A method and system for linkage regulation and control of working parameters of a medium-deep geothermal water source heat pump

Abstract

The present application relates to a kind of medium deep geothermal water source heat pump's working parameter linkage control method and system, specifically relates to medium deep geothermal water source heat pump field, first, disturbance signal is generated by reinforcement learning and is injected into geothermal circuit, to improve the dynamic mapping precision of underground thermal reservoir;Second, the thermal diffusivity and fracture ratio of rock formation are inversed by physical information neural network model, to realize the real-time monitoring of underground thermal reservoir;Then, the difference between heating load and geothermal heat absorption capacity is predicted using meteorological data, room temperature data and heat pump operating parameters, control decisions are generated and hierarchical trigger conditions are set;Finally, the phase change heat storage device and compressor are accurately controlled by double-actuator cooperative algorithm, to maximize the response speed and energy efficiency of heat pump system, to alleviate room temperature fluctuation problem, to avoid frequent frequency modulation and to reduce the start of standby boiler, to optimize the operating efficiency and stability of heating system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medium-deep geothermal water source heat pump, more particularly, the present application relates to a medium-deep geothermal water source heat pump working parameter linkage control method and system. BACKGROUND

[0002] Under the extreme climate conditions such as cold wave, the challenge faced by the heating system increases significantly, especially when the winter weather strikes, the regional heating load often rises sharply in a short time, for example, when the cold wave comes, the external temperature drops sharply, and the regional heating load may increase by 50% in 2 hours, this sudden load change requires the heating system to have enough response ability, but due to the high thermal inertia of the geothermal system, the heat recovery rate is relatively slow, so the geothermal well cannot provide enough heat in a short time, in addition, the response lag of the geothermal system also affects the operating efficiency of the heat pump unit, due to insufficient heat absorption, the heat pump unit has to run at a lower frequency, which leads to the user's room temperature fluctuation exceeding the standard, affecting the user's comfort, external factors such as rapid changes in weather, combined with internal factors such as the thermal inertia of the geothermal system not being fully considered, form a contradiction in the design of the heating system.

[0003] The current heating system design usually relies on weather forecasts to start the unit in advance to cope with temperature changes, however, the error of weather forecast is usually large, especially for short-term extreme weather prediction, the error can exceed 2℃, it is difficult to accurately predict the sudden load change, at the same time, the existing design often only considers the temperature factor, ignoring the thermal inertia of the building and the response characteristics of the geothermal heat storage, which makes the system react slowly when facing short-term load surge, and cannot quickly adapt to demand changes, in addition, due to the lack of real-time monitoring of key geological parameters such as soil thermal conductivity and rock fracture rate that affect geothermal recovery, the heat recovery capacity of the geothermal system cannot be accurately evaluated, the existing technology fails to achieve efficient coupling between weather, geology and unit, and fails to quickly and accurately respond to heating load changes through precise pre-compensation algorithm, resulting in room temperature fluctuation, frequent frequency adjustment of the unit and start-up of the standby gas boiler, further increasing carbon emissions and operating costs. SUMMARY

[0004] The present application provides a medium-deep geothermal water source heat pump working parameter linkage control method and system to solve the problems in the above background technology through geological disturbance injection-physical constraint inversion-dynamic threshold prediction-double actuator coordination technology.

[0005] The technical solution of the present application to solve the above technical problems is as follows, a medium-deep geothermal water source heat pump working parameter linkage control method, specifically comprising the following steps:

[0006] Step S1, when monitoring the underground temperature fluctuations of the geothermal system or external meteorological mutations, triggering the disturbance signal injection step, generating the optimal disturbance signal based on the reinforcement learning algorithm and injecting it into the geothermal water circuit, the core data operation object includes flow and pressure data, the feature processing means is dynamic adjustment of flow step amplitude and duration, to produce a response data set to maximize the sensor information entropy and establish the underground heat reservoir dynamic mapping;

[0007] Step S2, after obtaining the response data set, triggering the geology parameter inversion step, the core data operation object is pressure and temperature response data, the feature processing means is a physical information neural network model, to produce the rock layer thermal diffusivity and fracture rate inversion results, to construct the underground heat response real-time image and quantify the thermal inertia effect;

[0008] Step S3, after obtaining the geology parameter inversion result, triggering the load difference prediction step, the core data operation object is meteorological data, user room temperature data and heat pump operation parameters, the feature processing means is a time series prediction model and a thermal decay model, to produce a predicted difference between the heating load and the geothermal heat absorption capacity, and generate a control decision when the predicted difference exceeds a dynamic threshold, to predict the supply-demand conflict and set a hierarchical trigger condition;

[0009] Step S4, after generating the control decision, triggering the linkage control step, the core data operation object is the predicted difference data and the heat pump actuator parameters, the feature processing means is a double-actuator collaborative algorithm, to produce the phase change heat storage device valve opening control instruction and the compressor frequency adjustment value, to execute the pre-compensation control and finally optimize the heat pump system energy efficiency;

[0010] In a preferred embodiment, in step S1, the specific operation of dynamically adjusting the flow step amplitude and duration is:

[0011] When the absolute value of the difference between the real-time temperature data at the wellhead and the average well temperature at the same period in the same area exceeds the temperature fluctuation threshold, or the absolute value of the air temperature change rate calculated by the meteorological radar data exceeds the meteorological warning threshold, the disturbance strategy generation mechanism is triggered; based on the near-optimal policy optimization algorithm in reinforcement learning, the optimal combination parameters are searched within the preset flow step amplitude range and disturbance duration range, and the optimization objective of the search process is to maximize the weighted sum of the information entropy of all sensor data and the physical rule constraint term; wherein the physical rule constraint term is calculated by the thermal diffusivity and underground water flow rate in the prior geology parameter set; the generated optimal flow step amplitude and duration are injected into the geothermal water circuit through the frequency conversion pump, and the pressure data and temperature data are collected synchronously at a frequency of once per second to form the heat reservoir feature response data set.

[0012] In a preferred embodiment, the specific construction of the physical rule constraint term is:

[0013] The mean square error of the temperature versus depth partial derivative value and the thermal dispersion equation theoretical value is calculated, wherein the thermal dispersion equation theoretical value is obtained by dividing the a priori thermal diffusion coefficient by the groundwater flow rate and then multiplying the second-order partial derivative of the temperature versus time; the information entropy calculation covers the real-time reading distribution probability of the pressure sensor and the temperature sensor, and the distribution probability is obtained by simulating the data distribution under different perturbation strategies through Monte Carlo sampling; the search space of the optimal combination parameters is limited by the rated flow rate ratio of the heat pump, and the lower limit of the flow rate step amplitude is 10% of the rated flow rate, and the upper limit is 30%, and the lower limit of the perturbation duration is 5 minutes, and the upper limit is 30 minutes.

[0014] In a preferred embodiment, in the step S2, the feature processing operation of the physical information neural network model is:

[0015] Firstly, the response data set output by the step S1 is subjected to spatio-temporal alignment and noise reduction processing, and a pressure data matrix and a temperature data matrix are extracted as core inputs; an axial temperature gradient field is calculated based on the temperature data matrix along the sensor burial depth direction; a ternary loss function including a heat conduction constraint term, a Darcy flow constraint term and a data fitting term is constructed; wherein the Darcy flow constraint term introduces a permeability function, which quantifies the topological structure of the rock fracture network through an exponential decay relationship constructed by a fracture connectivity index and a critical fracture density value, and the fracture connectivity index is calculated according to the a priori geological parameter set delivered by the step S1; an adaptive Monte Carlo sampling algorithm is used to iteratively optimize the network parameters within a preset constraint range, and the constraint range is set according to the upper and lower floating thresholds of the thermal diffusion coefficient and the fracture density in the a priori geological parameter set delivered by the step S1; finally, a parameter set including the thermal diffusion coefficient inversion value and the fracture density inversion value is output, and a corresponding confidence evaluation matrix is generated, which is used to judge the effectiveness of the inversion result, and if the fracture density confidence is lower than 0.7, the step S1 is returned to start additional perturbation.

[0016] In a preferred embodiment, the construction process of the confidence evaluation matrix is:

[0017] According to the modulus characteristic value of the second-order derivative matrix of the network parameters of the total loss function of the physical information neural network model, a negative exponential mapping function is applied to calculate the confidence scalar of the thermal diffusion coefficient and the fracture density inversion value;

[0018] The operation of the Darcy flow constraint term is to minimize the sum of squares of errors between the measured pressure gradient data of the sensor array and the theoretical pressure gradient data calculated based on the fluid dynamic viscosity, the groundwater flow rate and the permeability function as an optimization sub-objective; wherein the permeability function is constructed depending on the fracture connectivity index and the critical fracture density value, the fracture connectivity index calls the pre-calibrated parameters of the regional geological database, and the critical fracture density value is set to 0.18;

[0019] The iteration termination condition of the adaptive Monte Carlo sampling algorithm is to stop optimization when the Frobenius norm of the Hessian matrix of the total loss function with respect to the network parameters is lower than a preset convergence threshold.

[0020] In a preferred embodiment, in step 3, the load difference prediction step is specifically:

[0021] Firstly, based on the geothermal thermal diffusion coefficient and fracture porosity parameters obtained by inversion in step S2, the geothermal thermal recovery time constant is calculated, the value of which is obtained through a combined expression of rock density, specific heat capacity, wellbore radius and fracture lag gain, wherein the fracture lag gain is set according to the calibration range of high temperature rheological experiment of granite; secondly, a bidirectional thermal decay function of building thermal inertia and geothermal delay is constructed, the function structure of which includes building heat capacity and surface convection coefficient calculation item, permeability related heat conduction efficiency item and frequency domain convolution operation item; then the long short-term memory network unit is improved, and the geological feature vector composed of thermal diffusion coefficient, fracture porosity and geothermal thermal recovery time constant is injected to perform dynamic prediction of heating load; finally, combined with the maximum absolute value of air temperature change rate, a dynamic threshold is generated through the geothermal lag sensitive factor and the weight coefficient, when the ratio of the predicted difference and the dynamic threshold meets the preset grading condition, the grading mark of the regulation and control decision is output to trigger the subsequent linkage control.

[0022] In a preferred embodiment, the frequency domain convolution operation of the bidirectional thermal decay function is specifically:

[0023] The product operation is performed after the fast Fourier transform on the building thermal inertia index decay term and the geothermal delay index decay term;

[0024] The injection mode of the geological feature vector in the long short-term memory network is to increase the three-dimensional vector splicing operation in the input gate, the forget gate and the output gate calculation formula; the weight coefficient of the dynamic threshold generation formula is obtained by reinforcement learning in the historical data of the heating season, and the geothermal lag sensitive factor is selected according to the rock type in the preset interval;

[0025] The grading trigger condition is defined as: when the predicted load difference exceeds fifteen percent but is less than twenty-five percent of the dynamic threshold, the first level regulation is started, and when it exceeds twenty-five percent, the second level regulation is started, and at the same time, the geothermal thermal recovery time constant is transmitted to the lag compensator optimization module.

[0026] In a preferred embodiment, in step S4, the linkage regulation step is specifically:

[0027] Firstly, a phase lead compensator is constructed based on the regulation decision grade flag and the geothermal heat recovery time constant output in step S3, and a time domain compensation quantity is generated by setting a phase lead gain and a differential time constant; then, a dual actuator collaborative control is executed, the phase change heat storage device valve opening control is activated when the decision grade flag is level one, and the compressor frequency adjustment is superimposed and started when the decision grade flag is level two, the control law contains a proportional term, an integral term and a feedforward term based on the user room temperature change rate, wherein the compressor adjustment value and the valve opening degree are dynamically associated through a coupling coefficient; finally, a dynamic amplitude limiting constraint is applied, the valve opening degree is forcibly limited in the range of zero to one hundred percent, and the compressor frequency is adaptively limited according to the geothermal heat recovery time constant in a negative exponential rule, so that the actuator action is within the physical safety boundary.

[0028] In a preferred embodiment, the compensation time quantity of the phase lead compensator is obtained by multiplying a fixed proportional coefficient by the geothermal heat recovery time constant, and the proportional coefficient is set according to the calibration range of the high temperature rheological experiment of granite;

[0029] The coupling activation condition of the dual actuator collaborative control is defined as being effective only when the decision grade flag value is greater than or equal to one; and the weight parameters of the proportional term, the integral term and the feedforward term are determined by reinforcement learning in the heating history data;

[0030] The upper limit of the compressor frequency limiting function decreases with the increase of the heat recovery time constant, and specifically, the maximum allowed frequency is multiplied by a negative exponential decay function.

[0031] The application also provides a working parameter linkage control system of a middle-deep geothermal water source heat pump, which specifically comprises: a disturbance signal injection module, a geological parameter inversion module, a load difference prediction module and a linkage control execution module;

[0032] The disturbance signal injection module: when the underground temperature sensor detects that the geothermal well water temperature deviates from the historical same period average value by more than a set threshold value or the weather radar detects that the air temperature change rate exceeds a warning value, a flow step disturbance signal is generated and injected into the geothermal water circuit through a reinforcement learning algorithm, and the flow amplitude and duration are dynamically adjusted by a variable frequency water pump as the core control means, a multi-sensor response data set containing pressure and temperature spatio-temporal distribution is generated, and a dynamic mapping atlas of the underground heat reservoir is established;

[0033] The geological parameter inversion module: after receiving the response data set, a physical information neural network model is used to process the pressure gradient field and temperature gradient field data, and the rock layer thermal diffusivity and fracture rate are output by joint inversion of heat conduction equation constraint and Darcy flow constraint, and an inversion result confidence evaluation matrix is generated;

[0034] The load difference prediction module: after obtaining the set of geological parameters, the meteorological forecast data and the real-time user room temperature data are fused, the time series prediction model of the rock thermal attenuation characteristic is fused, the dynamic difference between the building heating load and the geothermal heat absorption capacity is calculated, and when the difference value caused by the geothermal heat recovery hysteresis effect exceeds the adaptive grading threshold, the control decision instruction containing the regulation level identifier is generated;

[0035] The linkage regulation execution module: after receiving the control decision instruction, the predicted difference data and the heat pump working condition parameters are analyzed based on the double-actuator collaborative algorithm, the phase advance compensation mechanism is used to output the phase change heat storage device valve opening degree instruction and the compressor frequency correction value, and the dynamic pre-compensation of the geothermal response hysteresis and the system energy efficiency optimization control are realized.

[0036] The method has the advantages that: the method realizes effective relief of the contradiction between the heating load mutation under extreme climate and the geothermal response hysteresis through multi-step linkage regulation; first, the disturbance signal is generated through reinforcement learning and injected into the geothermal loop, so as to improve the dynamic mapping accuracy of the underground heat reservoir; second, the thermal diffusion coefficient and the fracture rate of the rock layer are inversed through the physical information neural network model, so as to realize real-time monitoring of the underground heat reservoir; then, the difference between the heating load and the geothermal heat absorption capacity is predicted by using meteorological data, room temperature data and heat pump operation parameters, a regulation decision is generated, and a grading trigger condition is set; finally, the phase change heat storage device and the compressor are accurately regulated through the double-actuator collaborative algorithm, the response speed and the energy efficiency of the heat pump system are maximized, the room temperature fluctuation problem is relieved, the frequency modulation is avoided, and the standby boiler is reduced, so that the operation efficiency and the stability of the heating system are optimized. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The method flowchart of the present application is shown in the figure;

[0038] Figure 2 The system structure block diagram of the present application is shown in the figure. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0040] In the description of the present application, the terms "first", "second" are only for descriptive purposes and cannot be understood to indicate or imply relative importance or implicitly indicate the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise explicitly specified.

[0041] In the description of the present application, the term "for example" is used to indicate "as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given in order to enable any person skilled in the art to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that those skilled in the art can recognize that the present application can be implemented without using these specific details. In other examples, well-known structures and processes will not be described in detail to avoid unnecessary details making the description of the present application obscure. Therefore, the present application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope in accordance with the principles and characteristics disclosed in the present application.

[0042] Embodiment 1

[0043] The present embodiment provides a working parameter linkage control method for a middle-deep geothermal water source heat pump as shown in Figure 1 The present embodiment provides a working parameter linkage control method for a middle-deep geothermal water source heat pump as shown in

[0044] Step S1, when the underground temperature fluctuation of the geothermal system or external meteorological mutation is monitored, a disturbance signal injection step is triggered, an optimal disturbance signal is generated based on a reinforcement learning algorithm and injected into the geothermal water circuit, the core data operation object includes flow and pressure data, the feature processing means is dynamic adjustment of flow step amplitude and duration, a response data set is generated to maximize sensor information entropy and establish an underground thermal reservoir dynamic mapping, this step uses active excitation of the characteristic response of the underground rock layer to obtain high information density observation data through controllable disturbance;

[0045] Step S2, after obtaining the response data set, a geology parameter inversion step is triggered, the core data operation object is pressure and temperature response data, the feature processing means is a physical information neural network model, the rock layer thermal diffusivity and fracture rate inversion results are generated to construct an underground thermal response real-time image and quantify thermal inertia effect, this step inverts key parameters of the underground rock layer through deep learning with physical constraints, solves the defects of traditional methods relying on steady-state assumption and ignoring geological dynamic response, and simultaneously fuses the thermal-hydraulic coupling equation and the fracture network topological constraint to realize high-precision parameter identification under non-steady-state disturbance;

[0046] Step S3, after obtaining the geological parameter inversion result, trigger the load difference prediction step, the core data operation object is meteorological data, user room temperature data and heat pump operation parameters, the characteristic processing means is the time series prediction model and the heat attenuation model, the predicted difference of heating load and geothermal heat absorption capacity is generated, its expression is: , represents the heating load, represents the geothermal heat absorption capacity, and , wherein, represents the permeability function (based on the crack rate calculation of step S2 inversion), represents the inverted crack rate, represents the effective heat transfer area, the product of the wellbore cross-sectional area and the heat exchange depth: , represents the wellbore radius, represents the characteristic depth, that is, the length of the effective section of the temperature gradient, which is the wellbore sensor spacing: [5, 50] m, represents the rock temperature gradient, the vertical temperature difference of the geothermal well: , represents the fluid dynamic viscosity, that is, the flow resistance of the geothermal fluid, the value is: pure water (90℃): ; high salinity water: In addition, The higher the permeability, the greater the temperature difference, the lower the viscosity, the stronger the geothermal energy supply, and when the predicted difference exceeds the dynamic threshold, a control decision is generated to predict the supply-demand conflict and set a hierarchical trigger condition;

[0047] Step S4, after generating the control decision, trigger the linkage control step, the core data operation object is the predicted difference data and the heat pump actuator parameters, the characteristic processing means is the double actuator collaborative algorithm, the phase change heat storage device valve opening control instruction and the compressor frequency adjustment value are generated to execute the pre-compensation control and relieve the geothermal response lag problem, and finally optimize the energy efficiency of the heat pump system.

[0048] In this embodiment, it is specifically necessary to explain step S1, the specific operation of dynamically adjusting the flow step amplitude and duration is:

[0049] The wellhead temperature sensor array (satisfying the precision level) is used to collect real-time geothermal water loop outlet temperature data, and the prediction data provided by the meteorological department is transmitted to the central control system through the optical cable, the temperature sensor selects the platinum resistance temperature measuring element, when the absolute value of the difference between the real-time temperature data of the wellhead and the average well temperature of the same period in the same area is greater than the temperature fluctuation threshold, or the absolute value of the air temperature change rate calculated by the weather radar data is greater than the meteorological warning threshold, the disturbance strategy generation mechanism is triggered, and the trigger condition expression is:

[0050] ;

[0051] in, This indicates the actual measured temperature at the wellhead at the current moment, i.e., the real-time temperature sensor reading at the geothermal well outlet. This represents the average well temperature for the same historical period in the same location, i.e., the average well temperature during periods of similar climatic conditions calculated based on historical databases. This represents the temperature fluctuation threshold, a critical value used to determine abnormal temperature changes; its set value is 3℃. This indicates the ambient temperature predicted by weather radar, specifically the predicted atmospheric temperature for the next hour. This indicates the predicted rate of temperature change, that is, the trend of temperature change per unit time. The threshold value for temperature change rate is used to determine the critical rate of change for sudden meteorological changes. Its set value is 2℃ / h. The triggering logic of this triggering condition formula is to activate the perturbation strategy generation mechanism when the actual well temperature deviates from the historical average by more than the threshold value or the predicted temperature change rate exceeds the threshold value. Based on the near-end policy optimization algorithm in reinforcement learning, its formula is:

[0052] ;

[0053] in, This represents the flow step amplitude, i.e., the change in flow rate (relative to the rated flow rate) during a disturbance, and its value range is... , This indicates the duration of the disturbance, i.e., the duration of the flow step, and its value ranges from [value missing]. , This represents the a priori thermal diffusivity coefficient, taken from the thermal diffusivity parameter of rock strata in a historical database, and its value range is... , This indicates the fluid velocity in the wellbore, the speed at which groundwater flows within the wellbore. The physical constraint weight coefficient is a hyperparameter used to balance information entropy and the importance of physical rules. This indicates the total number of sensors, specifically the number of sensors deployed at the wellhead and in the pipeline network. Indicates the first The measurement status of each sensor, including temperature and pressure Composite data, This represents the probability distribution of sensor states, i.e., generated by Monte Carlo simulation. Probability of occurrence It represents the partial derivative of temperature with respect to depth, and the temperature gradient along the well depth direction (reflecting the thermal conductivity of the rock formation). represents the second-order partial derivative of temperature with respect to time, reflecting the instantaneous acceleration of temperature change (capturing thermal inertia effects), searching for the optimal combination of parameters within the preset flow step amplitude range and the disturbance duration range, and the optimization objective of the search process is to maximize the weighted sum of the information entropy of all sensor data and the physical rule constraint term; wherein the physical rule constraint term is calculated by the thermal diffusion coefficient and the groundwater flow rate in the prior geological parameter set, ensuring that the disturbance signal conforms to the law of underground heat conduction; the generated optimal flow step amplitude and duration are injected into the geothermal water circuit through a variable frequency pump, and pressure data and temperature data are collected synchronously at a frequency of once per second using a pressure transmitter array (distributed at the inlet and outlet of the water source heat pump, the geothermal well casing) and a temperature sensor group (arranged every fifty meters along the shaft axis), forming a thermal reservoir characteristic response data set; in addition, the data integration and transmission operation specifically comprises: all sensor data are transmitted to the edge computing gateway through the industrial field bus, and after time stamp alignment and data verification are completed, they are packaged into a standardized response data set; and the data set is stored as a three-dimensional tensor: the first dimension is the sensor number, the second dimension is the time sequence, and the third dimension contains the numerical values of the three types of physical quantities of flow, pressure and temperature;

[0054] The specific construction method of the physical rule constraint term is:

[0055] The mean square error of the temperature derivative with respect to depth and the theoretical value of the thermal dispersion equation is calculated, wherein the theoretical value of the thermal dispersion equation is obtained by dividing the prior thermal diffusion coefficient by the groundwater flow rate and then multiplying the second-order derivative of temperature with respect to time; the information entropy calculation covers the real-time reading distribution probability of the pressure sensor and the temperature sensor, and the distribution probability is obtained by simulating the data distribution under different disturbance strategies through Monte Carlo sampling; the search space of the optimal combination of parameters is limited by the rated flow ratio of the heat pump, the lower limit of the flow step amplitude is 10% of the rated flow, and the upper limit is 30%, the lower limit of the disturbance duration is five minutes, and the upper limit is thirty minutes.

[0056] In this embodiment, it is specifically necessary to explain step S2, the feature processing operation of the physical information neural network model is:

[0057] Firstly, the response data set output by step S1 is subjected to spatio-temporal alignment and noise reduction processing, and the pressure data matrix and the temperature data matrix are extracted as the core input; the axial temperature gradient field is calculated based on the temperature data matrix along the sensor depth direction; a three-element loss function including a heat conduction constraint term, a Darcy flow constraint term and a data fitting term is constructed, and its expression is:

[0058] ;

[0059] Among them, represents the loss term weight coefficient, and its value range is , represents the predicted temperature field, represents the thermal diffusivity, and the value range of the thermal diffusivity is , represents the Laplacian of the temperature field, represents the predicted pressure field, represents the fluid dynamic viscosity, and the value range of the fluid dynamic viscosity is , represents the permeability function, and the value range of the permeability function is , represents the groundwater flow velocity vector, and the value range of the groundwater flow velocity vector is , represents the measured temperature vector; wherein the Darcy flow constraint term introduces the permeability function, and the expression of the permeability function is:

[0060] ;

[0061] wherein, represents the reference permeability, and the value range of the reference permeability is , represents the fracture connectivity index, and the value range of the fracture connectivity index is , represents the fracture ratio, and the value range of the fracture ratio is , represents the critical value of the fracture ratio, and the value of the critical value of the fracture ratio is 0.18, and the formula represents that when , the permeability drops sharply, and the function quantifies the topological structure of the fracture network of the rock formation by constructing an exponential decay relationship through the fracture connectivity index and the critical value of the fracture ratio, and the fracture connectivity index is calculated according to the set of prior geological parameters transmitted in step S1; the adaptive Monte Carlo sampling algorithm is used to iteratively optimize the network parameters within a preset constraint range, and the constraint range is set according to the thermal diffusivity and the upper and lower floating threshold of the fracture ratio in the set of prior geological parameters transmitted in step S1; finally, a parameter set containing the inverted value of the thermal diffusivity and the inverted value of the fracture ratio is output, and a corresponding confidence evaluation matrix is generated, and the confidence evaluation matrix is used to judge the effectiveness of the inversion result, and if the confidence of the fracture ratio is lower than 0.7, the step S1 is returned to start additional disturbance;

[0062] The construction process of the confidence evaluation matrix is:

[0063] According to the modulus eigenvalue of the second-order derivative matrix of the network parameters of the total loss function of the physical information neural network model, the confidence scalar of the inverted value of the thermal diffusivity and the fracture ratio is calculated by applying a negative exponential mapping function, and the expression is:

[0064] ;

[0065] wherein, represents the Hessian matrix, represents the Frobenius norm (convergence threshold ), denotes the confidence scalar, whose value range is ;

[0066] The operation of the Darcy flow constraint term is to minimize the error sum of squares between the pressure gradient data measured by the sensor array and the theoretical pressure gradient data calculated based on the fluid dynamic viscosity, groundwater flow rate and permeability function as the optimization sub-objective; wherein the permeability function is constructed depending on the fracture connectivity index and the fracture rate critical value, the fracture connectivity index calls the pre-calibration parameters of the regional geological database, and the fracture rate critical value is set to 0.18;

[0067] The iteration termination condition of the adaptive Monte Carlo sampling algorithm is: when the Frobenius norm of the Hessian matrix of the total loss function with respect to the network parameters is lower than the preset convergence threshold (Frobenius norm of the Hessian matrix of the total loss function with respect to the network parameters is lower than the preset convergence threshold ) stop optimization to ensure that the inversion result meets the physical law constraint and numerical stability requirement at the same time.

[0068] In this embodiment, it is particularly necessary to explain that the load difference prediction step S3 is specifically:

[0069] First, based on the geothermal heat diffusion coefficient and fracture rate parameters obtained by inversion in step S2, the geothermal heat recovery time constant is calculated, the value of which is obtained through a combined expression of rock density, specific heat capacity, wellbore radius and fracture lag gain, wherein the fracture lag gain is set according to the calibration range of granite high temperature rheological experiment, and the calculation formula of the geothermal heat recovery time constant is:

[0070] ;

[0071] wherein, denotes the rock density (granite: 2700 kg / m³), which quantifies the heat storage capacity of the rock layer, denotes the rock specific heat capacity, denotes the wellbore radius, whose value range is 0.1-0.5 m, which is used to control the heat recovery space scale, denotes the inversion heat diffusion coefficient, whose value range is , denotes the fracture lag gain, which is used to amplify the influence of fracture rate, whose value range is , denotes the fracture connectivity index, whose value range is [5, 20], denotes the inversion fracture rate, whose value range is , 0.18 represents the critical value of fracture rate; secondly, the bidirectional thermal decay function of building thermal inertia and geothermal delay is constructed, the function structure contains the calculation items of building heat capacity and surface convection coefficient, the heat conduction efficiency item related to permeability and the frequency domain convolution operation item; then the long short-term memory network unit is improved, and the geological feature vector composed of thermal diffusivity, fracture rate and geothermal heat recovery time constant is injected, and its expression is:

[0072] ;

[0073] wherein, represents the geological feature vector, represents the inversion of thermal diffusivity, and its value range is , represents the inversion of fracture rate, and its value range is , represents the geothermal heat recovery time constant, and its value range is [0.5, 12]h, represents the LSTM hidden state vector, that is, the memory history of heating law, represents the current time input data, which is used to perceive the current heating demand and contains air temperature, room temperature and heat pump parameters, and is used for dynamic prediction of heating load; finally, the maximum absolute value of air temperature change rate is combined to generate a dynamic threshold through the geothermal delay sensitive factor and the weight coefficient, and its expression is:

[0074] ;

[0075] wherein, represents the maximum air temperature change rate, which is used to capture the intensity of cold wave, and 0.6 represents the meteorological mutation weight coefficient, which is used to strengthen the influence of air temperature sudden change, represents the geothermal delay decay term, The greater the value is, the lower the threshold is, and 1.2 represents the geothermal weight coefficient, which amplifies the influence on the threshold, when the ratio of the prediction difference and the dynamic threshold meets the preset grading condition, the grading mark of the control decision is output to trigger the subsequent linkage control;

[0076] The frequency domain convolution operation of the bidirectional thermal decay function is specifically:

[0077] The product operation is performed on the building thermal inertia index decay term and the geothermal delay index decay term after fast Fourier transform, and its expression is:

[0078] ;

[0079] wherein, represents fast Fourier transform (FFT), which is used to convert time domain decay into frequency domain signal, represents inverse Fourier transform, i.e. back to time domain executable data, represents building thermal efficiency factor, and , represents building thermal inertia constant, which is in the range of [2, 8]h (including thermal insulation buildings), represents geothermal conduction efficiency factor, and , represents geothermal heat recovery time constant, represents FFT of building thermal inertia index decay term, represents FFT of geothermal delay index decay term, represents inverse transform after frequency domain multiplication operation;

[0080] The injection mode of the geology feature vector in the long short-term memory network is to increase a three-dimensional vector splicing operation in the input gate, the forgetting gate and the output gate calculation formula; the weight coefficient of the dynamic threshold generation formula is obtained by reinforcement learning optimization in the historical data of the heating season, and the geothermal delay sensitive factor is selected in a preset interval according to the rock type;

[0081] The hierarchical triggering condition is defined as follows: when the predicted load difference exceeds the dynamic threshold of 15% but is less than 25%, the first-stage regulation is started (the phase change heat storage device is activated), and the expression is as follows: , represents the load prediction difference, i.e. the gap between the heating demand and the geothermal energy supply, represents a dynamic threshold, 0.15, 0.25 represents a hierarchical proportion coefficient, i.e. the first / second regulation boundary is set, and when the gap exceeds 25%, the second regulation (the frequency of the compressor is increased) is started, and the expression is as follows At the same time, the geothermal heat recovery time constant is transmitted to the lag compensator optimization module.

[0082] In the embodiment, the step S4 needs to be specifically explained, and the linkage regulation step is specifically as follows:

[0083] Firstly, based on the regulation decision hierarchical flag and the geothermal heat recovery time constant output in the step S3, a phase lead compensator is constructed, a time domain compensation amount is generated by setting a phase lead gain and a differential time constant, and the specific formula of the phase lead compensator is as follows:

[0084] ;

[0085] Among them, represents the compensated control instruction, represents the original valve opening instruction, represents the phase lead time, which is in the range of , and the calculation method is as follows: , denotes the geothermal heat recovery time constant, whose value range is [0.5, 12] h, 0.7 denotes the phase advance gain coefficient, which advances the execution time of the control instruction to offset the hysteresis effect of the geothermal response; then the double-actuator collaborative control is executed, when the decision classification flag is level one, the phase change heat storage device valve opening control is activated, when the decision classification flag is level two, the compressor frequency adjustment is superimposed and started, the control law contains a proportional term, an integral term and a feedforward term based on the user room temperature change rate, wherein the compressor adjustment value and the valve opening degree are dynamically associated through a coupling coefficient, and its expression is:

[0086] ;

[0087] wherein, denotes the phase change heat storage valve opening degree, denotes the compressor frequency adjustment value, denotes the load prediction error at the moment, denotes the historical load error integral, denotes the user room temperature change amount, denotes the sampling period, denotes the geothermal water mass flow rate, and 4180 denotes the specific heat capacity of water, denotes the geothermal water inlet temperature safety margin, denotes the comprehensive efficiency coefficient, and 0.4 denotes the actuator coupling coefficient, whose value range is [0.3, 0.6], denotes the adjustment decision activation function, whose value range is , denotes the regulation decision classification flag, whose value range is ; finally, a dynamic limiting constraint is applied, the valve opening degree is forcibly limited in the range of zero to one hundred percent, and the compressor frequency is adaptively limited according to the geothermal heat recovery time constant in the form of a negative exponential rule, and its expression is:

[0088] ;

[0089] wherein, denotes the valve opening degree boundary constraint, whose value range is , and 60 denotes the maximum allowed frequency of the compressor, denotes the geothermal adaptive attenuation coefficient, whose value range is , denotes the geothermal heat recovery time constant, whose value range is , and 0.3 denotes the compressor frequency attenuation gain coefficient, which is determined by 10 years of historical data regression analysis, so that the actuator action is within the physical safety boundary;

[0090] The compensation time of the phase lead compensator is obtained by multiplying a fixed proportional coefficient by the geothermal thermal recovery time constant. The proportional coefficient is set according to the calibration range of the high-temperature rheological test of granite.

[0091] The coupling activation condition for dual-actuator cooperative control is defined as taking effect only when the decision level flag value is greater than or equal to one; the weight parameters of the proportional term, integral term, and feedforward term are optimized and determined from heating history data through reinforcement learning;

[0092] The upper bound of the compressor frequency limiting function decreases as the thermal recovery time constant increases. Specifically, it is the product of the maximum allowable frequency and the negative exponential decay function, in order to prevent the heat pump from shutting down due to overload and to improve system stability.

[0093] Example 2

[0094] This embodiment provides, for example Figure 2 The present invention relates to a working parameter linkage control system for a medium-deep geothermal water source heat pump, which specifically includes: a disturbance signal injection module, a geological parameter inversion module, a load difference prediction module, and a linkage control execution module;

[0095] Disturbance signal injection module: When the underground temperature sensor detects that the temperature of the geothermal well water deviates from the historical average value by more than a set threshold or the weather radar detects that the rate of change of air temperature exceeds the warning value, the module generates a flow step disturbance signal through reinforcement learning algorithm and injects it into the geothermal water circuit. The variable frequency pump dynamically adjusts the flow amplitude and duration as the core control means, generates a multi-sensor response dataset containing the spatiotemporal distribution of pressure and temperature, and establishes a dynamic mapping map of the underground thermal reservoir.

[0096] Geological parameter inversion module: After receiving the response dataset, the module uses a physical information neural network model to process the pressure gradient field and temperature gradient field data. It performs joint inversion by constraining the heat conduction equation and Darcy flow, outputting a set of geological parameters consisting of the thermal diffusivity and fracture rate of the rock strata, and generating a confidence assessment matrix for the inversion results.

[0097] Load difference prediction module: After acquiring the set of geological parameters, it integrates meteorological forecast data and real-time user room temperature data, and calculates the dynamic difference between building heating load and geothermal heat absorption capacity by integrating a time series prediction model of rock thermal decay characteristics. When the difference caused by the geological heat recovery lag effect exceeds the adaptive grading threshold, it generates a control decision instruction containing a control level identifier.

[0098] Linkage control execution module: After receiving control decision commands, it analyzes and predicts the difference data and heat pump operating parameters based on the dual actuator collaborative algorithm, and outputs the valve opening command of the phase change heat storage device and the compressor frequency correction value through the phase advance compensation mechanism, so as to realize dynamic pre-compensation for geothermal response lag and system energy efficiency optimization control.

[0099] It should be noted that the descriptions of the various embodiments are each given with emphasis on certain features of the embodiments. The descriptions of the various embodiments are not meant to be taken in a literal sense, and the features of the various embodiments can be combined with each other.

[0100] Those skilled in the art will appreciate that embodiments of the present application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0101] The present application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 means for performing the function of one or more of the steps in the flowchart illustrations and / or block diagrams.

[0102] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 means for performing the function of one or more of the steps in the flowchart illustrations and / or block diagrams.

[0103] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 means for performing the function of one or more of the steps in the flowchart illustrations and / or block diagrams.

[0104] While the preferred embodiments of the application have been described, additional variations and modifications can be made to these embodiments by those skilled in the art once they have the benefit of the present disclosure without departing from the spirit and scope of the application. Accordingly, it is intended that the appended claims include all such modifications and variations as fall within the scope of the present application.

[0105] It is apparent that those skilled in the art can make various changes and modifications to the application without departing from the spirit and scope of the application. It is therefore intended that the present application cover all such changes and modifications that are within its scope.

Claims

1. A method for linkage control of working parameters of a medium-deep geothermal water source heat pump, characterized in that, Specifically comprising the following steps: Step S1, when the deviation of the geothermal well water temperature from the historical average value of the same period exceeds the set threshold or the meteorological radar detects that the air temperature change rate exceeds the warning value, a disturbance signal injection step is triggered, a flow step disturbance signal is generated and injected into the geothermal water circuit through a reinforcement learning algorithm optimization, the frequency of the water pump is dynamically adjusted to control the flow amplitude and duration, a response data set is generated to maximize the sensor information entropy and establish a dynamic mapping of the underground thermal reservoir; Step S2, after obtaining the response data set, a geological parameter inversion step is triggered, a physical information neural network model is used to process pressure gradient field and temperature gradient field data, and a joint inversion is performed through thermal conduction equation constraint and Darcy flow constraint to generate rock layer thermal diffusion coefficient and fracture rate inversion results, to construct a real-time image of the underground thermal response and quantify the thermal inertia effect; Step S3, after obtaining the geological parameter inversion results, a load difference prediction step is triggered, meteorological forecast data and real-time user room temperature data are fused, a time series prediction model of rock thermal attenuation characteristics is used to generate a predicted difference between heating load and geothermal heat absorption capacity, and a control decision is generated when the predicted difference exceeds a dynamic threshold to predict the supply-demand conflict and set a hierarchical trigger condition; Step S4, after generating the control decision, a linkage control step is triggered, based on a double-actuator collaborative algorithm, the predicted difference data and the heat pump working condition parameters are analyzed to generate a phase change heat storage device valve opening control instruction and a compressor frequency adjustment value to perform pre-compensation control and finally optimize the energy efficiency of the heat pump system.

2. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 1, characterized in that: In step S1, the specific operation of dynamically adjusting the flow step amplitude and duration is as follows: When the absolute value of the difference between the real-time temperature data at the wellhead and the average well temperature of the same period in the same area exceeds the temperature fluctuation threshold, or the absolute value of the air temperature change rate calculated from the meteorological radar data exceeds the meteorological warning threshold, a disturbance strategy generation mechanism is triggered; Based on the proximal policy optimization algorithm in reinforcement learning, the optimal combination parameters are searched within the preset flow step amplitude range and disturbance duration range, and the optimization target of the search process is to maximize the weighted sum of the information entropy of all sensor data and the physical rule constraint term; wherein the physical rule constraint term is calculated from the thermal diffusion coefficient and underground water flow rate in the prior geological parameter set; the generated optimal flow step amplitude and duration are injected into the geothermal water circuit through a variable frequency pump, and pressure data and temperature data are collected at a frequency of once per second to form a thermal reservoir characteristic response data set.

3. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 2, characterized in that: The specific construction method of the physical rule constraint term is as follows: The mean square error of the temperature with respect to the depth partial derivative value and the thermal dispersion equation theoretical value is calculated, wherein the thermal dispersion equation theoretical value is obtained by dividing the prior thermal diffusion coefficient by the underground water flow rate and multiplying the second-order partial derivative of the temperature with respect to time; the information entropy calculation covers the real-time reading distribution probability of the pressure sensor and the temperature sensor, and the distribution probability is obtained by simulating the data distribution under different perturbation strategies through Monte Carlo sampling; the search space of the optimal combination parameter is limited by the rated flow rate ratio of the heat pump, and the lower limit of the flow rate step amplitude is 10% of the rated flow rate, and the upper limit is 30%, and the lower limit of the perturbation duration is 5 minutes, and the upper limit is 30 minutes.

4. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 3, characterized in that: In the step S2, the feature processing operation of the physical information neural network model is: Firstly, the response data set output by the step S1 is subjected to space-time alignment and noise reduction processing, and a pressure data matrix and a temperature data matrix are extracted as core inputs; an axial temperature gradient field is calculated based on the temperature data matrix along the sensor depth direction; a ternary loss function including a heat conduction constraint term, a Darcy flow constraint term and a data fitting term is constructed; wherein the Darcy flow constraint term introduces a permeability function, which quantifies the topological structure of the rock fracture network through an exponential decay relationship constructed by a fracture connectivity index and a critical fracture rate, and the fracture connectivity index is calculated according to the prior geological parameter set transmitted by the step S1; an adaptive Monte Carlo sampling algorithm is used to iteratively optimize the network parameters within a preset constraint range, and the constraint range is set according to the upper and lower floating thresholds of the thermal diffusion coefficient and the fracture rate in the prior geological parameter set transmitted by the step S1; finally, a parameter set containing the inversion value of the thermal diffusion coefficient and the inversion value of the fracture rate is output, and a corresponding confidence evaluation matrix is generated, which is used to judge the effectiveness of the inversion result, and if the fracture rate confidence is less than 0.7, the step S1 is returned to start additional perturbation.

5. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 4, characterized in that: The construction process of the confidence evaluation matrix is: According to the modulus characteristic value of the second-order derivative matrix of the network parameters of the total loss function of the physical information neural network model, a negative exponential mapping function is applied to calculate the confidence scalar of the inversion value of the thermal diffusion coefficient and the fracture rate; The operation of the Darcy flow constraint term is to minimize the sum of squares of errors between the measured pressure gradient data of the sensor array and the theoretical pressure gradient data calculated based on the fluid dynamic viscosity, the underground water flow rate and the permeability function as an optimization sub-objective; Wherein the permeability function is constructed depending on the fracture connectivity index and the critical fracture rate, the fracture connectivity index calls the pre-calibrated parameters of the regional geological database, and the critical fracture rate is set to 0.18; The iteration termination condition of the adaptive Monte Carlo sampling algorithm is that when the Frobenius norm of the Hessian matrix of the total loss function with respect to the network parameters is lower than the preset convergence threshold, the optimization is stopped.

6. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 5, characterized in that: In the step S3, the load difference prediction step is specifically: Firstly, the geothermal heat recovery time constant is calculated based on the geothermal thermal diffusivity and fracture porosity parameters obtained by step S2, and the value is obtained through a combined expression of rock density, specific heat capacity, wellbore radius and fracture lag gain, wherein the fracture lag gain is set according to the calibration range of granite high temperature rheological experiment; secondly, a bidirectional thermal decay function of building thermal inertia and geothermal delay is constructed, and the function structure includes building heat capacity and surface convection coefficient calculation item, permeability related heat conduction efficiency item and frequency domain convolution operation item; Then, the long short-term memory network unit is improved, and a geological feature vector composed of the thermal diffusivity, fracture porosity and geothermal heat recovery time constant is injected to perform dynamic prediction of the heating load; finally, combined with the maximum absolute value of air temperature change rate, a dynamic threshold is generated through a geothermal hysteresis sensitive factor and a weight coefficient, and when the ratio of the prediction difference and the dynamic threshold meets the preset grading condition, a grading flag of the control decision is output to trigger the subsequent linkage control.

7. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 6, characterized in that: The frequency domain convolution operation of the bidirectional thermal decay function is specifically: The product operation is performed on the building thermal inertia index decay term and the geothermal delay index decay term after the fast Fourier transform; The injection mode of the geological feature vector in the long short-term memory network is to increase the three-dimensional vector splicing operation in the input gate, the forgetting gate and the output gate calculation formula; the weight coefficient of the dynamic threshold generation formula is obtained by reinforcement learning in the heating season historical data, and the geothermal hysteresis sensitive factor is selected according to the rock type in the preset interval; The grading trigger condition is defined as: when the predicted load difference exceeds fifteen percent but is less than twenty-five percent of the dynamic threshold, the first level control is started, and when it exceeds twenty-five percent, the second level control is started, and the geothermal heat recovery time constant is transmitted to the lag compensator optimization module.

8. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 7, characterized in that: In step S4, the linkage control step is specifically: Firstly, based on the control decision grading flag and the geothermal heat recovery time constant output by step S3, a phase lead compensator is constructed, and a time domain compensation amount is generated by setting a phase lead gain and a differential time constant; then, a double actuator cooperative control is performed, and when the decision grading flag is one, the valve opening degree control of the phase change heat storage device is activated, and when the decision grading flag is two, the compressor frequency adjustment is started, and the control law includes a proportional term, an integral term and a feedforward term based on the user room temperature change rate, wherein the compressor adjustment value and the valve opening degree are dynamically associated through a coupling coefficient; finally, a dynamic amplitude constraint is applied, the valve opening degree is in the range of zero to one hundred percent, and the compressor frequency is adaptively limited according to the geothermal heat recovery time constant in a negative exponential rule, so that the actuator action is within the physical safety boundary.

9. The working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to claim 8, characterized in that: The compensation time amount of the phase lead compensator is obtained by multiplying the geothermal heat recovery time constant by a fixed proportion coefficient, and the proportion coefficient is set according to the calibration range of granite high temperature rheological experiment; The coupling activation condition of the double actuator cooperative control is defined as being effective only when the decision grading flag value is greater than or equal to one; the weight parameters of the proportional term, the integral term and the feedforward term are optimized and determined in the heating history data through reinforcement learning; The upper bound of the compressor frequency limiting function decreases with the increase of the thermal recovery time constant, specifically manifested as the product of the maximum allowed frequency and a negative exponential decay function.

10. The working parameter linkage regulation system of a medium-deep geothermal water source heat pump is applied to the working parameter linkage regulation method of a medium-deep geothermal water source heat pump according to any one of claims 1-9, characterized in that: Specifically, the method comprises the following steps: The disturbance signal injection module, the geological parameter inversion module, the load difference prediction module, and the linkage control execution module; When the underground temperature sensor detects that the water temperature of the geothermal well deviates from the historical average by more than a set threshold, or the weather radar detects that the air temperature change rate exceeds the warning value, the disturbance signal injection module generates a flow step disturbance signal by reinforcement learning algorithm optimization and injects it into the geothermal water circuit. The variable frequency water pump dynamically adjusts the flow amplitude and duration as the core control means to generate a multi-sensor response data set containing the spatial and temporal distribution of pressure and temperature, and establish a dynamic mapping atlas of the underground thermal reservoir. The geological parameter inversion module processes the pressure gradient field and temperature gradient field data using a physical information neural network model after receiving the response data set. Through the joint inversion of the heat conduction equation constraint and the Darcy flow constraint, the geological parameter set composed of the rock thermal diffusivity and the fracture rate is output, and the inversion result confidence evaluation matrix is generated. The load difference prediction module fuses meteorological forecast data and real-time user room temperature data after obtaining the geological parameter set. Through the time series prediction model of the rock thermal attenuation characteristics, the dynamic difference between the building heating load and the geothermal heat absorption capacity is calculated. When the difference value caused by the geological thermal recovery lag effect exceeds the adaptive hierarchical threshold, the control decision instruction containing the control level identifier is generated. The linkage control execution module analyzes the predicted difference data and the heat pump working condition parameters based on the double actuator collaborative algorithm after receiving the control decision instruction. Through the phase advance compensation mechanism, the valve opening degree instruction of the phase change heat storage device and the compressor frequency correction value are output, realizing the dynamic pre-compensation of the geothermal response lag and the system energy efficiency optimization control.

Images