Ground source heat pump buried pipe partition layout and cold and heat balance control method

Abstract

The invention relates to the technical field of ground source heat pumps, and discloses a ground source heat pump buried pipe zoning layout and cold and heat balance control method which comprises the following steps that the current three-dimensional ground temperature field state of the area where a ground source heat pump buried pipe is located is obtained; predicting a future state sequence of the ground temperature field under different control strategies in a future prediction time domain; solving to obtain an optimal control sequence based on a multi-objective optimization function which aims at minimizing the comprehensive operation cost of the system in the prediction time domain; the multi-objective optimization function at least comprises an energy consumption item representing system short-term operation energy consumption and a balance item representing a ground temperature long-term balance state; and executing the first control action in the optimal control sequence. The prediction model constructed by the invention can combine a physical rule with a data residual continuously learned from actual operation, and ensures that prediction information on which a control decision depends always keeps high fidelity in long-term operation of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ground source heat pump, and particularly to a ground source heat pump buried pipe partition layout and cold-heat balance control method. BACKGROUND

[0002] As an energy-saving technology for efficient utilization of shallow geothermal energy, ground source heat pump has been widely used in the field of building heating and refrigeration. The core performance of ground source heat pump depends largely on the performance of ground heat exchanger (GHE), and the comprehensive performance of ground heat exchanger is determined by the physical layout design and the operation control strategy in the whole life cycle.

[0003] In the prior art, the design process of ground heat exchanger is usually prior to and independent of the development of its operation control strategy. In the design stage, the engineers usually determine the key parameters such as drilling depth and spacing according to industry manuals, empirical formulas or simplified static load estimation. The primary purpose of this method is to ensure that the system can meet the peak load demand of the building under extreme working conditions, and essentially regards the ground heat exchanger as a passive and static component. Due to the lack of forward-looking and coordinated planning in ground heat exchanger design, the physical layout scheme fails to provide optimization space for advanced dynamic control strategy, which fundamentally limits the potential performance of the system.

[0004] When the system is built and put into operation, the traditional control system usually adopts reactive logic. For example, whether to enable a certain partition is determined by setting a fixed soil temperature threshold, or a conventional PID controller is used to track the building load. Such controllers lack the ability to predict future situations. For a controlled object like the ground temperature field with huge thermal inertia and long time lag, this control method based only on current measurement values is not sufficient, and often makes the system in an unnecessary frequent start-stop or large fluctuation state, thereby causing energy waste.

[0005] More importantly, the above-mentioned deficiencies in design and control jointly lead to a deeper problem: the sustainability of long-term operation of the system is difficult to guarantee. In the area where cooling load is dominant, the heat injected into the ground year after year cannot be effectively dissipated, which will cause the ground temperature to rise year by year; vice versa. The existing simple control rules cannot finely and dynamically balance between meeting the instantaneous comfort demand and maintaining the long-term ground temperature balance. The lack of this ability has become a core bottleneck restricting the long-term sustainable operation of the system.

[0006] Even in some schemes that attempt to implement control using more advanced models, an inherent flaw is prevalent. The predictive models upon which these schemes rely are typically static, with parameters calibrated once during system commissioning and then left unchanged. However, the real underground environment is not constant; seasonal fluctuations in groundwater levels and changes in soil moisture content both affect its thermal conductivity. Fixed models cannot track these real-world changes, and over time, model mismatch worsens, leading to a significant decrease in predictive accuracy. Consequently, control decisions based on this flawed model increasingly deviate from the true optimal state. Summary of the Invention

[0007] This invention aims to provide a method for zoned layout and thermal balance control of ground source heat pump buried pipes. This method solves the problem of long-term performance degradation and ground temperature imbalance caused by the disconnect between design and control in the prior art by coordinating offline optimization design of buried pipes with online predictive control of the system.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for zoned layout and thermal balance control of ground source heat pump buried pipes, comprising the following steps: S1: Obtain the current three-dimensional geothermal field status of the area where the ground source heat pump buried pipe is located; S2: Based on a prediction model that can characterize the dynamic evolution of the geothermal field, and according to the current three-dimensional geothermal field state, predict the future state sequence of the geothermal field under different control strategies in a future prediction time domain; S3: Based on a multi-objective optimization function that aims to minimize the overall system operating cost within the predicted time domain, an optimal control sequence is obtained; the multi-objective optimization function includes at least an energy consumption term characterizing the short-term operating energy consumption of the system and a balance term characterizing the long-term geothermal equilibrium state; S4: Execute the first control action in the optimal control sequence to control the operation of the ground source heat pump system.

[0009] Preferably, the prediction model in step S2 is a hybrid driving model, which includes: A physical model based on the physical mechanism of heat conduction is used to predict the main trends in geothermal field changes; A residual compensation model based on historical data is used to learn and compensate for the prediction errors of the physical model.

[0010] Preferably, step S2 further includes an online self-calibration step: After each control cycle, the actual measured value of the geothermal field is compared with the predicted value of the physical model to obtain new residual data, and the residual compensation model is trained and updated online using the new residual data.

[0011] Preferably, in step S3, the weights of each sub-item in the multi-objective optimization function are dynamically adjusted; Step S3 also includes: Based on the current three-dimensional geothermal field state, a geothermal health index is calculated to quantify the degree of geothermal imbalance. Based on the geothermal health index, the weights of the energy consumption term and the balance term in the multi-objective optimization function are dynamically adjusted.

[0012] Preferably, the calculation of the geothermal health index is based on at least one or more of the following: The deviation between the spatial average temperature of the current geothermal field and the spatial average temperature of the initial geothermal field; the spatial distribution variance of the current geothermal field.

[0013] Preferably, the control method is a model predictive control method, wherein steps S1 to S4 are repeatedly executed in a rolling time domain manner.

[0014] Preferably, the method further includes an offline design step performed before the underground pipe construction, the offline design step including: The physical layout scheme of the buried pipe is determined through full life cycle simulation optimization; The process of full lifecycle simulation optimization includes different candidate physical layout schemes.

[0015] Preferably, in the offline design step, the physical layout scheme of the buried pipe is a zoned layout scheme; The goal of the simulation optimization is to determine the optimal combination of physical parameters for each partition, including at least the borehole depth and pipe spacing.

[0016] Preferably, the partitioning scheme includes at least two functionally differentiated partitions and adopts a gradient layout scheme; The gradient arrangement scheme is characterized by a gradient increase in drilling depth and / or pipe spacing from the outer partition to the core partition, which is used to actively optimize the long-term heat exchange performance of the core area.

[0017] Preferably, in regions where cooling load is dominant, the physical layout scheme determined through simulation optimization has the following characteristics: The pipe spacing in the inner zone is greater than that in the outer zone, and the drilling depth in the inner zone is greater than that in the outer zone, which is used for cold dissipation and long-term heat absorption in the core area.

[0018] In summary, the present invention has at least one of the following beneficial technical effects: 1. This invention provides an offline simulation optimization design method that pre-binds the zonal layout of underground pipes with the operation and control throughout the entire lifecycle, thereby obtaining an optimal physical structure that adapts to advanced control strategies from the outset. Compared to existing technologies that rely solely on empirical formulas or static load estimations for design, this invention overcomes the inherent defect of a disconnect between physical layout and control strategies, which limits the overall performance potential of the system.

[0019] 2. This invention innovatively constructs a geothermal health assessment model and uses it as the basis for adjusting the dynamic weights of the optimization function. As a result, the system has the ability to intelligently balance short-term operational economy and long-term geothermal sustainability. This is in stark contrast to the reactive control logic based on fixed temperature thresholds in existing technologies. It effectively solves the fundamental problem that existing technologies cannot predict and suppress the slow and cumulative imbalance of the geothermal field, which ultimately leads to the gradual decline of system energy efficiency.

[0020] 3. The method of this invention applies a model predictive control framework to the operation of ground source heat pumps. The controller no longer passively responds based solely on the current state, but actively optimizes by predicting future time domain conditions. This forward-looking control approach, compared to existing technologies based on instantaneous parameters for start-stop or PID control, fundamentally solves the problems of frequent operational fluctuations and energy waste caused by their inability to effectively handle the large time delays and inertia characteristics of the system.

[0021] 4. This invention constructs a hybrid-driven predictive model with online self-correction capabilities. This model combines physical laws with data residuals continuously learned from actual operation to achieve self-evolution, ensuring that the predictive information upon which control decisions rely maintains high fidelity throughout the long-term operation of the system. This technical solution effectively overcomes the inherent defects of static and fixed models in existing technologies. Static and fixed models cannot adapt to changes in the real environment, and will eventually lead to deterioration of control performance due to model mismatch. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the zoning plan layout of the underground pipes according to the present invention; Figure 2 This is a flowchart of the operation control logic of the present invention. Detailed Implementation

[0023] The following is in conjunction with the appendix Figure 1 - Appendix Figure 2 The present invention will be further described in detail below.

[0024] This invention proposes a gradient layout design concept, the core of which lies in using differentiated zoning design to present a gradient change in the physical layout of buried pipes from the periphery to the core. Specifically, the drilling depth gradually increases from the outside to the inside, and the pipe spacing gradually increases from the outside to the inside. This design is not a passive response, but a proactive and forward-looking optimization of the physical structure, aiming to effectively guide the heat dissipation from the core area outward or provide more space for seasonal energy storage, fundamentally solving the problem of long-term heat accumulation in the center of the site, and ensuring the long-term health and balance of the entire geothermal field.

[0025] S1: Obtain the current three-dimensional geothermal field status of the area where the ground source heat pump buried pipe is located; Specifically, this step aims to transform discrete, limited measurement information in the physical world into a complete, high-resolution digital state representation that can be computed by subsequent prediction models.

[0026] The specific implementation of this step includes the deployment of a physical-level sensor network and an algorithm-level data reconstruction process. First, while the underground pipe project is under construction, a three-dimensional temperature sensor network is deployed in the underground area. This sensor network, exemplarily, includes multiple independent temperature sensor modules and a central data acquisition unit that communicates with the multiple temperature sensor modules.

[0027] The multiple temperature sensor modules are pre-planned and arranged at different spatial coordinate positions within the underground area to form a three-dimensional monitoring matrix. To accurately reflect the thermal state of different functional zones, the sensor module placement is specifically arranged according to the zoning plan. Furthermore, within each functional zone (e.g., inner zone, middle zone, outer zone), at least one temperature sensor module is deployed at different planar positions and different burial depths to ensure effective capture of temperature gradients in both the vertical and horizontal directions.

[0028] The central data acquisition unit is connected to all temperature sensor modules via wired or wireless communication. Its function is to send data acquisition commands to all sensor modules at the start of a preset control cycle, or to receive real-time temperature data actively reported by each module. The acquired raw data constitutes a discrete temperature dataset, which takes the form of a set of spatial coordinates at time t and their corresponding temperature measurements {(x...}). i ,y i ,z i ),T i (t)}, where i = 1, 2, ..., n, and n is the total number of sensors.

[0029] Since the number of sensors is always limited, the raw data collected is sparse and discrete, and cannot be directly used as input for subsequent physical model calculations. Therefore, after obtaining the discrete temperature dataset, a spatial interpolation algorithm is needed to reconstruct the data of the entire underground region, thereby generating a complete and regularized three-dimensional geothermal field state matrix.

[0030] For example, this invention uses the Kriging interpolation algorithm to complete this data reconstruction process. The Kriging algorithm is an unbiased optimal estimation algorithm based on spatial autocorrelation, whose core lies in using the spatial relationships between known sample points to predict the values ​​of unknown points. The implementation of this algorithm includes the following steps: First, using real-time temperature data from all n sensor points, the experimental variability function γ(h) is calculated to characterize the spatial correlation and variability of the geothermal field. The calculation relationship is as follows: In the formula, γ(h) is a function of h. This is a normalization factor, where N(h) represents the number of samples used in the calculation for a specific value of h. T(x) represents the summation over indices i from 1 to N(h), where N(h) is the number of valid samples under a specific condition h, and T(x) represents the summation over indices i from 1 to N(h). i ) indicates at position x i The value of a certain quantity at a certain point, T(x) i +h) indicates that at position x i The same value at +h, where h is usually a small increment and represents the observation after a position shift of h, [T(x i )-T(x i +h)] 2 Indicates position x i and position x i The square of the difference between the measures at +h.

[0031] Subsequently, a theoretical variogram model (e.g., a spherical model, an exponential model, or a Gaussian model) is selected to fit the calculated experimental variogram. This process yields a continuous variogram model that quantitatively describes the expected variance of temperature values ​​between two points at any distance.

[0032] Finally, based on the fitted variogram model, the temperature value is estimated for each pre-divided grid node x0 within the entire underground region to generate a three-dimensional geothermal field. The estimated temperature value at any unknown point x0 is... It can be calculated using a weighted linear combination of the temperatures of known measuring points in its vicinity, and the calculation relationship is as follows: In its formula, This is the estimated temperature at location x0. Indicates to The summation operation, This represents the summation over indices i from 1 to n, where n is the number of data points used to describe the summation, and w i (x0) is the weighting coefficient of the i-th sensor for the estimated point x0, T(x i ) represents the measured temperature value of the i-th sensor.

[0033] The weighting coefficient w i (x0) is obtained by solving a system of equations designed to satisfy unbiasedness and minimum variance constraints. The construction of this system of equations directly depends on the aforementioned well-fitted variogram model.

[0034] By repeatedly performing the above interpolation calculation on all nodes in the entire underground region, the final output is the three-dimensional geothermal field state matrix T(t) at the current time t required by this invention. It is a high-resolution, complete digital object that can be directly processed by a computer.

[0035] This step reliably transforms limited, discrete sensor measurement data from the physical world into a digital matrix that comprehensively characterizes the current macroscopic thermal state of the geothermal field. This matrix is ​​not only spatially continuous and complete, but its generation process also considers the inherent spatial structure characteristics of the geothermal distribution, resulting in high fidelity. This three-dimensional geothermal field state matrix T(t) will serve as an indispensable initial condition for the prediction model to extrapolate future state evolution in subsequent steps, laying a solid data foundation for the accuracy and reliability of the entire predictive control method.

[0036] S2: Based on a prediction model that can characterize the dynamic evolution of the geothermal field, and according to the current three-dimensional geothermal field state, predict the future state sequence of the geothermal field under different control strategies in a future prediction time domain; Specifically, after obtaining the three-dimensional geothermal field state matrix T(t) at the current time t, the method of this invention enters the prediction stage. The core objective of this stage is to extrapolate the changes in the geothermal field within a predetermined time domain based on a prediction model that can accurately characterize the complex dynamic evolution of the geothermal field. This prediction is not singular, but rather generates corresponding future state sequences based on different given candidate control strategies, providing an analytical basis for subsequent optimization decisions.

[0037] The prediction model used in this invention is not a single physical model or data model, but a hybrid-driven prediction model. This model organically integrates modeling methods based on physical mechanisms with modeling methods driven by historical data, thus balancing the universality of physical laws with the specificity of complex disturbances in real-world environments, thereby achieving high-precision prediction results. The hybrid-driven prediction model, exemplarily, includes a physical mechanism module and a data-driven residual compensation module.

[0038] The physical mechanism module describes the main trends and fundamental laws governing the evolution of the geothermal field. This module is based on the partial differential equation of heat conduction, which characterizes the heat conduction process in porous underground media. The governing equations of this physical model are as follows: In the formula, T represents temperature, which is a function of spatial location and time. This represents the partial derivative of temperature T with respect to time t. This represents the rate of energy change per unit volume of the material, where t is time, ρ is the density of the underground medium, c is the specific heat capacity of the underground medium, and k is the thermal conductivity of the underground medium. Q is the divergence operator; source (t) represents the source and sink terms, which characterize the heat exchange power between the buried pipe heat exchanger and the surrounding soil.

[0039] Furthermore, the source and sink terms Q source The source-sink term u(t) is determined by the control actions of the ground source heat pump system. In a specific embodiment, the system's control actions are the operating frequency of the circulating water pumps in each zone and the opening degree of the electric valve assembly. These actions collectively determine the flow rate and temperature of the circulating fluid flowing through the buried pipes in each zone, and thus the heat exchange power. Therefore, this source-sink term is a function of the candidate control strategy u(t) and serves as a bridge connecting control and state evolution. This physical model can be numerically solved using the finite difference method or the finite element method.

[0040] The data-driven residual compensation module is designed to learn and compensate for the prediction errors of the physical mechanism module. Because the real underground environment contains complex factors that are difficult to describe precisely with physical equations, such as groundwater seepage and seasonal variations in soil moisture, the predictions of the physical mechanism module will inevitably deviate from the actual situation, resulting in residuals. This module aims to learn the behavioral patterns of these residuals using historical data.

[0041] For example, this data-driven residual compensation module is constructed using a Long Short-Term Memory (LSTM) network. LSTM is a special type of recurrent neural network, well-suited for processing and predicting long-term dependencies in time series data. The input to this LSTM network is the true residual sequence over a past period, and the output is a prediction of the residual for the next time step.

[0042] Specifically, at the end of each preset control cycle (e.g., a 15-minute time step) and after the control command has been executed, the system's central controller will automatically trigger and execute an online self-calibration program.

[0043] Real-time data acquisition and residual calculation: First, the system collects and integrates the actual measured value of the geothermal field at the current time t through a distributed high-precision temperature sensor network deployed in the underground heat exchange area, denoted as vector T. actual (t). Simultaneously, the controller retrieves the predicted geothermal field value for the current time t, generated by the physical model during the previous planning cycle, denoted as vector T. phy_pred (t). Subsequently, the latest actual residual vector e is calculated using the following formula. actual (t): Online training of the residual compensation model: the newly obtained actual residual vector e actual (t) constitutes a valuable training sample with precise timestamps, which will serve as a true label for online training and updating of the residual compensation model.

[0044] The goal of training is to fine-tune the internal adjustable parameter θ of the residual compensation model to improve its predictions under the same input conditions. The goal is to approximate the newly acquired actual residual value as closely as possible. This is achieved by minimizing a pre-defined loss function L. A typical loss function is the mean squared error, as shown in the following equation: In the formula, L represents the loss function. This represents the average of M samples. This represents the summation over the range 1 to M, where M is the total number of samples. This represents the square of the difference between the predicted and actual values. The model represents the input x. i The predicted value, e actual,i (t) represents the actual or target value at time t.

[0045] Setting key training parameters: Optimizing and updating model parameters, preferably using gradient descent or its advanced variants (such as the Adam optimizer). The iterative rule for parameter updates is shown in the following equation: In its formula, θ new This represents the updated parameter value, θ. old This represents the parameter values ​​before the update, and η represents the learning rate. This represents the gradient of the loss function L with respect to the parameter θ.

[0046] By periodically and automatically executing the aforementioned online self-calibration steps, the residual compensation model continuously improves and evolves. This innovative mechanism enables the entire hybrid-driven model to effectively track and adapt to slowly changing dynamic characteristics not precisely described in the physical model, such as seasonal variations in groundwater conditions and property drift caused by long-term soil heat extraction and release, ensuring the effectiveness of the prediction model under long-term operation.

[0047] During prediction, the physical mechanism module and the data-driven residual compensation module work together. First, the physical mechanism module calculates the predicted value of the main geothermal field for the next time step based on the current geothermal state and the given control input. Simultaneously, the data-driven residual compensation module predicts the residual for the next time step based on the historical residual sequence. The two are then added together to obtain the final compensated single-step prediction result. The calculation relationship is as follows: In its formula, This is the final geothermal field prediction matrix at time t+1. Let represent a function related to the physical model, describing the influence of the temperature T(t) at time t and the control input u(t) on the future temperature, where T(t) represents the current temperature value at time t. The evolution function representing the physical mechanism module, u(t), is the control action applied at time t. This represents a function that uses a Long Short-Term Memory (LSTM) network. The prediction function representing the LSTM residual compensation module, ε t For the true predicted residual at time t, ∈ t-1 This indicates the error at an earlier time.

[0048] To obtain the state sequence in the future prediction time domain, this step will perform the above single-step prediction iteratively. First, based on the current three-dimensional geothermal field state T(t) obtained in the previous stage, and a given candidate control strategy sequence U = {u t ,u t+1 ,...,u t+N-1}. Then, starting from the current time t, the formula (4) is called repeatedly to perform the deduction, and the prediction output of the previous step is used as the prediction input of the next step, until all predictions for the next N steps are completed.

[0049] Finally, the output of this step is a predicted sequence of future geothermal field states corresponding to the candidate control strategy U. This step, through a self-correcting hybrid driving model, enables reliable prediction of the system's future dynamics, providing essential decision-making basis for finding the optimal solution among multiple candidate control strategies, and giving the control method a forward-looking effect.

[0050] S3: Based on a multi-objective optimization function that aims to minimize the overall system operating cost within the predicted time domain, an optimal control sequence is obtained; the multi-objective optimization function includes at least an energy consumption term characterizing the short-term operating energy consumption of the system and a balance term characterizing the long-term geothermal equilibrium state; Specifically, the purpose of this step is to transform the predictive information generated in the previous step into a specific sequence of optimal control actions that can guide the system's operation. This transformation process is achieved by constructing and solving a multi-objective optimization problem in a finite-time domain. The fundamental goal of this problem is to find a series of future control actions that minimize the overall operating cost of the system throughout the entire predictive time domain while meeting the building's comfort requirements.

[0051] To achieve this goal, it is first necessary to construct a multi-objective optimization function capable of accurately quantifying the overall operating cost of the system. This function integrates multiple interrelated and even mutually restrictive performance indicators, such as short-term energy efficiency, long-term sustainability, and user experience, into a unified mathematical framework, thereby enabling synergistic optimization among them. For example, the multi-objective optimization function J is defined as the weighted sum of various costs over the next N control cycles. Its specific mathematical expression is as follows: In the formula, min U J represents the objective function J to be minimized, the optimization variable is the control input U, N is the length of the control prediction, where k is the index of the time step, and α(H) k C represents a certain adjustment coefficient or weight related to time step k. energy (u k ) represents the control input u k The relevant energy cost function, β·C comfort (k) represents the comfort cost function associated with time step k, γ(H) k ) is the adjustment coefficient related to time step k, c balance (k+1) represents the balance cost function associated with time step k+1, where α, β, and γ are the weighting coefficients for the above costs, respectively.

[0052] Furthermore, the energy consumption cost item C energy (u k Its physical meaning lies in quantifying the execution of control actions. kThe required instantaneous power consumption. This item mainly represents the total power consumption of all circulating water pumps in the ground source heat pump system. Its value can be calculated using the pump performance curve model based on the pump operating frequency as the control variable.

[0053] The comfort cost item C comfort (k) aims to quantify the degree of matching between the system's energy supply and the building's actual needs, thereby ensuring a thermally comfortable environment for users. For example, this term can be defined as the squared deviation between the system's predicted total heat exchange at that moment and the predicted building load. Its calculation relationship is as follows: Q exch To control action u k and predict ground temperature The calculated predicted heat transfer, and L k These are known building load forecasts provided by the building automation system.

[0054] The geothermal balance cost item C balance (k+1) aims to evaluate the health of the geothermal field from a spatial perspective, in order to suppress the accumulation of underground hot and cold substances. For example, this term can be defined as being determined by the control action u. k The resulting prediction of the geothermal field matrix at the next moment The spatial distribution variance, i.e. The smaller the value of this item, the more uniform the distribution of geothermal temperature throughout the underground area, and the better the long-term sustainability of operation.

[0055] A key and innovative technical feature of this invention is that the weight coefficients in the optimization function are not static constants, but variables that can be dynamically and adaptively adjusted according to the system state. Specifically, the weight α of the energy consumption term, which characterizes short-term economic efficiency, and the weight γ of the balance term, which characterizes long-term sustainability, are designed as a function of the overall health status of the geothermal field. To achieve this function, before formally entering the optimization solution, this method also includes calculating a geothermal health index H used to quantify the current risk of geothermal imbalance. k The steps. Further, the geothermal health index H... k The calculation is based on the current moment. Predicted geothermal field This index was obtained through comprehensive analysis. Its calculation incorporates information from two dimensions: the overall temperature shift of the geothermal field and the uneven temperature distribution within the geothermal field. The specific calculation relationship is as follows: In the formula, H k For the geothermal health index in the k-th control period, express and The L2 norm between them For the predicted geothermal field in the k-th control period The average temperature of the space, The spatial average temperature of the initial geothermal field, obtained through initial surveys before system operation, is used as a reference value. For the predicted geothermal field in the k-th control period Spatial variance, w dev and w var These are pre-set constant weighting coefficients used to adjust the magnitude of the influence of the two parts.

[0056] Obtain the geothermal health index H at each moment. k Then, the system will dynamically generate the weighting coefficient α(H) based on the preset adjustment function. k ) and γ(H k The underlying logic of this adjustment function is: when the health index H... k When the temperature is in the lower ideal range, it indicates good ground temperature conditions. In this case, the system assigns a higher weight α to the energy consumption term and a lower weight γ to the balance term, making the optimization algorithm focus more on finding the highest energy efficiency for short-term operation. Conversely, when the health index H... k When the temperature rises due to long-term accumulation of thermal imbalance and exceeds a preset safety threshold, the weight γ(H) of the balance term increases. k The variance of the geothermal field will increase dramatically and non-linearly, significantly enhancing its dominance in the total cost function. At this point, the optimization algorithm will be forced to prioritize control actions that can most effectively reduce the variance of the geothermal field and restore geothermal balance, such as forcibly driving inactive zones to perform reverse heat exchange, even if this action may lead to a temporary increase in energy consumption in the short term.

[0057] After constructing the complete objective function and determining all dynamic weights, this step employs an efficient nonlinear programming solver, such as the interior-point method or sequential quadratic programming algorithm, to numerically solve the optimization problem. The solution process must be conducted under a series of physical and operational constraints, including but not limited to: upper and lower limits of the operating frequency of each pump, the opening range of each electric valve, and the requirement that the outlet fluid temperature of the buried pipe must be higher than the antifreeze lower limit. The final output of this optimization solution process is an optimal sequence of control actions from the current time t to the future time t+N-1.

[0058] S4: Execute the first control action in the optimal control sequence to control the operation of the ground source heat pump system; Specifically, in the preceding steps, by solving a multi-objective optimization problem, an optimal control sequence aimed at minimizing the overall future operating cost of the system has been obtained. This step is the final execution stage of the entire online control closed loop, and its purpose is to transform the calculated optimal decision, located in the virtual time domain, into actual and precise control of the ground source heat pump system in the physical world. The core of this step lies in the implementation of the rolling time domain execution strategy. After the optimization solution in the previous step is completed, the controller obtains an optimal control sequence U covering the entire prediction time domain N. * Its specific mathematical form is as follows: In its formula, U * Represents a set or vector of control inputs. This represents the optimal control input at time step t. This represents the optimal control input at time step t+1. This represents the optimal control input at time step t+N-1, where N is an integer.

[0059] The method of this invention does not execute the complete optimal control sequence U. * Instead, from this sequence, only the first element is extracted and selected, which corresponds to the optimal control action for the current control period t. The rest of the sequence, i.e. It will then be actively discarded by the system.

[0060] The strategy of executing only the first control action is the core principle of Model Predictive Control (MPC). Its technical logic lies in the fact that the optimal future control sequence calculated at the current moment is the best decision derived from all available information up to that moment (including the model, measurements, and predictions). However, as time progresses, new measurement information will be obtained, and the actual system state may deviate from the predicted trajectory due to unmodeled disturbances. Therefore, re-optimizing the global system at the next moment will yield a better decision.

[0061] The optimal control action at the current moment has been determined. Subsequently, the central controller in the system translates these into specific physical execution instructions. The central controller, exemplarily, includes a control signal generation module and a communication interface. The control signal generation module is responsible for converting the vector... The values ​​contained therein (for example, the frequency setting of the water pump in zone A is 45.5 Hz, and the opening setting of the electric valve group in zone B is 85%) are converted into industrial standard electrical signals.

[0062] These electrical signals are transmitted to the corresponding field actuators in the ground source heat pump system via the communication interface and fieldbus network. The actuators specifically include variable frequency drives (VFDs) connected to the circulating water pumps of each zone, and electrically operated regulating valves installed on the pipelines of each zone.

[0063] Upon receiving the corresponding frequency command signal, the variable frequency drive precisely adjusts the power supply frequency of the water pump motor, thereby changing the pump speed and controlling the flow rate of the circulating working fluid. Simultaneously, upon receiving the opening command signal, the actuator of the electric regulating valve drives the valve to the specified opening degree, thus regulating the proportion of the working fluid flowing through that zone. In this way, the calculated optimal control action can be precisely implemented at the physical level.

[0064] In the optimal control action After being executed and maintained for a complete control cycle (e.g., 15 minutes), the system time steps to the next time point t+1. At this point, the method of the present invention will not execute the previously discarded steps.

[0065] Instead, the entire control process returns to the first step of the method of this invention, namely, re-collecting and reconstructing the new current three-dimensional geothermal field state at a new time t+1, incorporating the control effects of the previous cycle, through the sensor network. Subsequently, based on this latest state, the system again executes the complete closed loop of prediction, optimization, and execution. This iterative process of sensing-prediction-optimization-execution of the first step is the implementation method of rolling time-domain control.

[0066] This step, by executing only the first action of the optimal sequence and continuously optimizing it, constitutes a closed-loop feedback control system. This mechanism enables the control system to continuously use the latest field measurement data to correct deviations in model predictions and to respond quickly to unforeseen disturbances, ensuring that the system always approaches the theoretical optimal state during actual operation.

[0067] The first and essential prerequisite for the offline design steps performed before the construction of the underground pipe is to conduct a detailed site geological survey and geotechnical thermal property test. Due to the significant differences in heat exchange effects among different strata (such as clay, sand, and bedrock), the heat exchange capacity of a borehole at the same depth may differ by several times under different geological conditions.

[0068] Therefore, it is essential to accurately obtain key thermophysical parameters such as thermal conductivity, specific heat capacity, and groundwater seepage velocity of soil and rock layers at different depths within the site through methods such as borehole sampling and field thermal response testing (TRT). These measured data will serve as the most fundamental and core inputs to the simulation optimization model, directly determining the accuracy and reliability of subsequent physical layout design, and are the fundamental basis for ensuring the rationality of borehole design.

[0069] Taking a typical commercial building primarily used for refrigeration as an example, the offline design steps of this invention may include: Step 1: Zoning and gradient layout: The underground pipe site is simplified into two functionally distinct zones: the outer zone and the inner zone.

[0070] Outer Zone: Defined as the ring-shaped area surrounding the site, primarily bearing the majority of the system's base load and peak load. Drilling is densely distributed in this zone to ensure rapid response.

[0071] Inner Zone: Defined as the core area at the center of the site, primarily used to bear seasonal loads and achieve long-term heat / cold storage functions. This zone features a relatively sparse but deeper borehole layout to enhance its long-term heat exchange and energy storage capabilities.

[0072] Step 2: Determining and optimizing key parameters: Before proceeding with subsequent simulation optimization, it is necessary to clarify the value range of key parameters. It should be emphasized that the following values ​​are merely examples; the optimal parameters for actual engineering projects must be ultimately determined based on on-site geological test results and full lifecycle simulation optimization.

[0073] In this case, based on local construction experience and geological conditions, the preliminary parameter range is set as follows: Drilling depth in the outer zone: 80-120 meters.

[0074] Drilling depth in the inner zone: 120-150 meters.

[0075] Through simulation optimization, the optimal combination of drilling depth and pipe spacing for each zone was finally determined to minimize the total life cycle cost.

[0076] The full lifecycle simulation optimization is not a vague concept, but a rigorous, quantifiable, multi-parameter optimization process. This process clearly defines the inputs required for computation, the variables to be optimized, and the final optimization objective, as follows: Optimize Input: Before starting simulation optimization, the following basic data must be provided to the model: Building and environmental data: including the project building's hourly heating and cooling load curves throughout the year, as well as local hourly meteorological data (such as dry-bulb temperature, wet-bulb temperature, etc.).

[0077] Geological physical parameters: including the thermal conductivity, specific heat capacity and groundwater conditions of each soil and rock layer in the site, measured by on-site thermal response testing (TRT) and other exploration methods.

[0078] Economic and cost data: including the engineering cost per unit of borehole, pipe material cost, and the time-of-use electricity pricing structure of the project location.

[0079] Optimization variables: During simulation, the algorithm will optimize the following key design parameters that have a decisive impact on system performance and cost: Macro-level layout plan: including the number of functional zones, the area of ​​each zone, and the boundary division.

[0080] Microscopic design parameters: including the specific drilling depth and pipe spacing within each zone.

[0081] Optimization Objective: The core objective of this simulation optimization is to find a set of optimal variables that minimizes the system's total lifecycle cost (LCC). Mathematically, this objective function can be expressed as solving the following cost function: Minimize LCC = C_initial + C_operation; Among them: Initial Investment Cost (C_initial): This part of the cost mainly depends on the scale of the underground heat exchanger project and is a direct function of the total number of boreholes (i.e., borehole depth, number, and pipe spacing).

[0082] Lifecycle Operating Cost (C_operation): This cost is calculated by establishing a dynamic simulation model that couples the performance of the buried pipe heat exchanger and the heat pump unit. The model simulates the hourly energy consumption of each candidate physical layout scheme over a long period (e.g., 20 years), and then incorporates local electricity pricing policies to arrive at the total cost. This simulation process can accurately predict the long-term evolution trend of the geothermal field under different layout schemes and its impact on system energy efficiency (COP / EER).

[0083] To further clarify the technical solution of this invention, it should be noted that the system described in this invention comprises two core models applied to different stages throughout its complete lifecycle: a full lifecycle simulation optimization model applied to the offline design stage, and a hybrid-driven prediction model applied to the online operation stage. The aforementioned hybrid-driven prediction model is used in the online operation stage, where it performs short-term future ground temperature predictions based on real-time ground temperature data to generate optimal real-time operation control commands. The simulation optimization model, which is of interest to customers, is formally called the full lifecycle simulation optimization model, and it is applied to the offline design stage, as detailed below.

[0084] In one embodiment, the full life cycle simulation optimization model is a dynamic simulation system that couples the performance of the buried pipe heat exchanger, the heat pump unit, and the building load. Its purpose is to evaluate and determine the optimal physical structure that can achieve the lowest overall cost throughout the entire life cycle by conducting long-term virtual operation on multiple candidate physical layout schemes before the construction of the buried pipe heat exchanger system.

[0085] The core objective function of this full lifecycle simulation optimization model is to find a set of physical layout parameters that minimizes the system's total lifecycle cost (LCC), and its mathematical expression is: minLCC=C initial +C operation ; Wherein, LCC stands for Total Life Cycle Cost, and C initial For initial investment costs, C operation Cost of operation over the entire lifecycle.

[0086] The initial investment cost C initial The total borehole depth is mainly determined by the scale of the underground heat exchanger project and is a direct function of the total borehole length. Specifically, it depends on the physical layout parameters such as the borehole depth, pipe spacing, and number of boreholes in each zone.

[0087] The total lifecycle operating cost C operation The calculations are performed using a core simulation engine. This engine simulates the hourly energy consumption of the system over a long period (e.g., twenty years) with hourly increments, and then combines this with time-of-use pricing to arrive at the final result. The mathematical expression is as follows: Among them, P k Let Price be the total power consumption of the system in hour k. k Let t be the time-of-use electricity price for the kth hour, Δt be the time step (1 hour), and LifecycleHours be the total number of hours in the entire lifecycle.

[0088] The core simulation engine iteratively calculates the total power consumption P of the system at each time step k through the following coupled steps. k : Obtain the building's cooling or heating load L in the kth hour k As known input.

[0089] The instantaneous power of a heat pump unit is calculated based on a heat pump performance model. The coefficient of performance (EER for cooling and COP for heating) of the heat pump unit is the outlet temperature of the circulating fluid in the buried pipe, i.e., the average underground temperature T. g,k The function is established through polynomial fitting and other methods. The instantaneous power P of the heat pump unit... k Calculated by the following formula: and Among them, P k L represents a certain type of power or performance indicator. kThe load or metric of the k-th device or system is represented by COP (Coefficient of Performance), EER (Energy Efficiency Ratio) represents the energy efficiency of refrigeration equipment when handling cooling load, and TP... k T represents a temperature or performance index related to k. g,k This represents a certain temperature associated with the k-th device.

[0090] Update the underground average temperature T based on the buried pipe heat transfer model (GHE Model) g,k Average underground temperature T g,k The evolution depends on the temperature T at the previous moment. g,k-1 And the heat Q discharged or extracted from the heat pump unit per hour. g,k , where Q g,k With building load L k It is directly related to the performance of the generating unit. This evolution process is simulated by a heat conduction physics model, and its mathematical form can be abstracted as follows: Among them, T g,k T represents the temperature value at the k-th stage or location. g,k-1 Q represents the temperature value at the previous stage or position (k-1). g,k This represents the heat or energy input associated with the k-th stage or location. This is the evolution function for long-term heat exchange of buried pipes. Layout parameters refer to parameters related to the system layout, and rock parameters refer to parameters related to the properties of soil and rock.

[0091] For each candidate physical layout scheme, the full lifecycle simulation optimization model will perform the aforementioned long-term dynamic simulation to calculate its corresponding total lifecycle cost (LCC). By comparing the LCC values ​​of different layout schemes, the optimal physical layout scheme is finally determined and output.

Claims

1. A ground source heat pump ground heat exchanger zoning and cold and heat balance control method, characterized by, The method comprises the following steps: S1: obtaining the current three-dimensional ground temperature field state of the area where the ground source heat pump borehole is located; S2: based on a prediction model capable of representing the dynamic evolution of the ground temperature field, and according to the current three-dimensional ground temperature field state, predicting the future state sequence of the ground temperature field in a prediction time domain in the future under different control strategies; S3: based on a multi-objective optimization function with the minimum system comprehensive operation cost in the prediction time domain as the target, an optimal control sequence is obtained; the multi-objective optimization function at least includes an energy consumption term representing the short-term operation energy consumption of the system and a balance term representing the long-term balance state of the ground temperature; S4: executing the first control action in the optimal control sequence to control the operation of the ground source heat pump system.

2. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 1, characterized in that, The prediction model in step S2 is a hybrid driving model, which comprises: a physical model based on heat conduction physical mechanism, used to predict the main change trend of the ground temperature field; a residual compensation model based on historical data driving, used to learn and compensate the prediction error of the physical model.

3. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 2, characterized in that, The step S2 further comprises an online self-correction step: After each control cycle, the actual measured value of the ground temperature field is compared with the predicted value of the physical model to obtain new residual data, and the residual compensation model is updated online using the new residual data.

4. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 1, characterized in that, In step S3, the weights of each sub-term in the multi-objective optimization function are dynamically adjusted; In step S3, it further comprises: According to the current three-dimensional ground temperature field state, a ground temperature health index is calculated to quantify the degree of ground temperature imbalance; And according to the ground temperature health index, the weights of the energy consumption term and the balance term in the multi-objective optimization function are dynamically adjusted.

5. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 4, characterized in that, The calculation of the ground temperature health index is based on at least one or a combination of the following: deviation between the spatial average temperature of the current ground temperature field and the spatial average temperature of the initial ground temperature field; spatial distribution variance of the current ground temperature field.

6. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 1, characterized in that, The control method is a model predictive control method, and the method repeatedly executes steps S1 to S4 in a rolling time domain.

7. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 1, characterized in that, The method further comprises an offline design step executed before the construction of the borehole, which comprises: determining the physical layout scheme of the borehole through life cycle simulation optimization; The process of the life cycle simulation optimization comprises different candidate physical layout schemes.

8. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 7, characterized in that, In the offline design step, the physical layout scheme of the borehole is a zoning layout scheme; The simulation optimization aims to determine the optimal combination of physical parameters of each zone, including at least drilling depth and pipe spacing.

9. The ground-source heat pump ground loop zoning layout and cold-heat balance control method according to claim 8, characterized in that, The zoning layout scheme at least includes two functionally differentiated zones, and adopts a gradient arrangement scheme; The gradient arrangement scheme is characterized by the fact that the drilling depth and / or pipe spacing gradually increases from the peripheral zone to the core zone, which is used to actively optimize the long-term heat exchange performance of the core area.

10. The ground-source heat pump ground loop zoning and thermal balance control method of claim 9, wherein, In the region dominated by refrigeration, the physical layout scheme determined through the simulation optimization has the following characteristics: The inner zone has a tube spacing greater than the tube spacing of the outer zone and a borehole depth greater than the borehole depth of the outer zone for cold energy extraction and long-term heat sequestration of the core region.

Images