressure difference between the inlet and outlet of the circulating pump at all times, for Circulation pump frequency at all times, for The electrical power of the circulating pump at all times, for Secondary side water supply temperature at this moment, for Secondary side return water temperature at time for The predicted heat load of the building at any given moment;

[0023] The actions of the agent of the neural network control model include: ;

[0024] Where, for The adjustment increment of the circulating pump frequency at all times, is the maximum frequency change step allowed by the action, is the frequency change of the circulating pump.

[0025] The reward function of the agent of the neural network control model includes: ;

[0026] Where, for The moment reward function, for The actual heat load of the building at any moment, for The electrical power of the circulating pump at all times, is the weighting coefficient.

[0027] In a second aspect, an embodiment of the present invention further provides a heat exchange station circulating pump characteristic adaptive optimization control system, comprising: a first establishment module, a second establishment module, a third establishment module, a fourth establishment module, a fifth establishment module and a control module; wherein the first establishment module is used to establish a building terminal heat transfer model and a building energy balance model based on the temperature distribution of the radiator surface; the second establishment module is used to establish a building load prediction model based on the building terminal heat transfer model and the building energy balance model; the third establishment module is used to numerically model the performance curve of the heat exchange station circulating pump and establish a circulating pump performance model; the fourth establishment module is used to establish a neural network control model for adaptive control of the heat exchange station circulating pump based on a deep deterministic policy gradient algorithm; the fifth establishment module is used to establish the state, action and reward function of the intelligent agent of the neural network control model based on the building load prediction model and the circulating pump performance model; the control module is used to train the neural network control model based on the state, action and reward function of the intelligent agent to obtain a trained control model, and perform real-time control of the heat exchange station circulating pump based on the trained control model.

[0028] In a third aspect, an embodiment of the present invention further provides an electronic device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the method provided in the embodiment of the present invention when executing the computer program.

[0029] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method provided in the embodiment of the present invention is implemented.

[0030] The present invention provides a method and system for adaptive optimization control of the characteristics of a heat exchange station circulation pump. By introducing a reinforcement learning agent, it can perceive the system operating status in real time, dynamically adjust the operating frequency of the circulation pump, and make the operating point of the circulation pump continuously approach the high-efficiency operating range, thereby significantly reducing the energy consumption of the circulation pump while ensuring the stable supply of the building's heat load. Compared with traditional fixed logic or PID control methods, the present invention has self-learning and self-adaptation capabilities, can cope with changes in complex working conditions such as load fluctuations and system aging, and improve the robustness and intelligence level of the system operation. At the same time, the present invention reduces the dependence on empirical parameter debugging and manual intervention by constructing an end-to-end intelligent control strategy, which helps to achieve standardization, automation and remote control of pump station control, and alleviates the technical problems of low operating efficiency and serious energy waste in traditional control methods of circulation pumps. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the implementation methods of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the specific implementation methods or the description of the prior art. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0032] Figure 1 A flow chart of a heat exchange station circulating pump characteristic adaptive optimization control method provided by an embodiment of the present invention;

[0033] Figure 2 A schematic diagram of a heat exchange station circulating pump characteristic adaptive optimization control system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0035] Example 1

[0036] Figure 1 This is a flow chart of a heat exchange station circulating pump characteristic adaptive optimization control method provided by an embodiment of the present invention. Figure 1As shown, the method specifically includes the following steps:

[0037] Step S102: establishing a building terminal heat transfer model and a building energy balance model based on the temperature distribution on the radiator surface.

[0038] Step S104: establishing a building load prediction model based on the building terminal heat transfer model and the building energy balance model.

[0039] Step S106 , numerically modeling the performance curve of the circulating pump of the heat exchange station to establish a circulating pump performance model.

[0040] Step S108: Based on the deep deterministic policy gradient algorithm, a neural network control model for adaptive control of the heat exchange station circulating pump is established.

[0041] Step S110: Based on the building load prediction model and the circulating pump performance model, the state, action and reward function of the intelligent agent of the neural network control model are established.

[0042] Step S112: Based on the state, action and reward function of the intelligent agent, the neural network control model is trained to obtain a trained control model, and the heat exchange station circulation pump is controlled in real time based on the trained control model.

[0043] Specifically, the specific heat capacity of the water within the radiator is approximately 10 times that of its steel casing, so the steel casing is ignored when constructing the radiator's dynamic heat transfer model. The temperature distribution on the radiator surface decreases vertically but is uniformly distributed horizontally, simplifying the heat dissipation process to a one-dimensional heat transfer model. The building terminal heat transfer model includes: ;

[0044] Where, T(t,y) is the temperature distribution of the radiator in the vertical direction, unit ℃; t is time, unit s; y is the vertical position of the radiator, Y is the height of the radiator, unit m; V is the volume of the radiator, unit m³; q(t) is the volume flow rate of water flowing through the radiator at time t, unit m³ / s; K is the heat transfer coefficient of the radiator, W / (㎡·℃); F is the heat exchange area of ​​the radiator, unit ㎡; c is the specific heat capacity of water, unit kJ / (kg·℃); ρ is the density of water, unit kg / m³; T in is the indoor temperature of the building room, unit is ℃.

[0045] Specifically, the building energy balance model includes the exterior wall energy balance equation, the interior wall energy balance equation, and the indoor air energy balance equation;

[0046] The exterior wall (including the roof) only considers one-dimensional heat transfer in the thickness direction. The energy balance equation for the exterior wall includes:

[0047] ;

[0048] Where, is the specific heat capacity of the exterior wall, in J / (kg·℃); is the mass of the exterior wall, in kg; for The temperature of the exterior wall at the moment, in °C; is the exterior wall area, unit: m 2 ; is the convection heat transfer coefficient between the inner surface of the exterior wall and the indoor air, unit W / (m 2 K); for Indoor temperature at the moment, unit: ℃; is the convection heat transfer coefficient between the exterior wall surface and the outdoor air, in W / (m 2 K); for Outdoor temperature at the moment, unit: ℃; is the absorption rate of solar radiation by the exterior wall surface; for Solar radiation intensity at the moment, unit: W / m 2 .

[0049] For interior walls (including floors), only one-dimensional heat transfer in the thickness direction is considered, and heat transfer between adjacent rooms and between upper and lower floors is ignored. The energy balance equation for the interior wall includes:

[0050] ;

[0051] Where, is the specific heat capacity of the inner wall, unit is J / (kg·℃); is the mass of the inner wall, in kg; for The temperature of the inner wall at the moment, in °C; is the interior wall area, unit: m 2 .

[0052] The indoor air energy balance equation includes: ;

[0053] Where, is the specific heat capacity of indoor air, unit is J / (kg·K); is the indoor air quality, unit is kg; is the heat transfer coefficient of the exterior window, unit W / (m 2 K); is the exterior window area, in m 2 ; for The outdoor temperature at the moment, in ℃.

[0054] To achieve active optimization control of circulating pump operation, it is necessary to first obtain a predicted value for the building load. The terminal building load directly determines the system's required heat and circulation flow rate. This invention uses multiple linear regression to perform short-term load predictions, providing a basis for subsequent pump control decisions. Multiple linear regression constructs a linear combination model to fit the relationship between the terminal building heat load and a set of influencing factors. Specifically, the building load prediction model includes:

[0055] ;

[0056] Where, for The predicted heat load of the building at the moment, in kW; for The actual heat load of the building at the moment, in kW; for The indoor temperature of the building at the moment, in °C; for Solar radiation intensity at the moment, in W / m²; for Outdoor relative humidity at the moment, unit: % is the model coefficient to be estimated, is the residual term, which obeys the normal distribution;

[0057] The actual heat load calculation formula for the building includes: ;

[0058] Where, for The actual heat load of the building at the moment, in kW; is the specific heat capacity of water, in kJ / (kg·℃); is the density of water, in kg / m 3 ; is the circulation pump flow rate, unit is m 3 / s; for Secondary side water supply temperature at this moment, unit: °C; for Secondary side return water temperature at the moment, unit: °C.

[0059] The model's estimated coefficients are determined through training with historical operating data. Specifically, the initial building load forecasting model is trained using this historical data. Once sufficient data for the current heating season is collected, the model is continuously regressed using this new data. By inputting the next-moment weather forecast and current-moment indoor data, the building load at the next moment can be predicted. This prediction result is directly fed into the reinforcement learning controller to assist in optimizing the pump speed control strategy.

[0060] Specifically, the circulation pump performance model includes: ;

[0061] Where, is the circulation pump efficiency, unit: % is the circulation pump flow rate, unit is m 3 / s; is the head of the circulating pump, in m·H2O; is the circulation pump head coefficient, is the effective power of the circulation pump, in kW; is the density of water, in kg / m 3 ; is the acceleration due to gravity, 9.81 m / s²; is the shaft power of the circulation pump, in kW; is the motor efficiency of the circulation pump, unit: % is the electrical power of the circulation pump, in kW; It is the ratio of the operating frequency of the circulating pump to the rated frequency.

[0062] Specifically, in order to enable the deep reinforcement learning algorithm to effectively explore and optimize the operating status of the circulation pump, the circulation pump performance curve was numerically modeled, and the head expression of the circulation pump at different flow rates was established. Based on this, a response model of the reinforcement learning interactive environment was constructed. Binomial fitting was used at a constant frequency to obtain the flow-head characteristic curve of the circulation pump: ;

[0063] In actual projects, secondary network circulation pumps generally use variable frequency circulation pumps to reduce transmission and distribution energy consumption. According to the similarity law, when the same circulation pump is running at variable speed, the performance parameters between similar operating points meet the following formula:

[0064] ;

[0065] The subscript 0 indicates the parameters of the circulating pump at the rated frequency of 50 Hz, and no subscript indicates the parameters of the circulating pump at the operating frequency. is the speed of the circulation pump, is the frequency of the circulation pump. is the frequency ratio, which represents the ratio of the operating frequency of the circulating pump to the rated frequency. Substituting the proportional relationship of the circulating pump parameters in formula (2) into formula (1), the flow-head relationship of the variable frequency circulating pump is obtained: ;

[0066] Effective power of circulation pump The total energy obtained by the fluid passing through the pump per unit time is: ;

[0067] Shaft power of circulation pump The mechanical power transmitted from the motor to the circulation pump shaft is: ;

[0068] Efficiency of the circulation pump The degree to which the input shaft power is utilized by the fluid is: ;

[0069] Using binomial fitting, the flow-efficiency characteristic curve of the circulation pump is obtained as follows: ;

[0070] Where A, B, and C are the efficiency coefficients of the circulation pump.

[0071] The flow rate and pressure at the operating point at different frequencies are fitted using binomial equations to obtain the system resistance characteristic curve. The system resistance characteristic curve reflects the relationship between flow resistance and flow rate in the pipe network. The resistance pressure drop is proportional to the square of the flow rate: ;

[0072] Where: For the system flow The required head is in m·H2O; The system resistance coefficient comprehensively reflects the influence of pipe length, diameter, local resistance, valve opening, etc. on water flow. The unit is s 2 / m 5 .

[0073] Specifically, the neural network control model includes a policy network Actor and a value network Critic; the policy network is a policy network that inputs the current system state and outputs continuous actions, namely the control frequency of the circulating pump, including two layers of fully connected networks and one layer of rectified linear unit activation network (Rectified Linear Unit, ReLU); the value network includes a value function network, which inputs and actions and outputs the Q value of the action in this state (i.e., the estimated future cumulative reward). It is a three-layer structure.

[0074] Reinforcement learning is an intelligent learning method that learns optimal behavioral strategies through interaction with the environment, continuous trial and error, and accumulation of experience. Its basic idea is derived from the "reward and punishment mechanism" in psychology: the intelligent agent gradually improves its behavioral strategy by trying different actions based on the rewards or punishments given by the environment to maximize long-term cumulative returns. The embodiment of the present invention adopts the Deep Deterministic Policy Gradient (DDPG) algorithm to achieve adaptive operating point control of the circulating pump. DDPG is a reinforcement learning method based on the actor-critic structure and is suitable for continuous action space scenarios. The control variable of the circulating pump is the variable frequency, which is a typical continuous control problem, so the use of DDPG is particularly appropriate.

[0075] Specifically, the state of the agent in the neural network control model is the state observed by the agent at each control step, including:

[0076] ;

[0077] Where, For intelligent agents The state vector at the moment, for Circulation pump flow rate at any moment, unit: m 3 / s; for The pressure difference between the inlet and outlet of the circulation pump at any moment, in m·H2O; for The frequency of the circulating pump at any moment, in Hz; for The electrical power of the circulation pump at any moment, in kW; for Secondary side water supply temperature at this moment, unit: °C; for Secondary side return water temperature at this moment, unit: °C; for The predicted heat load of the building at the moment, in kW.

[0078] The action of the agent in the neural network control model, that is, how the agent controls the environment, in this embodiment of the present invention, the action is the frequency increment of the circulating pump, specifically including: ;

[0079] Where, for The adjustment increment of the circulating pump frequency at all times, in Hz; The maximum frequency change step allowed for the action, in Hz; is the frequency change of the circulating pump.

[0080] The reward function of the agent in the neural network control model significantly affects the convergence of deep reinforcement learning and the final control performance, and determines the control optimization goal. The goal of the embodiment of the present invention is to make the circulating pump operate in the high-efficiency zone while matching the terminal heat load. Specifically, the reward function includes: ;

[0081] Where, for The moment reward function, for The actual heat load of the building at the moment, in kW; is the weighting coefficient.

[0082] Specifically, the neural network control model also includes environmental feedback: the neural network control model receives the frequency , use the circulation pump flow-head characteristic curve and system characteristic curve to calculate the flow Heyangcheng , further calculate the effective power ; Use the circulation pump flow-efficiency characteristic curve to obtain the circulation pump operating efficiency at the current flow rate , further calculate the circulating pump electric power through the effective power ; Through the current flow and water temperature , using the building terminal radiator model and the building's most unfavorable user to calculate the most unfavorable user's return water temperature as the building's return water temperature and the indoor temperature that is most unfavorable to users , further calculate the actual heat load of the building And predict the building heat load at the next moment .

[0083] Specifically, the Actor-Critic structure update strategy in the neural network control model is as follows:

[0084] Initialize the experience pool Replay Buffer and randomly explore and collect , each round samples a training batch from the experience pool and updates the Critic network, with the goal of minimizing the temporal difference error:

[0085] ;

[0086] Where, is the loss function of the Critic network; Parameters that need to be optimized for the Critic network; is the batch size of sampling; is the status of the i-th piece of experience data; is the action of the i-th piece of experience data; is the current critic network pair of Value estimation; The goal of the i-th experience value;

[0087] .

[0088] Where, The immediate reward obtained for the i-th experience; is the discount factor, which controls the impact of future rewards; is the next state after executing the action in step i; For the target Actor network The following actions are given; For the target Critic network of Value assessment; are the parameters of the target Actor and target Critic networks respectively.

[0089] Use policy gradient to optimize the Actor network, the goal is to maximize the Critic network given value:

[0090] ;

[0091] Where, Parameterize the policy network's objective function (total expected reward) gradient; Parameters of the Actor network; is the batch size of sampling; Given by the Critic network Value function; For The function finds the gradient with respect to action a and is calculated on the action output by the current policy of the Actor network; Output the gradient of the action to its parameters for the policy network; is the policy objective function.

[0092] Specifically, the target network soft replacement includes:

[0093] To ensure stable training, the target network is introduced to smoothly iterate and slowly follow the main network update to avoid violent fluctuations during the policy update process: ;

[0094] Where, are the parameters of the target network; are the parameters of the current main network; is the soft update factor.

[0095] The neural network control model is trained through multiple rounds, repeating the sampling-update-soft replacement process. When the average reward steadily increases and the control frequency tends to be reasonable, it indicates that the strategy has converged. The final derived Actor network can be used as a control strategy and used in real-time systems according to the state. Output frequency , used to adjust the circulation pump in real time.

[0096] This embodiment of the present invention provides a method for adaptive optimization control of circulating pump characteristics in a heat exchange station. This method constructs a neural network model consisting of a policy network (actor) and a value network (critic) to control the circulating pump frequency in real time for efficient and energy-saving operation. The system state design encompasses key operating parameters of the circulating pump, including flow rate, pressure differential, frequency, electrical power, supply and return water temperature, and predicted building heat load, forming the state vector perceived by the agent at each control step. The control action is set as the pump frequency adjustment increment to guide the dynamic adjustment of the circulating pump operating point. The reward function integrates load tracking accuracy and operating efficiency deviation objectives, aiming to guide the system to operate within the high-efficiency range while meeting heat load requirements. The environmental model provides feedback on various state information based on the pump's flow-head characteristic curve, flow-efficiency characteristic curve, system resistance curve, building terminal model, and building user model, providing a realistic interactive environment. During training, the critic network is updated using an experience replay mechanism and temporal difference method. The actor network is optimized using policy gradients, and the target network is introduced for soft replacement to enhance training stability and convergence. When the average reward of the agent tends to be stable and the control behavior is reasonable during training, it indicates that the strategy has converged. The final trained Actor network can be deployed as an online control strategy to output the circulating pump frequency according to the real-time status, realizing efficient and adaptive operation of the system.

[0097] Example 2

[0098] Figure 2 Schematic diagram of a heat exchange station circulating pump characteristic adaptive optimization control system according to an embodiment of the present invention. Figure 2 As shown, the system includes: a first establishing module 10 , a second establishing module 20 , a third establishing module 30 , a fourth establishing module 40 , a fifth establishing module 50 and a control module 60 .

[0099] Specifically, the first establishing module 10 is used to establish a building terminal heat transfer model and a building energy balance model based on the temperature distribution of the radiator surface;

[0100] The second establishing module 20 is used to establish a building load prediction model based on the building terminal heat transfer model and the building energy balance model;

[0101] The third building module 30 is used to perform numerical modeling on the performance curve of the heat exchange station circulating pump and establish a circulating pump performance model;

[0102] A fourth establishing module 40 is used to establish a neural network control model for adaptive control of a heat exchange station circulating pump based on a deep deterministic policy gradient algorithm;

[0103] a fifth establishing module 50 for establishing states, actions, and reward functions of an agent of a neural network control model based on the building load prediction model and the circulating pump performance model;

[0104] The control module 60 is used to train the neural network control model based on the state, action and reward function of the intelligent agent to obtain the trained control model, and to perform real-time control of the heat exchange station circulation pump based on the trained control model.

[0105] The present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method provided in the embodiment of the present invention when executing the computer program.

[0106] The present invention also provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the method provided in the embodiment of the present invention is implemented.

[0107] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

[0108] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A heat exchange station circulating pump characteristic adaptive optimization control method, characterized in that: include: Based on the temperature distribution on the radiator surface, a building terminal heat transfer model and a building energy balance model are established; Establishing a building load prediction model based on the building terminal heat transfer model and the building energy balance model; Conduct numerical modeling on the performance curve of the heat exchange station circulation pump and establish a circulation pump performance model; Based on the deep deterministic policy gradient algorithm, a neural network control model for adaptive control of the heat exchange station circulating pump is established; Establishing the state, action, and reward functions of the agent of the neural network control model based on the building load prediction model and the circulating pump performance model; Based on the state, action and reward function of the intelligent agent, the neural network control model is trained to obtain a trained control model, and the heat exchange station circulation pump is controlled in real time based on the trained control model; The building load prediction model includes: ； Where, for The predicted heat load of the building at any given moment, for The actual heat load of the building at any moment, for The indoor temperature of the building at all times, for Outdoor temperature at any time, for Solar radiation intensity at any moment, for The relative humidity outside at any time, is the model coefficient to be estimated, is the residual term; The actual heat load calculation formula for the building includes: ; Where, for The actual heat load of the building at any moment, is the specific heat capacity of water, is the density of water, is the circulation pump flow rate, for Secondary side water supply temperature at this moment, for Secondary side return water temperature at the moment; The model coefficients to be estimated are determined through training with historical operating data; The circulating pump performance model includes: ； Where, is the circulation pump efficiency, is the circulation pump flow rate, is the circulating pump head, is the circulation pump head coefficient, is the effective power of the circulation pump, is the density of water, is the acceleration due to gravity, is the shaft power of the circulation pump, is the motor efficiency of the circulation pump, is the electrical power of the circulation pump, It is the ratio of the operating frequency of the circulating pump to the rated frequency.

2. The method according to claim 1, wherein: The building terminal heat transfer model includes: ； Where T(t,y) is the temperature distribution of the radiator in the vertical direction, t is time, y is the vertical position of the radiator, Y is the height of the radiator, V is the volume of the radiator, q(t) is the volume flow rate of water flowing through the radiator at time t, K is the heat transfer coefficient of the radiator, F is the heat exchange area of ​​the radiator, c is the specific heat capacity of water, ρ is the density of water, T in is the indoor temperature of the building room.

3. The method according to claim 1, wherein: The building energy balance model includes the exterior wall energy balance equation, the interior wall energy balance equation and the indoor air energy balance equation; wherein, The energy balance equation of the exterior wall includes: ； The energy balance equation of the interior wall includes: ； The indoor air energy balance equation includes: ； Where, is the specific heat capacity of the exterior wall, For the quality of the exterior wall, for The temperature of the exterior wall at all times, is the exterior wall area, is the convection heat transfer coefficient between the inner surface of the exterior wall and the indoor air, for Indoor temperature at all times, is the convective heat transfer coefficient between the exterior wall surface and the outdoor air, for Outdoor temperature at any time, is the absorption rate of solar radiation by the exterior wall surface, for Solar radiation intensity at any moment, is the specific heat capacity of the inner wall, For the quality of the interior wall, for The temperature of the inner wall at all times, is the interior wall area, is the specific heat capacity of indoor air, For indoor air quality, is the heat transfer coefficient of the exterior window, The area of ​​the exterior windows.

4. The method according to claim 1, wherein: The neural network control model includes a policy network and a value network; wherein the policy network includes two layers of fully connected networks and one layer of rectified linear unit activation network; and the value network includes a value function network.

5. The method according to claim 1, wherein: The state of the agent of the neural network control model includes: ； Where, For the intelligent agent The state vector at the moment, for Circulation pump flow at all times, for The pressure difference between the inlet and outlet of the circulating pump at all times, for Circulation pump frequency at all times, for The electrical power of the circulating pump at all times, for Secondary side water supply temperature at this moment, for Secondary side return water temperature at time for The predicted heat load of the building at any given moment; The actions of the agent of the neural network control model include: ； Where, for The adjustment increment of the circulating pump frequency at all times, is the maximum frequency change step allowed by the action, is the frequency change of the circulating pump; The reward function of the agent of the neural network control model includes: ； Where, for The moment reward function, for The actual heat load of the building at any moment, for The electrical power of the circulating pump at all times, is the weighting coefficient.

6. A heat exchange station circulating pump characteristic adaptive optimization control system, characterized in that: include: A first establishment module, a second establishment module, a third establishment module, a fourth establishment module, a fifth establishment module and a control module; wherein, The first establishing module is used to establish a building terminal heat transfer model and a building energy balance model based on the temperature distribution of the radiator surface; The second establishing module is used to establish a building load prediction model based on the building terminal heat transfer model and the building energy balance model; The third building module is used to perform numerical modeling on the performance curve of the heat exchange station circulating pump and establish a circulating pump performance model; The fourth establishing module is used to establish a neural network control model for adaptive control of the heat exchange station circulating pump based on a deep deterministic policy gradient algorithm; The fifth establishing module is used to establish the state, action and reward function of the agent of the neural network control model based on the building load prediction model and the circulating pump performance model; The control module is configured to train the neural network control model based on the state, action, and reward function of the agent to obtain a trained control model, and to perform real-time control of the heat exchange station circulating pump based on the trained control model; The building load prediction model includes: ； Where, for The predicted heat load of the building at any given moment, for The actual heat load of the building at any moment, for The indoor temperature of the building at all times, for Outdoor temperature at any time, for Solar radiation intensity at any moment, for The relative humidity outside at any time, is the model coefficient to be estimated, is the residual term; The actual heat load calculation formula for the building includes: ； Where, for The actual heat load of the building at any moment, is the specific heat capacity of water, is the density of water, is the circulation pump flow rate, for Secondary side water supply temperature at this moment, for Secondary side return water temperature at the moment; The model coefficients to be estimated are determined through training with historical operating data; The circulating pump performance model includes: ； Where, is the circulation pump efficiency, is the circulation pump flow rate, is the circulating pump head, is the circulation pump head coefficient, The effective power of the circulation pump, is the density of water, is the acceleration due to gravity, is the shaft power of the circulation pump, is the motor efficiency of the circulation pump, is the electrical power of the circulation pump, It is the ratio of the operating frequency of the circulating pump to the rated frequency.

7. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images