Optimal Scheduling Method for Ground Source Heat Pump System Based on Probabilistic Graph Digital Twin Model

Abstract

The present invention discloses an optimized scheduling method for a ground source heat pump system based on a probabilistic graph digital twin model, belonging to the field of operation optimization of ground source heat pump systems. The method comprises the following steps: S1, constructing a digital twin model based on a dynamic probabilistic graph decision network in the modeling stage; S2, calibrating the digital twin model in the calibration stage; S3, updating the digital twin model in the operation stage, and real-time outputting the optimal operation method of the ground source heat pump system through the updated digital twin model. The physical entity-twin system is abstracted as a set of coupled dynamic systems, which evolve over time in the physical state space and the digital state space and interact through observation data and control inputs, so as to solve the problems of complex multi-physical field coupling modeling and high computational resource requirements for dynamic optimization control of the ground source heat pump system, and realize real-time and automatic optimal control of the ground source heat pump system under complex and variable underground environments and load demands.

Description

Technical Field

[0001] The present invention belongs to the field of ground source heat pump system operation optimization, and specifically relates to a ground source heat pump system optimization scheduling method based on a probability graph digital twin model. Background Art

[0002] Although the ground source heat pump unit shows a high energy efficiency ratio in laboratory testing and production processes, it is often difficult to achieve the energy efficiency level calibrated at the factory in actual applications. The reasons for this difference are complex and varied, including the diversity of geological conditions, the complexity of underground heat sources, the difference between actual working conditions and experimental conditions, unreasonable initial design, and insufficient equipment operation and maintenance. The actual efficiency is significantly lower than the theoretical value, which will lead to low energy utilization and failure to achieve energy saving effects. In addition, due to the high initial investment in the ground source heat pump system, poor actual operating efficiency will extend the investment payback period, weaken its economy and market competitiveness, and thus affect the popularization and promotion of ground source heat pump technology. Therefore, improving the operating efficiency of the ground source heat pump system under actual conditions and optimizing the initial design plan are the core keys to the widespread application of the ground source heat pump system.

[0003] In order to solve the above problems, it is necessary to perceive the operating status of the ground source heat pump system in real time and accurately, and automatically optimize the operating parameters according to dynamic factors such as the load demand of the application scenario and the characteristics of the geological heat source, and quickly diagnose equipment failures and respond to abnormal operation in a timely manner. Digital twin technology provides an effective solution to achieve these goals. The digital twin of the ground source heat pump system, by building a virtual model corresponding to the ground source heat pump system, obtains the operating data, environmental parameters and application scenario load changes of the ground source heat pump in real time, and realizes the seamless connection between the physical system and the digital space, so as to accurately model, monitor in real time and dynamically adjust the system operation. At the same time, based on the real-time collected information, the physical system of the ground source heat pump and the characteristics of the geological heat source, the digital twin combines big data and artificial intelligence technology for in-depth analysis, identifies the inefficient state in the system operation, simulates and dynamically optimizes the control under different working conditions and load conditions, and implements predictive maintenance for the equipment, thereby minimizing the gap between the actual energy efficiency and the calibrated energy efficiency. In addition, the digital twin's simulation function is based on geological exploration data and building load forecasts. It evaluates the performance of the ground source heat pump system in different scenarios and working conditions during the design phase, helping designers evaluate the feasibility of the system and find the optimal heat exchanger layout, equipment selection, geothermal well network layout, and underground pipe size and quantity configuration to ensure that the system design is highly matched with the actual working conditions and achieve the best balance between initial investment and subsequent operating efficiency.

[0004] In recent years, significant progress has been made in digital twin technology, especially in data management, real-time monitoring, and visualization. There are also related research and applications on digital twin HVAC systems. However, currently, there is a lack of systematic practice and specialized research on the digital twin system of ground source heat pumps at home and abroad, especially the modeling methods and mathematical forms for constructing digital twins of ground source heat pump systems. Compared with establishing a digital twin for an HVAC system, establishing a digital twin for a ground source heat pump system involves modeling the underground environment and its coupling with above-ground equipment, detailed calculations of heat conduction processes, and optimization of long-term operation, requiring the handling of more variables and more complex model calculations. Summary of the Invention

[0005] In view of the above deficiencies in the prior art, the optimized scheduling method for a ground source heat pump system based on a probabilistic graph digital twin model provided by the present invention uses digital twin technology with low computational costs to achieve real-time and automatic optimized control of the ground source heat pump system, and solves the problems of complex multi-physical field coupling modeling and dynamic optimized control calculations of the ground source heat pump system, as well as high computational resource requirements.

[0006] To achieve the above invention objective, the technical solution adopted by the present invention is: an optimized scheduling method for a ground source heat pump system based on a probabilistic graph digital twin model, comprising the following steps:

[0007] S1. Construct a digital twin model based on a dynamic probabilistic graph decision network in the modeling stage;

[0008] S2. Calibrate the digital twin model in the calibration stage;

[0009] S3. Update the digital twin model in the operation stage, and output the optimal operation method of the ground source heat pump system in real time through the updated digital twin model.

[0010] Furthermore: In S1, the nodes in the digital twin model based on the dynamic probabilistic graph decision network include random variables of physical states, digital states, observed data, research quantities, control inputs, and reward equations at each time point.

[0011] Furthermore: The set of physical states , where is a virtual time, represents the modeling stage, represents the calibration stage of the digital twin model, is the prediction and decision-making time, is the value of relevant parameter variables reflecting the operation state of the ground source heat pump system at different times;

[0012] The set of digital states , where The digital state values of relevant parameter variables for constructing the digital twin calculation model of the ground source heat pump system at different times;

[0013] The set of observed data , where is the observed data reflecting the physical state changes collected by sensors or detection devices in the ground source heat pump system at different times;

[0014] The set of research quantities , where is the coefficient of performance of the ground source heat pump system at different times;

[0015]

[0016] In the formula, is the average value of the heat at the load end from time to time, is the average value of the input power of the heat pump unit from time to time, is the average value of the input power of the water pump from time to time, is the average value of the power consumption of the fan coil at the load end from time to time, is the average value of the energy consumption of the control system from time to time;

[0017] The set of control inputs is , where is the operation control scheme of the ground source heat pump system at different times;

[0018] The set of reward equations is , where is the reward equation in the model calibration stage of the ground source heat pump digital twin system at time;

[0019]

[0020] In the formula, is the first duration of operation in the model calibration stage, and are the observed values and predicted values of the digital twin model of the load at each time respectively, s is the observation value collection time;

[0021] is for the ground source heat pump digital twin system at Reward equation for the moment model calibration stage;

[0022]

[0023] Wherein, is the second duration of operation in the model calibration stage, and are the observed value and the digital twin model prediction value of the ground source side outlet temperature at each moment respectively;

[0024] is the reward equation of the ground source heat pump digital twin system in the moment model calibration stage;

[0025]

[0026] Wherein, is the third duration of operation in the model calibration stage, and are the observed value and the digital twin model prediction value of the load side return water temperature at each moment respectively;

[0027] is the reward equation of the digital twin model in t the moment model operation stage;

[0028]

[0029] Wherein, is the coefficient of performance of the ground source heat pump system at each observed value collection moment.

[0030] Furthermore: In S1, the expression of the digital twin model based on the dynamic probability graph decision network is specifically:

[0031]

[0032] Wherein, is the conditional probability distribution function of the system showing the digital state , calculating the research quantity and obtaining the reward at the cut-off moment given the historical observation data and the control input, is the conditional probability distribution function of updating the digital twin to obtain the digital state when the digital state of the ground source heat pump system at the previous moment is , the control input is executed, and the observation data is collected at this moment, is the conditional probability distribution function of the digital state Obtain the research quantity The conditional probability distribution function of is the conditional probability distribution function of obtaining the reward value given the digital state at the current moment, the research quantity , the control input and the observed data .

[0033] Furthermore: Specifically, S2 is as follows:

[0034] The ground source heat pump system mainly consists of three parts: the load end, the ground source side, and the ground source heat pump unit. The load end refers to the users such as buildings that are heated, cooled, or provided with hot water by the ground source heat pump system; the ground source side is the underground heat source, and the system usually exchanges heat with the soil, groundwater, or surface water through buried pipes, etc.; the ground source heat pump unit consists of a compressor, a heat exchanger, etc., and is the core equipment of the system, transferring the heat pump on the ground source side to the load end, or vice versa. Correspondingly, the digital twin model of the ground source heat pump system contains three parts: the load end model, the ground source side model, and the ground source heat pump unit model. In the model calibration stage, these three models need to be calibrated separately. At calibrate the load end model and related parameters at the moment to minimize as the goal and adjust the load end model; at calibrate the ground source side model and related parameters at the moment to minimize as the goal and adjust the ground source side model; at calibrate the ground source heat pump unit model and related parameters at the moment to minimize as the goal and adjust the unit model.

[0035] Furthermore: In S3, the method for updating the digital twin model is specifically as follows:

[0036] At the moment t in the operation stage, collect the observed data , and according to the observed data update the predicted values of the digital state , the research quantity and the reward equation to the true values calculated according to the observed data . Use the expectation maximization algorithm to update the marginal / conditional probability distribution functions in the model, obtain the joint probability distribution function of the random variables of all nodes in the digital twin model, and then output the predicted values through this joint probability distribution function, including the digital state t from moment +1 to moment , the research quantity and the reward equation ;

[0037]

[0038] Wherein, is the joint probability distribution function, is that the digital state of the ground source heat pump system at the previous moment is , and the control input is taken for the ground source heat pump system, and then the conditional probability distribution function of the digital state is obtained, is the conditional probability distribution function of obtaining the observation data given the digital state , is at the moment given the digital state of the ground source heat pump system and the research quantity , and the conditional probability distribution function of taking the control input . This conditional probability distribution function is based on the optimal control scheme solved at the previous moment.

[0039] Furthermore, in S3, the specific method for obtaining the optimal operation method of the ground source heat pump system is as follows:

[0040] Based on the predicted value output at the moment t , a Markov decision process model is constructed to solve and select the optimal control scheme as the optimal operation method of the ground source heat pump system;

[0041] Among them, the expression of the Markov decision process model is specifically:

[0042]

[0043] Wherein, is the reward equation of the operation energy efficiency of the ground source heat pump system, , is the performance coefficient of the ground source heat pump system at each observation value collection moment before the observation value collection moment , is the discount factor for the future moment reward, , is to find the ground source heat pump system operation strategy that maximizes .

[0044] The beneficial effects of the present invention are as follows: The present invention provides an optimized scheduling method for a ground source heat pump system, and constructs a digital twin model by using a dynamic probabilistic graph decision network. Compared with the prior art, the following effects are achieved:

[0045] (1)Abstract the physical entity-twin system as a set of coupled dynamic systems that evolve over time in the physical state space and the digital state space and interact through observation data and control inputs to solve the problems of complex multi-physics field coupling modeling and high computational resource requirements in the ground source heat pump system, and realize real-time and automatic optimal control of the ground source heat pump system under complex and variable underground environments and load demands.

[0046] (2)Abstract the physical-data optimization decision-making process as multi-dimensional time-series variables and their correlations. The form of the digital twin model construction can be adjusted according to the application scenarios, decision-making requirements, and data collection capabilities of the ground source heat pump system. Moreover, it can utilize mature theories and methods such as Bayesian statistics, control theory, and stochastic processes, featuring generality, flexibility, and economy, transforming the one-time digital twin based on customization into a robust digital twin that can be deployed on a large scale. Description of the Drawings

[0047] Figure 1 It is a flowchart of the optimization scheduling method for the ground source heat pump system based on the probabilistic graph digital twin model of the present invention.

[0048] Figure 2 It is a schematic diagram of the digital twin model based on the dynamic probabilistic graph decision network. Detailed Embodiments

[0049] The following describes the detailed embodiments of the present invention to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the detailed embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions created using the concept of the present invention are within the scope of protection.

[0050] As Figure 1 shown, in an embodiment of the present invention, the optimization scheduling method for the ground source heat pump system based on the probabilistic graph digital twin model includes the following steps:

[0051] S1. Construct a digital twin model based on the dynamic probabilistic graph decision network in the modeling stage;

[0052] S2. Calibrate the digital twin model in the calibration stage;

[0053] S3. Update the digital twin model in the operation stage, and the optimal operation method of the ground source heat pump system is output in real time through the updated digital twin model.

[0054] In this embodiment, a digital twin model based on the dynamic probabilistic graph decision network is constructed for the ground source heat pump system, as Figure 2As shown. Time is discretized in hours or days (depending on decision-making requirements or data storage capabilities). Nodes in the digital twin model represent random variables of physical states, digital states, observed data, research quantities, control inputs, and reward equations at various time points.

[0055] In S1, nodes in the digital twin model based on the dynamic probabilistic graph decision network include random variables of physical states, digital states, observed data, research quantities, control inputs, and reward equations at various time points.

[0056] Set of physical states , where is a virtual moment, represents the modeling phase, represents the calibration phase of the digital twin model, is the prediction and decision-making moment, are the values of relevant parameter variables reflecting the operating state of the ground source heat pump system at different times;

[0057] In this embodiment, represents the calibration phases of three different types of parameters, and each phase can include multiple moments in hours or days. are the values of relevant parameter variables reflecting the operating state of the ground source heat pump system at different times, including the inlet temperature of the ground source side, the inlet pressure of the ground source side, the inlet flow rate of the ground source side, the outlet temperature of the ground source side, the outlet pressure of the ground source side, the outlet flow rate of the ground source side, the supply water temperature at the load end, the supply water pressure at the load end, the supply water flow rate at the load end, the return water temperature at the load end, the return water pressure at the load end, the return water flow rate at the load end, the inlet temperature of the heat pump unit, the outlet temperature of the heat pump unit, the inlet pressure of the heat pump unit, the outlet pressure of the heat pump unit, the inlet flow rate of the heat pump unit, the outlet flow rate of the heat pump unit, the flow rate of the ground source side pump, the inlet pressure of the ground source side pump, the outlet pressure of the ground source side pump, the flow rate of the load end pump, the inlet pressure of the load end pump, the outlet pressure of the load end pump, the filter differential pressure, the pressure of the main and branch collectors of the buried pipe, the inlet temperature of the ground source side heat exchanger, the inlet pressure of the ground source side heat exchanger, the inlet flow rate of the ground source side heat exchanger, the outlet temperature of the ground source side heat exchanger, the outlet pressure of the ground source side heat exchanger, the outlet flow rate of the ground source side heat exchanger, the inlet temperature of the load end heat exchanger, the inlet pressure of the load end heat exchanger, the inlet flow rate of the load end heat exchanger, the outlet temperature of the load end heat exchanger, the outlet pressure of the load end heat exchanger, the outlet flow rate of the load end heat exchanger, the compressor energy consumption, the pump energy consumption, the energy consumption of the load end fan coil, the energy consumption of the control system, the working state (fault) of equipment such as the unit, auxiliary equipment, and valves, the load, the temperature of each measurement point in the soil, etc.

[0058] Set of digital states , where The digital state values of relevant parameter variables for constructing the digital twin calculation model of the ground source heat pump system at different times;

[0059] In this embodiment, are the digital state values of relevant parameter variables for constructing the digital twin calculation model of the ground source heat pump system at different times, and the parameters are consistent with the physical state parameters.

[0060] Set of observed data , where are the observed data reflecting the physical state changes collected by sensors or detection devices in the ground source heat pump system at different times;

[0061] In this embodiment, are the observed data reflecting the physical state changes collected by sensors or detection devices in the ground source heat pump system at different times, including the inlet temperature of the ground source side, the inlet pressure of the ground source side, the inlet flow rate of the ground source side, the outlet temperature of the ground source side, the outlet pressure of the ground source side, the outlet flow rate of the ground source side, the supply water temperature of the load end, the supply water pressure of the load end, the supply water flow rate of the load end, the return water temperature of the load end, the return water pressure of the load end, the return water flow rate of the load end, the energy consumption of the compressor, the energy consumption of the water pump, the energy consumption of the load end fan coil, the energy consumption of the control system, the inlet temperature of the heat pump unit, the outlet temperature of the heat pump unit, the inlet pressure of the heat pump unit, the outlet pressure of the heat pump unit, the inlet flow rate of the heat pump unit, the outlet flow rate of the heat pump unit, the flow rate of the ground source side water pump, the inlet pressure of the ground source side water pump, the outlet pressure of the ground source side water pump, the flow rate of the load end water pump, the inlet pressure of the load end water pump, the outlet pressure of the load end water pump, the inlet pressure of the filter, the outlet pressure of the filter, the inlet temperature of the ground source side heat exchanger, the inlet pressure of the ground source side heat exchanger, the inlet flow rate of the ground source side heat exchanger, the outlet temperature of the ground source side heat exchanger, the outlet pressure of the ground source side heat exchanger, the outlet flow rate of the ground source side heat exchanger, the inlet temperature of the load end heat exchanger, the inlet pressure of the load end heat exchanger, the inlet flow rate of the load end heat exchanger, the outlet temperature of the load end heat exchanger, the outlet pressure of the load end heat exchanger, the outlet flow rate of the load end heat exchanger, the working state (fault) of equipment such as the unit, auxiliary equipment, and valves, the outdoor temperature of the load end, the outdoor humidity of the load end, the outdoor solar radiation of the load end, the operating rules of the load end, the number of people and occupancy rate in the load end, the number and usage rate of lighting in the load end, the number and usage rate of energy-consuming equipment in the load end, the total length of the buried pipe, the pressure of the manifold, and the temperature of each measurement point of the soil.

[0062] Set of research quantities , where is the coefficient of performance of the ground source heat pump system at different times;

[0063]

[0064] In the formula, is from Time arrives The average heat at the load end at a certain time is from a certain time to the average input power of the heat pump unit at a certain time is from a certain time to the average input power of the water pump at a certain time is from a certain time to the average power consumption of the fan coil at the load end at a certain time is from a certain time to the average energy consumption of the control system at a certain time

[0065] The set of control inputs is wherein is the operation control scheme of the ground source heat pump system at different times

[0066] In this embodiment is the operation control scheme of the ground source heat pump system at different times, including operation mode (cooling, heating, shutdown), ground source side water pump flow rate, ground source side water pump head, ground load end water pump flow rate, load end water pump head, load end supply water temperature, compressor flow rate

[0067] The set of reward equations is wherein is the reward equation of the ground source heat pump digital twin system in the model calibration stage at a certain time

[0068]

[0069] In the formula is the first duration of operation in the model calibration stage and are respectively the observed value and the predicted value of the digital twin model of the load at each time s is the observation value collection time

[0070] is the reward equation of the ground source heat pump digital twin system in the model calibration stage at a certain time

[0071]

[0072] In the formula is the second duration of operation in the model calibration stage and are respectively the observed value and the predicted value of the digital twin model of the ground source side outlet temperature at each time

[0073] For the reward equation in the model calibration stage of the ground source heat pump digital twin system at time;

[0074]

[0075] Wherein, is the third duration of operation in the model calibration stage, and are respectively the observed value and the predicted value of the digital twin model for the return water temperature at the load end at each moment;

[0076] For the reward equation in the model operation stage of the digital twin model at t time;

[0077]

[0078] Wherein, is the coefficient of performance of the ground source heat pump system at each observation value collection moment.

[0079] In this embodiment, the reward equation is used to evaluate the operation efficiency of the system. The higher this value is, the more efficient and energy-saving the system operation is.

[0080] In S1, the expression of the digital twin model based on the dynamic probability graph decision network is specifically:

[0081]

[0082] Wherein, is the cut-off time , given the historical observation data and control input, the system exhibits the digital state , calculates the research quantity and obtains the conditional probability distribution function of the reward , is the digital state of the ground source heat pump system at the previous moment being , executes the control input , and at this moment collects the observation data to update the digital twin to obtain the conditional probability distribution function of the digital state , is the conditional probability distribution function of obtaining the research quantity given the digital state of the updated digital twin at the current moment, is the digital state given at the current moment , research quantity , control input and observation data , to obtain the reward value The conditional probability distribution function.

[0083] As Figure 2 shown, the arrows connecting the nodes represent the correlation relationships between variables, and these correlation relationships are defined by the conditional probability density function so that the digital twin model has the ability to simulate uncertainty and make optimized decisions considering uncertainty.

[0084] The physical state of the ground source heat pump system and the digital state of its digital twin form two dynamic systems that evolve over time. The digital twin of the ground source heat pump system uses the observation data collected by sensors to predict the physical state of the system operation, so as to make the optimal control input for the ground source heat pump system. While minimizing the control cost and ensuring the observability of the system state, the coefficient of performance of the ground source heat pump system is improved. The optimized ground source heat pump system generates new observation data, updates the model and digital state in its digital twin, and makes it more accurately simulate the physical state of the real system. The updated digital twin further analyzes the operation process of the ground source heat pump system, predicts the future state evolution of the ground source heat pump system through a more accurate algorithm, and then better optimizes the decision-making of the next control input. This process evolves cyclically, forming a collaborative multi-directional coupling among the ground source heat pump system, digital twin, data collection, calculation model, and optimized control decision, and finally forming a ground source heat pump digital twin model that accurately simulates and supports predictive decision-making.

[0085] As Figure 2 shown, at the moment of the modeling stage, according to the mathematical theory of the ground source heat pump system, such as the building load calculation model, compressor model, condenser model, evaporator model, ground source heat pump system model, and buried pipe heat transfer model, a twin of the ground source heat pump system is constructed, including calculation models such as the heat absorption of the ground source side heat exchanger, compressor input work, heat dissipation of the load side heat exchanger, heat transfer of the load side equipment, energy consumption of the ground source side water pump, energy consumption of the load side water pump, buried pipe heat transfer, system coefficient of performance, load, indoor temperature and load balance at the load side, heat transfer between the load side and the fan coil, compressor power, soil temperature, inlet and outlet temperatures at the ground source side, etc.

[0086] S2 is specifically:

[0087] The ground source heat pump system mainly consists of three parts: the load end, the ground source side, and the ground source heat pump unit. The load end refers to the users such as buildings that are heated, cooled, or provided with hot water by the ground source heat pump system; the ground source side is the underground heat source, and the system usually exchanges heat with the soil, groundwater, or surface water through buried pipes, etc.; the ground source heat pump unit consists of a compressor, a heat exchanger, etc., and is the core equipment of the system, which transfers the heat pump on the ground source side to the load end, or vice versa. Correspondingly, the digital twin model of the ground source heat pump system contains three parts: the load end model, the ground source side model, and the ground source heat pump unit model. In the model calibration stage, these three models need to be calibrated separately. At time, calibrate the load end model and related parameters. At time, calibrate the ground source side model and related parameters. At time, calibrate the ground source heat pump unit model and related parameters.

[0088] In the calibration stage, at time, calibrate the load end model and related parameters. According to the laws of outdoor weather, personnel flow, equipment and lighting usage at the user end, construct a simulation model of the load end operation, and conduct simulation experiments to adjust parameters such as the heat transfer coefficient and the heat dissipation of personnel, equipment, and lighting. The control input at this time is the decision of the simulation experiment design scheme.

[0089] At time, calibrate the ground source side model and related parameters. According to the "Technical Specification for Ground Source Heat Pump System Engineering" (GB-50366-2009) and the "Standard for Geotechnical Investigation of Ground Source Heat Pump System Engineering" (CJJ / T-291-2019), carry out geotechnical investigation and thermal response test to adjust parameters such as the convective heat transfer coefficient, soil thermal conductivity, volumetric specific heat coefficient, heat transfer per unit length of the heat exchange hole, and comprehensive heat transfer coefficient of the heat exchange well. The control input at this time is the decision of the on-site test scheme such as the number of boreholes, borehole depth, borehole location, buried pipe form, and buried pipe outer diameter.

[0090] At time, calibrate the ground source heat pump unit model and related parameters. Construct a simulation model of the unit side loop and the load end loop in the ground source heat pump system, and conduct simulation experiments to adjust parameters such as the actual power consumption of the compressor, refrigerant specific volume, polytropic index, enthalpy value, and valve regulation constant. The control input at this time is the decision of the simulation experiment design scheme.

[0091] In S3, the method for updating the digital twin model is specifically as follows:

[0092] At the time t in the operation stage, collect observation data , and according to the observation data set the digital state and the predicted values of the research quantity and the reward equation are updated to the true values calculated based on the observed data. The expectation-maximization algorithm is used to update the marginal / conditional probability distribution functions in the model, obtaining the joint probability distribution function of the random variables of all nodes in the digital twin model. Then, the predicted values are output through this joint probability distribution function, including the digital states t from time +1 to time the research quantity and the reward equation ;

[0093]

[0094] In the formula, is the joint probability distribution function, is the conditional probability distribution function when the digital state of the ground source heat pump system at the previous time is , and the control input is taken for the ground source heat pump system to obtain the digital state . is the conditional probability distribution function for obtaining the observed data given the digital state . is the conditional probability distribution function for taking the control input at time given the digital state of the ground source heat pump system and the research quantity . This conditional probability distribution function is obtained from the optimal control scheme solved at the previous time.

[0095] In this embodiment, at each moment in the operation stage the digital twin model is updated through Bayes' rule and the principle of conditional independence.

[0096]

[0097] In S3, the specific method for obtaining the optimal operation method of the ground source heat pump system is as follows:

[0098] Based on the predicted values output at time t , a Markov decision process model is constructed to solve and select the optimal control scheme as the optimal operation method of the ground source heat pump system;

[0099] Among them, the expression of the Markov decision process model is specifically:

[0100]

[0101] In the formula, is the reward equation for the operating energy efficiency of the ground source heat pump system, , is the coefficient of performance of the ground source heat pump system at each observation value collection time before the observation value collection time ; is the discount factor for the future time reward, , is to find the operating strategy of the ground source heat pump system that maximizes ; .

[0102] In this embodiment, the state of the Markov decision process is a set of digital states and research quantities. The goal is to solve the optimization problem to maximize the expected cumulative reward within the selected prediction time range, and a control scheme is found using a deep learning algorithm,

[0103]

[0104] In the formula, is the true value of the digital state of the ground source heat pump system before time , is the true value of the research quantity of the ground source heat pump system before time .

[0105] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "center", "thickness", "upper", "lower", "horizontal", "top", "bottom", "inner", "outer", "radial", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the number of technical features. Therefore, the features defined by "first", "second", "third" may explicitly or implicitly include one or more of such features.

Claims

1. An optimization scheduling method for a ground source heat pump system based on a probabilistic graph digital twin model, characterized in that: The following steps are involved: S1. Construct a digital twin model based on a dynamic probability graph decision network in the modeling stage; S2. Calibrate the digital twin model in the calibration phase; S3. Update the digital twin model during the operation phase, and output the optimal operation method of the ground source heat pump system in real time through the updated digital twin model; In S1, the expression of the digital twin model based on the dynamic probability graph decision network is specifically: In the formula, Deadline , given historical observation data and control input, the system exhibits a digital state , calculate the research volume and get rewards The conditional probability distribution function of The digital state of the ground source heat pump system at the last moment is , control input executed , and collect observation data at this moment Update the digital twin to get the digital state The conditional probability distribution function of Given the updated digital state of the digital twin at the current moment Get research volume The conditional probability distribution function of For a given digital state at the current moment , research volume , control input and observational data , get the reward value The conditional probability distribution function of ; For virtual moments, represents the modeling phase, represents the calibration phase of the digital twin model, For prediction and decision-making moments, In order to construct the digital state values ​​of relevant parameter variables of the digital twin calculation model of the ground source heat pump system at different times, It is the observation data reflecting the changes of physical state collected by sensors or detection equipment in the ground source heat pump system at different times. For the operation control scheme of the ground source heat pump system at different times, Digital twin system for ground source heat pump The reward equation for the moment model calibration phase, Digital twin system for ground source heat pump The reward equation for the moment model calibration phase, Digital twin system for ground source heat pump The reward equation for the moment model calibration phase, For digital twin models t The reward equation for the moment model operation phase; In S3, the method for updating the digital twin model is as follows: At the moment of operation t , collect observation data , based on the observed data The digital state , research volume and the reward equation The predicted value of is updated to The calculated true value is used to update the marginal / conditional probability distribution function in the model using the expectation maximization algorithm to obtain the joint probability distribution function of the random variables of all nodes in the digital twin model, and then the predicted value is output through the joint probability distribution function, including the time t +1 to time Digital Status , research volume and the reward equation ; In the formula, is the joint probability distribution function, The digital state of the ground source heat pump system at the last moment is , control input for ground source heat pump system Then get the digital state The conditional probability distribution function of For a given digital state , get the observed data The conditional probability distribution function of For at the moment Given the digital status of the ground source heat pump system and research volume , taking control input The conditional probability distribution function of , which is the optimal control solution solved at the previous moment.

2. The method for optimizing and scheduling a ground source heat pump system based on a probabilistic graph digital twin model according to claim 1, characterized in that: In S1, the nodes in the digital twin model based on the dynamic probability graph decision network include physical states, digital states, observation data, research quantities, control inputs, and random variables of reward equations at each time point.

3. The method for optimizing and scheduling a ground source heat pump system based on a probabilistic graph digital twin model according to claim 2, characterized in that: A collection of physical states , To reflect the values ​​of relevant parameter variables of the ground source heat pump system operating status at different times; A collection of digital states ; A collection of observation data ; A collection of research ,in, is the coefficient of performance of the ground source heat pump system at different times; In the formula, For Time has come The average value of the heat at the load end at all times, For Time has come The average value of the input power of the heat pump unit at the moment, For Time has come The average value of the pump's input power at the moment, For Time has come The average power consumption of the fan coil at the load end at all times. For Time has come Control the average value of system energy consumption at all times; The set of control inputs is ; The set of reward equations is ,in, Digital twin system for ground source heat pump The reward equation during the moment-to-moment model calibration phase; In the formula, is the first duration of the model calibration phase. and are the observed values ​​of the load at each moment and the predicted values ​​of the digital twin model, s Collect moments for observations; Digital twin system for ground source heat pump The reward equation during the moment-to-moment model calibration phase; In the formula, The second duration of the model calibration phase, and are the observed value of the outlet temperature at the ground source side at each moment and the predicted value of the digital twin model; Digital twin system for ground source heat pump The reward equation during the moment-to-moment model calibration phase; In the formula, The third duration of the model calibration phase. and They are the observed value of the return water temperature at the load end at each moment and the predicted value of the digital twin model; For digital twin models t The reward equation for the moment model operation phase; In the formula, is the coefficient of performance of the ground source heat pump system at each observation collection moment.

4. The method for optimizing and scheduling a ground source heat pump system based on a probabilistic graph digital twin model according to claim 3, characterized in that: In S2, the digital twin model includes the load-side model, the ground-source side model, and the ground-source heat pump unit model; exist Always calibrate the load end model and related parameters to minimize As the target, adjust the load end model; Calibrate the ground source model and related parameters at all times to minimize As the target, adjust the ground source side model; Calibrate the ground source heat pump unit model and related parameters at all times to minimize As the goal, adjust the ground source heat pump unit model.

5. The method for optimizing and scheduling a ground source heat pump system based on a probabilistic graph digital twin model according to claim 3, characterized in that: In S3, the optimal operation method of the ground source heat pump system is obtained as follows: Based on time t The predicted value of the output is used to construct a Markov decision process model to solve and select the optimal control scheme as the optimal operation method of the ground source heat pump system; Among them, the Markov decision process model The specific expression is: In the formula, The reward equation for the energy efficiency of the ground source heat pump system is: , is the time at which the observations are collected The previous observation collection time The coefficient of performance of the ground source heat pump system, is the discount factor for future rewards, , To find the maximum Operation strategy of ground source heat pump system .

Images