Heat pump collaborative cooling recovery and hot water temperature intelligent regulation and control method

Abstract

The invention relates to the technical field of heat pump regulation and control, and discloses a heat pump collaborative cooling recovery and hot water temperature intelligent regulation and control method. The method comprises the steps of detecting cooling medium flow parameters and hot water tank temperature parameters, and establishing a dynamic correlation model; a cooling recovery stage and a hot water heating stage are divided, heat exchange characteristic parameters are extracted, and recoverable heat and heat needing to be supplemented are calculated; training a heat transfer prediction model, and generating a flow and power regulation instruction; executing deviation is monitored in real time, and model parameters are corrected; acquiring actual temperature distribution data of the hot water tank, reversely calibrating a heat exchange coefficient of the dynamic correlation model, and updating a recoverable heat calculation logic; and the updated calculation logic and the heat transfer prediction model are subjected to coupling iteration, a final heat pump cooperative regulation and control parameter set is output, closed-loop control over heat pump system cooling recovery and hot water heating is achieved, the energy utilization efficiency is improved, and the hot water temperature is precisely regulated and controlled.

Description

Technical Field

[0001] This invention relates to the field of heat pump control technology, specifically a method for heat pump synergistic cooling recovery and intelligent control of hot water temperature. Background Technology

[0002] Traditional heat pump systems typically operate independently in terms of cooling, heating, and hot water supply. Taking summer cooling as an example, the heat pump system runs at full capacity to meet indoor cooling needs; however, the large amount of condensation heat generated during this process is not effectively utilized, failing to simultaneously and efficiently provide heat for domestic hot water supply. This necessitates the operation of additional hot water heating equipment if domestic hot water is also needed, increasing equipment costs and wasting energy. For instance, in some hotels or large commercial buildings, air conditioning systems operate at full capacity in summer to maintain indoor coolness, but the supply of domestic hot water relies on separate gas or electric water heaters, failing to fully utilize the waste heat generated by the air conditioning system.

[0003] In the operation of traditional heat pump systems, the heat carried by the cooling medium is often overlooked. After the cooling medium completes heat exchange, a large amount of waste heat is directly released into the environment without any recovery or utilization. Similarly, in the hot water heating stage, the planning for heat replenishment lacks rationality and fails to fully consider the usable heat that may be generated in other parts of the system, resulting in a large amount of energy being wasted. Taking heat pump systems in industrial production as an example, in some processes, if the heat carried away by the cooling medium could be recovered and used to preheat raw materials or other heat-required processes, energy consumption could be significantly reduced, but traditional systems have failed to achieve this goal.

[0004] Traditional heat pump systems have significant shortcomings in hot water temperature control, making precise control difficult. They often only operate within a preset, simple temperature range, unable to flexibly adjust heating power and time based on changes in actual water demand or fluctuations in ambient temperature. This results in either excessively high hot water temperatures, leading to excessive energy consumption and water waste, or excessively low temperatures, failing to meet user needs and negatively impacting the user experience. For example, in household use, when the number of family members suddenly increases or water usage habits change, traditional heat pump systems cannot adjust the hot water temperature promptly, causing instability and inconvenience to users.

[0005] While current research on heat pump-assisted cooling recovery and hot water temperature control has made some progress, many problems remain. Regarding the accuracy of heat calculation models, existing models often oversimplify the actual heat exchange process, failing to fully consider the influence of various complex factors during system operation, such as pipeline heat loss and variations in heat exchange efficiency under different operating conditions. This leads to significant discrepancies between the recoverable heat and the required heat replenishment calculated based on these models and the actual values, making it impossible to provide accurate data support for the system's optimized control.

[0006] In predicting heat transfer efficiency, most existing studies are based on ideal operating conditions, neglecting the impact of factors such as equipment aging and changes in environmental conditions on heat transfer efficiency during actual operation. Therefore, the predicted results differ significantly from the actual heat transfer efficiency, making it difficult to accurately grasp the heat transfer situation in practical applications and hindering efficient system control.

[0007] From the perspective of system dynamic adjustment capabilities, existing research often lacks flexible and timely control strategies to cope with sudden changes or adjustments in operating conditions during system operation. When the system is subjected to external disturbances or changes in internal parameters, it cannot quickly and effectively adjust operating parameters to ensure stable operation and high efficiency. For example, when extreme weather causes a sudden and drastic change in ambient temperature, existing heat pump system control strategies may not be able to adjust the cooling recovery and hot water heating operation modes in a timely manner, thus affecting the overall system performance and energy utilization efficiency. Summary of the Invention

[0008] The purpose of this invention is to provide a method for heat pump-assisted cooling recovery and intelligent control of hot water temperature to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides a method for heat pump synergistic cooling recovery and intelligent hot water temperature control, the method comprising: The flow rate parameters of the cooling medium and the temperature parameters of the hot water tank are detected during the operation of the heat pump system, and a dynamic correlation model between the flow rate of the cooling medium and the temperature of the hot water tank is established. Based on the operating mode of the heat pump system, the cooling recovery stage and the hot water heating stage are divided, and the heat exchange characteristic parameters of the two stages are extracted respectively. Based on the dynamic correlation model and heat exchange characteristic parameters, the recoverable heat in the cooling recovery stage and the additional heat required in the hot water heating stage are calculated. A heat transfer prediction model is trained using historical operating data of the heat pump system to predict the heat transfer efficiency from the cooling recovery stage to the hot water heating stage. Based on the recoverable heat, the heat to be replenished, and the heat transfer efficiency, cooling medium flow rate regulation commands and heat pump compressor power regulation commands are generated. Real-time monitoring of the execution deviation between the cooling medium flow regulation command and the heat pump compressor power regulation command, and dynamic correction of the output parameters of the heat transfer prediction model; At the end of the hot water heating stage, the actual temperature distribution data of the hot water tank is collected, the heat exchange coefficient of the dynamic correlation model is calibrated in reverse, and the calculation logic of recoverable heat in the cooling recovery stage is updated based on the calibrated heat exchange coefficient. The updated recyclable heat calculation logic is coupled and iterated with the heat transfer prediction model to output the final heat pump collaborative control parameter set. Based on the final heat pump collaborative control parameter set, the cooling recovery and hot water heating of the heat pump system are controlled in a closed loop.

[0010] Preferably, the establishment of the dynamic correlation model between the cooling medium flow rate and the hot water tank temperature includes: Collect the instantaneous values ​​of the cooling medium flow rate and the corresponding temperature gradient changes of the hot water tank under different load rates of the heat pump system; Identify the nonlinear mapping relationship between the instantaneous value of the cooling medium flow rate and the change in the temperature gradient of the hot water tank; The terminal difference parameter between the heat pump evaporator and condenser is introduced as a compensation variable in the dynamic correlation model; By statistically analyzing the weight distribution of compensation variables through a sliding time window, a dynamic correlation model with weight correction is constructed.

[0011] Preferably, the extraction of heat exchange characteristic parameters in the two stages includes: During the cooling recovery phase, the subcooling of the cooling medium on the evaporator side and the rate of decrease in saturation temperature are captured. During the hot water heating stage, the temperature rise lag time and stratification temperature difference of the hot water tank on the condenser side were recorded. The supercooling, saturation temperature decrease rate, temperature rise lag time, and stratified temperature difference are normalized into a heat exchange characteristic matrix.

[0012] Preferably, the calculation of recoverable heat in the cooling recovery stage and the required supplementary heat in the hot water heating stage includes: Based on the subcooling and saturation temperature drop rate in the heat exchange characteristic matrix, the sensible heat recovery potential value of the evaporator is derived. By combining the temperature difference between the layers in the hot water tank and the temperature rise lag time, the latent heat demand gap of the condenser can be calculated. The sensible heat recovery potential value is converted into recoverable heat value through a dynamic correlation model, and the latent heat demand gap value is converted into heat that needs to be supplemented.

[0013] Preferably, the training heat transfer prediction model includes: Extract heat overflow events during the cooling recovery phase and heat shortage events during the hot water heating phase from historical operation data; Establish time-series matching relationships between heat overflow events and heat shortage events; By employing a bidirectional long short-term memory network to learn the transfer decay characteristics of time series matching relationships, a time-series predicted value of heat transfer efficiency is generated.

[0014] Preferably, the generation of cooling medium flow rate regulation command and heat pump compressor power regulation command includes: Multiply the recoverable heat by the time-series predicted value of the heat transfer efficiency to obtain the theoretical adjustment reference value of the cooling medium flow rate; Calculate the power compensation amount of the heat pump compressor based on the difference between the required heat supply and the theoretical adjustment benchmark value; Using the set value of the hot water tank temperature parameter as a constraint, boundary corrections are made to the theoretical adjustment reference value and the power compensation amount to generate adjustment commands.

[0015] Preferably, the output parameters of the dynamically corrected heat transfer prediction model include: Compare the cumulative error between the theoretical execution value of the cooling medium flow rate regulation command and the actual sensor feedback value; When the cumulative error exceeds the threshold, the attenuation factor of the heat transfer efficiency is recalculated; The attenuation factor is injected into the hidden layer nodes of the heat transfer prediction model to update the generation logic of the time series prediction values.

[0016] Preferably, the heat exchange coefficient of the reverse calibration dynamic correlation model includes: Based on the actual temperature distribution data of the hot water tank, the heat transfer efficiency curve on the condenser side is reconstructed. Extract the abrupt change points in the slope of the heat transfer efficiency curve and correlate them with evaporator pressure fluctuation data during the cooling recovery stage; By analyzing the correlation between the slope abrupt change point and pressure fluctuation data, the initial value of the heat exchange coefficient of the dynamic correlation model was adjusted.

[0017] Preferably, the logic for calculating recoverable heat in the updated cooling recovery stage includes: The process of substituting the adjusted initial value of the heat exchange coefficient into the derivation of the sensible heat recovery potential value; An adaptive filtering algorithm is used to eliminate high-frequency noise components in the initial value of the heat exchange coefficient. The calculation path for reconstructing recoverable heat is based on the filtered heat exchange coefficient.

[0018] Preferably, the coupling iteration process includes: The reconstructed recoverable heat calculation path and the heat transfer prediction model with the updated time-series prediction value generation logic are run in parallel. The intermediate parameters of the two are exchanged and a consistency check is performed at preset intervals; When the consistency check passes, the current parameter set is locked as the final heat pump collaborative control parameter set.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This invention establishes a dynamic correlation model between the cooling medium flow rate and the hot water tank temperature, enabling a comprehensive and detailed consideration of various complex factors during the operation of a heat pump system. In actual operation, the cooling medium flow rate is affected by multiple factors, such as system load changes and ambient temperature fluctuations, while the hot water tank temperature constantly changes due to water usage and insulation performance. This dynamic correlation model can accurately capture these changes and, combined with heat exchange characteristic parameters, precisely calculate the recoverable heat during the cooling recovery stage. Previously overlooked waste heat can now be effectively collected. For example, in industrial heat pump systems, the large amount of waste heat carried by the cooling medium is no longer directly discharged but is accurately calculated and recovered, avoiding unnecessary energy waste.

[0020] Based on the precise calculations above, the system can accurately replenish the required heat during the hot water heating stage. It's no longer the previous method of blindly heating, but rather supplying heat according to actual demand. When the water temperature in the tank drops to a certain level, the system will adjust the heating power and time appropriately based on the previously calculated heat replenishment, ensuring a precise match between heat supply and demand. This guarantees that every unit of energy is fully utilized, avoids excessive energy consumption, and maximizes energy efficiency.

[0021] A key innovation of this invention is the heat transfer prediction model. It learns the system's operating patterns under different conditions through in-depth mining and analysis of historical operating data of the heat pump system. This historical operating data contains rich information, such as heat transfer patterns under different seasons, time periods, and load conditions. Based on this data, the model can accurately predict the heat transfer efficiency from the cooling recovery stage to the hot water heating stage. According to the prediction results, the system can generate precise cooling medium flow rate adjustment commands and heat pump compressor power adjustment commands. During high summer temperatures, when the system determines that the heat transfer efficiency is high based on the prediction model, it will adjust the cooling medium flow rate and compressor power accordingly to achieve efficient heat transfer and utilization, thereby realizing intelligent system operation and greatly improving the system's operating efficiency and response speed.

[0022] During system operation, real-time monitoring of the execution deviation between the cooling medium flow rate adjustment command and the heat pump compressor power adjustment command is crucial. Once a deviation is detected, the system quickly and dynamically corrects the output parameters of the heat transfer prediction model. When the actual cooling medium flow rate deviates from the commanded flow rate, the system analyzes the cause, which could be due to changes in pipe resistance, valve malfunction, etc., and then adjusts the model parameters accordingly to more accurately reflect the actual operating state of the system. At the end of the hot water heating stage, the actual temperature distribution data of the hot water tank is collected, and the heat exchange coefficient of the dynamic correlation model is calibrated in reverse, thereby updating the recoverable heat calculation logic for the cooling recovery stage. Through this series of dynamic adjustments, the system can continuously optimize its performance based on actual operating conditions, maintaining a highly efficient operating state and adapting to various complex and changing operating environments. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating the working principle of the heat pump synergistic cooling recovery and intelligent hot water temperature control method described in this invention. Figure 2 A flowchart for establishing a dynamic correlation model between cooling medium flow rate and hot water tank temperature; Figure 3 This is a flowchart for calculating the recoverable heat in the cooling recovery stage and the additional heat required in the hot water heating stage. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1This invention provides a method for coordinated cooling recovery and intelligent hot water temperature control in a heat pump system. The method integrates real-time detection, model building, and intelligent control mechanisms to achieve efficient operation of the heat pump system. The method detects the cooling medium flow rate and hot water tank temperature parameters during heat pump system operation, establishing a dynamic correlation model between the cooling medium flow rate and hot water tank temperature. This dynamic correlation model is continuously optimized based on real-time data flow to reflect system state changes. The system is divided into a cooling recovery stage and a hot water heating stage according to its operating mode, and heat exchange characteristic parameters are extracted for each stage. These parameters capture key thermodynamic variables for subsequent calculations. Based on the dynamic correlation model and heat exchange characteristic parameters, the recoverable heat in the cooling recovery stage and the required heat in the hot water heating stage are calculated. These recoverable and required heat serve as control benchmarks to guide energy transfer. A heat transfer prediction model is trained using historical operating data of the heat pump system to predict the heat transfer efficiency from the cooling recovery stage to the hot water heating stage. The heat transfer prediction model uses machine learning algorithms to learn historical patterns. Based on the recoverable heat, required heat, and heat transfer efficiency, cooling medium flow rate adjustment commands and heat pump compressor power adjustment commands are generated. These commands are then sent to the actuators via the controller. The system monitors the execution deviation between the cooling medium flow rate regulation command and the heat pump compressor power regulation command in real time, dynamically correcting the output parameters of the heat transfer prediction model. This correction process ensures the model is synchronized with actual operation. At the end of the hot water heating phase, actual temperature distribution data of the hot water tank is collected, and the heat exchange coefficient of the dynamic correlation model is calibrated in reverse. Based on the calibrated heat exchange coefficient, the recoverable heat calculation logic for the cooling recovery phase is updated, improving model accuracy. The updated recoverable heat calculation logic is coupled and iteratively integrated with the heat transfer prediction model, outputting the final heat pump coordinated control parameter set. Based on this final parameter set, closed-loop control is implemented for the cooling recovery and hot water heating of the heat pump system, achieving system autonomous optimization.

[0026] Example 1: See Figure 2The system collects instantaneous values ​​of cooling medium flow rate and corresponding hot water tank temperature gradient changes under different load rates to form the basic data stream. The instantaneous cooling medium flow rate is acquired through a high-frequency sampling sensor with a sampling interval set to milliseconds to capture dynamic fluctuations. The hot water tank temperature gradient changes are calculated from a distributed temperature sensor array, reflecting the temporal evolution of heat distribution within the hot water tank. Load rate changes simulate actual operating conditions, with load rates set in segments from zero to full load. Each load segment runs continuously for a sufficient time to stabilize the data. The instantaneous cooling medium flow rate is recorded as time-series data points, stored in a circular buffer for real-time processing. The hot water tank temperature gradient changes are calculated using a difference algorithm to determine the temperature difference between adjacent time points, employing the central difference method to reduce errors. Identifying the nonlinear mapping relationship between the instantaneous cooling medium flow rate and the hot water tank temperature gradient changes is the core step in model construction. This nonlinear mapping relationship is fitted using a polynomial regression model, with the polynomial order determined based on the data scatter plot. The fitting process uses the least squares method to optimize parameters, and the least squares iterative solution minimizes the sum of squared residuals. The instantaneous flow rate of the cooling medium was used as the independent variable, and the temperature gradient change in the hot water tank was used as the dependent variable. Data preprocessing included outlier removal and noise filtering. Outlier detection employed the Z-score method, and noise filtering used a low-pass filter to smooth high-frequency fluctuations. Nonlinear mapping relationships were validated using cross-validation. The dataset was divided into training and test sets. The training set was used for model training, and the test set was used to evaluate goodness of fit. Goodness-of-fit metrics included R-squared and root mean square error.

[0027] The terminal temperature difference parameters between the evaporator and condenser of the heat pump are introduced as compensation variables in the dynamic correlation model to enhance its adaptability. These parameters are defined as the differences between the evaporator outlet temperature and the condenser inlet temperature, respectively. The terminal temperature difference parameters are directly measured from temperature sensors, and the measured values ​​are input into the model in real time. The compensation variables are integrated into the structure of the dynamic correlation model, which employs a multiple linear regression framework. The terminal temperature difference parameters are added as additional independent variables to the regression equation, and the coefficients of the regression equation are learned from training data. The introduction of compensation variables addresses non-ideal factors in the system, such as heat transfer losses and fluid dynamic effects. Changes in the terminal temperature difference parameters reflect fluctuations in heat exchange efficiency, and the fluctuation data are analyzed using time-series data to identify patterns. Dynamic adjustment of the model is achieved by statistically analyzing the weight distribution of the compensation variables using a sliding time window. The sliding time window size is set as a multiple of the system's thermal time constant, which is derived from system response experiments. The window sliding step size is synchronized with the data sampling rate. The weight distribution calculation is based on the moving average method, which smooths short-term fluctuations and highlights long-term trends. Weight values ​​are assigned to each terminal temperature difference parameter data point, and these weight values ​​are derived from the correlation in historical data. A weighted dynamic correlation model is constructed, in which coefficients are updated with each window slide. The coefficient update employs recursive least squares, which reduces computational complexity and adapts to real-time applications. The model output includes predicted hot water tank temperature gradient changes. Comparison between predicted and measured values ​​is used for model calibration, and the weight distribution is automatically adjusted during the calibration process.

[0028] During the cooling recovery phase, capturing the subcooling and saturation temperature drop rate of the cooling medium on the evaporator side involves precise measurements. Subcooling is measured jointly by temperature and pressure sensors installed at the evaporator outlet. The temperature sensor has an accuracy of 0.1 degrees Celsius, and the pressure sensor measures the saturation pressure, which is converted to saturation temperature using the refrigerant property table. The saturation temperature drop rate is calculated from the saturation temperature time series using a first-derivative approximation. The derivative calculation uses a numerical difference method. Subcooling is defined as the difference between the actual temperature and the saturation temperature, and this difference is monitored in real time to detect system status. Data acquisition during the cooling recovery phase is synchronized with the heat pump refrigeration cycle. Evaporator-side parameters are recorded once per second, and the data stream is continuously transmitted to the processing unit. Abnormal subcooling values, such as negative values, trigger a quality check routine. During the hot water heating phase, recording the temperature rise lag time and stratification temperature difference of the hot water tank on the condenser side requires multi-point measurements. The temperature rise lag time is defined as the delay between the time when the hot water tank temperature begins to rise and the heating start time. The delay is calculated from timestamp data, and the heating start signal comes from the heat pump controller. The stratified temperature difference is obtained through a vertically arranged array of temperature sensors inside the hot water tank. The sensor array is evenly spaced to cover the height of the tank. The stratified temperature difference is taken as the absolute value of the maximum temperature difference, and the sampling frequency of the temperature difference is matched with the heating stage. The parameter recording during the hot water heating stage is coordinated with the heat pump heating cycle. The temperature rise lag time is affected by the water flow rate and heating power. The stratified temperature difference reflects the intensity of thermal stratification, and the intensity of thermal stratification affects the heating efficiency.

[0029] Supercooling, saturation temperature decrease rate, temperature rise lag time, and stratified temperature difference are normalized into a heat exchange feature matrix to achieve data standardization. The normalization process uses the Z-score standardization method, which converts each parameter to zero mean and unit variance. The conversion formula is applied but not listed. The dimension of the heat exchange feature matrix is ​​consistent with the number of parameters; matrix rows correspond to time points, and columns correspond to parameter values. Matrix elements are scaled to eliminate the influence of dimensions. The normalized heat exchange feature matrix serves as a unified input for subsequent models. Matrix data is stored in a multidimensional array, and array operations utilize a linear algebra library. The update frequency of the heat exchange feature matrix is ​​synchronized with data acquisition. The construction of the feature matrix facilitates comparison and integration between different parameters. Matrix data is used for machine learning algorithm training, which includes feature selection and dimensionality reduction. The generation of the heat exchange feature matrix relies on a real-time data pipeline, which includes three stages: acquisition, preprocessing, and normalization. The acquisition stage uses a sensor network, the preprocessing stage denoises and imputes missing values, and the normalization stage dynamically adjusts the parameter range. A matrix data stream is continuously input into a dynamic correlation model, which uses the feature matrix to optimize predictions. The prediction results are then fed back to control the heat pump system. Data integration between the cooling recovery and hot water heating stages is achieved through time alignment based on a unified clock source, ensuring smooth stage transitions. The application of the heat exchange feature matrix is ​​extended to system diagnostics, with matrix pattern recognition identifying abnormal operating states and triggering alarm mechanisms in response.

[0030] The acquisition of instantaneous values ​​of cooling medium flow rate and changes in hot water tank temperature gradient covers the entire operating range. Load rate changes are simulated through automatic control, with simulation conditions including ambient temperature changes and sudden load changes. Identification of nonlinear mapping relationships utilizes advanced statistical tools integrated into embedded software that runs in real-time on the heat pump controller. Terminal difference parameters are introduced as compensation variables to improve model robustness; terminal difference parameter measurement avoids single points of failure, and redundant verification of measured values ​​is performed. A statistical method using sliding time windows adapts to system inertia; the window size is adaptively adjusted based on operating history, and weight distribution is automatically derived from the data. High-precision sensing is required for capturing subcooling during the cooling recovery phase; subcooling measurement considers sensor drift, which is compensated for through periodic calibration. The saturation temperature drop rate calculation uses sliding window averaging to reduce noise. Recording the temperature rise lag time during the hot water heating phase relies on time synchronization, which is obtained from a network time protocol. Layered temperature difference measurement uses high-resolution sensors, and sensor data fusion improves accuracy. The generation of a heat exchange characteristic matrix for the normalized process ensures data consistency; the matrix is ​​used for multivariate analysis, and the analysis results guide control decisions.

[0031] In the data acquisition phase, instantaneous cooling medium flow rate is measured using an electromagnetic flowmeter. The flowmeter is installed in the cooling medium pipeline; pipeline size and fluid properties affect measurement accuracy, which is calibrated by comparing with a benchmark flowmeter. The calculation of hot water tank temperature gradient changes uses a temperature sensor array. The array layout optimizes spatial coverage, and the temperature sampling frequency matches the thermal dynamic response. Load rate control is achieved through an external load simulator, which generates step and ramp changes, covering typical operating scenarios. A nonlinear mapping identification algorithm is deployed on an edge computing device, whose processing power meets real-time requirements. Algorithm parameter tuning is based on historical data. The introduction of terminal difference parameters addresses the nonlinear behavior of the heat pump system; evaporator and condenser terminal differences are calculated separately, and the calculated values ​​update the model in real time. A sliding time window is implemented using a circular buffer with configurable size. Weight calculation employs an exponentially weighted moving average, assigning higher weight to recent data. A weighted dynamic correlation model outputs predicted values, which are used for feedforward control to reduce feedback delay. In the cooling recovery phase, subcooling monitoring integrates fault detection; excessively low subcooling indicates insufficient refrigerant, and an excessively rapid rate of saturation temperature drop indicates a sudden load change. The temperature rise lag time analysis during the hot water heating stage considers the insulation effect of the water tank, which affects the lag time length. Layered temperature difference monitoring identifies stirring needs, which trigger the mixing pump. The generation of the normalized heat exchange characteristic matrix is ​​automated, and the matrix data is visualized for operator monitoring, with the monitoring interface displaying real-time trends.

[0032] Example 2: See Figure 3 The calculation begins with the sensible heat recovery potential of the evaporator, derived from the subcooling and saturation temperature drop rate in the heat exchange characteristic matrix. Subcooling data is obtained from precise measurements at the evaporator outlet, and the saturation temperature drop rate is derived from readings of the saturation pressure sensor, based on the thermodynamic properties of the refrigerant. The derivation of the sensible heat recovery potential is based on classical thermodynamic formulas, involving the mass flow rate, specific heat capacity, and temperature difference represented by the subcooling of the cooling medium. The mass flow rate is provided in real-time by a flow meter, and the specific heat capacity is retrieved from the built-in database as a medium property parameter. The saturation temperature drop rate, as a dynamic variable, reflects the instantaneous change in heat exchange intensity within the evaporator. This change is directly related to system load fluctuations, and the calculation process requires integrating the drop rate over a period of time to account for the cumulative effect.

[0033] The process of inversely deducing the latent heat demand gap of the condenser by combining the stratified temperature difference and temperature rise lag time of the hot water tank is a reverse derivation process. The stratified temperature difference is measured by an array of temperature sensors placed at different heights in the hot water tank. The data from the temperature sensor array is fused to generate a three-dimensional temperature field. The temperature rise lag time is defined as the time interval from the start of the heat pump compressor to the detection of a significant temperature rise at the inlet of the hot water tank. This time interval captures the magnitude of the system's thermal inertia. The inverse derivation of the latent heat demand gap relies on the principle of energy conservation. It requires estimating the difference between the theoretical energy required to heat the hot water to the set temperature and the actual energy supplied. The theoretical energy calculation considers the volume of the hot water tank and the specific heat capacity of water, while the actual energy supplied is derived from the condenser's operating parameters. The stratified temperature difference of the hot water tank indicates the uniformity of heat distribution within the tank. A larger stratified temperature difference means that additional energy is needed for mixing to achieve temperature equilibrium. The temperature rise lag time reflects the delayed effect of heat transfer from the condenser to the hot water tank. The application of the model is to convert the sensible heat recovery potential into recoverable heat through a dynamic correlation model. The internal coefficients of the dynamic correlation model have been optimized using training data. The conversion process involves using the sensible heat recovery potential as input to the dynamic correlation model. The model output is the quantified recoverable heat value, which represents the heat that can be effectively collected and stored during the cooling recovery stage. The latent heat demand gap is then converted into supplementary heat, completing another direction of quantification. This latent heat demand gap is also input into the dynamic correlation model, which outputs the supplementary heat requirement based on its learned system characteristics. The supplementary heat requirement represents the additional energy needed to compensate for the gap during the hot water heating stage. The dynamic correlation model acts as a translator in this conversion, mapping feature parameters from different stages and with different physical meanings to a unified energy metric.

[0034] The calculation of sensible heat recovery potential requires high-frequency data updates. The calculation module performs a complete derivation every second, and a moving average filter is used to smooth the subcooling data to reduce the impact of random fluctuations. The calculation of the saturation temperature drop rate relies on high-precision timestamps, making system clock synchronization crucial. The rate value is obtained by calculating the first derivative of the saturation temperature at consecutive time points. The derived sensible heat recovery potential is temporarily stored in a buffer, awaiting invocation by the dynamic correlation model. The buffer size is set to account for pipeline latency in data processing. The back-calculation process for the latent heat demand gap integrates historical data. The stratified temperature difference in the hot water tank is not an instantaneous value but is based on statistical characteristics within a time window, such as average or maximum values. The calculation of the temperature rise lag time requires precise identification of the heating start moment and the temperature response moment, which is achieved using a threshold comparison algorithm. The back-calculation uses an iterative method to approximate the latent heat demand gap value. In each iteration, the estimated gap value is adjusted until the calculated heat demand matches the observed temperature change. The back-calculation results undergo a plausibility check, such as verifying that the gap value is within a physically possible range, before being passed to the dynamic correlation model. The dynamic correlation model's conversion function is implemented through its internal weight matrix, which is determined during the model training phase. The training data contains a large number of correspondences between sensible heat recovery potential values ​​and actual recoverable heat. The conversion calculation can be viewed as a matrix multiplication operation; the vector of sensible heat recovery potential values ​​is multiplied by the weight matrix to obtain a scalar of recoverable heat. For conversions requiring additional heat, the dynamic correlation model uses another set of weight parameters, which focuses on learning the complex mapping between latent heat demand gap values ​​and actual additional heat required. The real-time nature of the model conversion requires the conversion calculation to be completed within milliseconds; therefore, the dynamic correlation model typically runs in optimized code on high-performance embedded processors.

[0035] Recoverable heat and required heat serve as the basis for generating subsequent control commands, and the accuracy of these two heat values ​​directly affects the energy efficiency of the entire system. The calculation logic for recoverable heat continuously receives real-time data streams from the cooling recovery stage, and the data streams drive the calculation logic to dynamically update its output. The calculation of required heat is synchronized with the hot water heating stage; each time a new heating cycle begins, the required heat is re-initialized and calculated. The dynamic correlation model records conversion errors during operation, i.e., the differences between the model's predicted recoverable heat and required heat and the actual observed heat values ​​in subsequent stages. These error data are collected for periodic retraining of the model, thus forming a self-improving cycle. Subcooling and saturation temperature drop rate serve as status indicators on the evaporator side, and their accuracy is ensured through sensor calibration, with the calibration cycle set according to the operating time. Stratified temperature difference and temperature rise lag time serve as status indicators on the condenser side and the hot water tank side. Their measurement is affected by the tank structure and fluid flow, and the measurement scheme is optimized during the system design phase. The calculation of the sensible heat recovery potential and the latent heat demand gap introduced a physical model, but was ultimately fine-tuned through a data-driven dynamic correlation model. This combination approach utilizes the constraints of physical laws while incorporating the experience of actual operating data.

[0036] The calculation process relies on a dedicated software module that receives data packets from the sensor network, containing timestamps and parameter values. The sensible heat recovery potential derivation module first verifies the validity of the subcooling data; invalid data is discarded and replaced with the previous valid value. The saturation temperature drop rate calculation module monitors data continuity and activates an interpolation algorithm when data loss occurs. The latent heat demand gap calculation module requires the initial temperature data of the hot water tank, obtained from sensor readings at the start of the heating phase. The dynamic correlation model conversion module, as the core, has optimized processing speed, and the conversion results are immediately published to the system's message bus for other modules to consume. The entire heat calculation implementation is tightly coupled with the heat pump system's control cycle, which determines the rhythm of data acquisition and calculation. The calculation results of recoverable heat are mainly used to guide the adjustment of the cooling medium flow rate, while the calculation results of the required supplemental heat are used to determine the power compensation needed by the heat pump compressor.

[0037] Example 3: Training and Control Command Generation of the Heat Transfer Prediction Model. The model training data is based on heat overflow events during the cooling recovery phase and heat shortage events during the hot water heating phase, extracted from historical operating data. Heat overflow events are triggered when the recoverable heat exceeds the system's storage capacity, identified by a set threshold. Heat shortage events are recorded when the required heat supply fails to meet the hot water tank temperature requirements. Event data is filtered from long-term operating logs. Establishing the time-series matching relationship between heat overflow and heat shortage events relies on a precise time alignment algorithm. Dynamic time warping is used to align event sequences of different lengths. The matching relationship reveals the heat transfer path from overflow to shortage, and path analysis considers system delays and energy losses. A bidirectional long short-term memory (LSTM) network is used to learn the transfer decay characteristics of the time-series matching relationship. The LSM network structure includes forward and backward recurrent layers to capture temporal dependencies. The transfer decay characteristics model the efficiency decrease of heat during the transfer process. The learning process uses historical data sequences to train the network weights. A time-series predicted value of heat transfer efficiency is generated as the model output. The time-series predicted value represents the efficiency value at future time points in the form of a probability distribution, and the update frequency of the predicted value is synchronized with the heat pump system control cycle.

[0038] The training process of the heat transfer prediction model begins with data preprocessing. Historical running data is cleaned to remove noise and outliers. Timestamps of heat overflow and heat shortage events are accurate to the millisecond level to ensure alignment accuracy. A dynamic time warping method handles the length variation of event sequences. The cost function of the warped path is based on Euclidean distance minimization. The matching results are stored as a list of event pairs for the network to learn from. The training configuration of the bidirectional long short-term memory network includes setting the number of hidden layer nodes and the number of training iterations. The network input is the temporal feature vector of the event pairs, and the output is an initial estimate of the transfer efficiency. Training uses the backpropagation algorithm to optimize the loss function. The generation of time-series prediction values ​​is integrated into the real-time inference engine. The engine loads the trained network model, performs forward computation on the new input event sequence, and outputs an efficiency value. The efficiency value is smoothed to reduce fluctuations.

[0039] Multiplying the recoverable heat by the time-series predicted value of the heat transfer efficiency yields the theoretical adjustment reference value for the cooling medium flow rate. This theoretical adjustment reference value serves as the reference for flow control. The multiplication operation is mathematically represented as follows:

[0040] in: Represents the theoretical adjustment benchmark value. Represents recyclable heat. The time-series predicted value represents the heat transfer efficiency. The power compensation amount of the heat pump compressor is calculated based on the difference between the required heat replenishment and the theoretical adjustment baseline value. The difference is calculated using arithmetic subtraction. The power compensation amount is dynamically adjusted by a proportional-integral controller, and the adjustment amount is converted into a compressor drive signal. The theoretical adjustment baseline value and the power compensation amount are boundary-corrected using the setpoint of the hot water tank temperature parameter as constraints. A minimum-maximum limit function is applied to the boundary correction to prevent the command from exceeding the limit. Adjustment commands are generated and encoded in a digital protocol format, then transmitted via the communication bus.

[0041] The theoretical adjustment benchmark calculation module receives real-time data streams of recoverable heat and time-series predicted values. Recoverable heat is output from the calculation logic, while time-series predicted values ​​are obtained from the heat transfer prediction model. Data stream synchronization is achieved through timestamp matching. Numerical verification is performed before multiplication to ensure that the recoverable heat is non-negative and the time-series predicted value is between zero and one. If verification fails, a default value is used instead. The theoretical adjustment benchmark value is temporarily stored in a register for subsequent processing. The register size is designed to avoid overflow, and the calculation cycle is consistent with the system sampling rate. The power compensation is calculated based on difference analysis. The difference between the supplementary heat and the theoretical adjustment benchmark reflects the energy gap. The gap value is input to the proportional-integral controller, whose parameters are tuned according to the system response characteristics. The output of the proportional-integral controller is mapped to the power range of the heat pump compressor. Whether the mapping relationship is linear or non-linear depends on the compressor characteristics. The power compensation output is an analog or digital signal. The boundary correction stage integrates user-defined hot water tank temperature parameters. These temperature parameters serve as upper limit constraints. The correction algorithm compares the theoretical value with the constraint conditions and dynamically adjusts the command amplitude. The module that generates adjustment instructions converts the numerical instructions into a device-readable format, such as a Modbus or CAN bus message, which includes the instruction type, value, and timestamp.

[0042] The training data for the heat transfer prediction model covers various operating conditions, including different ambient temperatures, load levels, and time periods. Data augmentation techniques increase sequence diversity to improve the model's generalization ability. The hidden layer activation functions of the bidirectional long short-term memory network use tanh or ReLU, and the loss function is chosen as mean squared error. Early stopping during training prevents overfitting, and the model is saved in a lightweight format for deployment on edge devices. The inference process for time-series predictions optimizes the computational graph and is accelerated using TensorFlow Lite or similar frameworks. Post-processing of the predictions includes Kalman filtering for smoothing. The formula for theoretically adjusting the baseline value is limited to cooling medium flow regulation, and the formula parameters are updated in real-time to respond to system changes. Derived from real-time computing, Derived from model reasoning, The output drives the actuator. The power compensation calculation takes into account the dynamic response of the heat pump compressor. The compressor power adjustment has a rate limit to avoid mechanical stress, and the rate limit value is obtained from the manufacturer's specifications. The boundary correction logic is implemented as a state machine. The state machine monitors the hot water tank temperature feedback. When the temperature exceeds the set value, it triggers an adjustment command. The adjustment strategy prioritizes ensuring system safety.

[0043] The training cycle of the heat transfer prediction model is aligned with the system maintenance cycle, and the training data is regularly updated to incorporate the latest operational experience. The generation of adjustment commands is executed cyclically, with the theoretical adjustment baseline value and power compensation amount recalculated in each control cycle to ensure timely commands. The real-time performance of the entire process is guaranteed through multi-task scheduling: the model training task runs in a low-priority background thread, while the command generation task runs in a high-priority real-time thread, with lock-free queues used for inter-thread communication. The detection sensitivity for heat overflow and heat shortage events is configurable; the sensitivity parameter affects the event frequency—too high a frequency increases computational load, while too low a frequency loses detail. Parameter tuning is based on experimental data. The inference latency measurement of the bidirectional long short-term memory network meets the control cycle requirements, and latency optimization is achieved through model quantization and hardware acceleration. A fault-tolerant mechanism is introduced in the calculation of the theoretical adjustment baseline value; when the heat transfer prediction model fails, it switches to a backup formula based on a fixed efficiency value. This backup formula is simple yet reliable. The calculation of the power compensation amount integrates fault detection; power adjustment is prohibited when the compressor malfunctions, and fault signals are obtained from the monitoring system.

[0044] Example 4: Dynamic correction and reverse calibration of model parameters. The cumulative error between the theoretical execution value and the actual sensor feedback value of the cooling medium flow regulation command constitutes the correction trigger mechanism. The theoretical execution value is output from the command generation module, and the actual sensor feedback value comes from the high-precision electromagnetic flowmeter installed on the cooling medium pipeline. The cumulative error is obtained by integrating the deviation value within a continuous time window. The integral operation uses the trapezoidal rule to approximate the error area. The time window length is set to an integer multiple of the system's main thermodynamic time constant to ensure statistical significance. The cumulative error value is compared with a preset threshold in real time. The threshold is dynamically adjusted based on the system tolerance and historical performance data. When the cumulative error exceeds the threshold, the attenuation factor of the heat transfer efficiency is recalculated. The attenuation factor represents the rate of efficiency decrease during the heat transfer from the cooling recovery stage to the hot water heating stage. The recalculation process calls an optimization algorithm to solve for the attenuation factor value that minimizes the prediction error. The attenuation factor is injected into the hidden layer node of the heat transfer prediction model to realize model update. The injection operation is completed by modifying the weight matrix of the hidden layer in the bidirectional long short-term memory network. The adjustment of the weight matrix is ​​based on the product of the attenuation factor and the error gradient. The generation logic of the updated time-series prediction value makes the prediction output closer to the actual system behavior.

[0045] The comparison process relies on high-frequency data acquisition and alignment. The theoretical execution value is emitted from the controller as a digital signal, while the actual sensor feedback value is read in via an analog-to-digital converter, achieving millisecond-level time synchronization accuracy. Cumulative error calculation is performed once per second, with error values ​​cyclically stored in a fixed-length array for trend analysis. Threshold comparison results trigger flag changes. The attenuation factor is recalculated using gradient descent for iterative optimization. The learning rate parameter of gradient descent is adaptively adjusted based on the error change rate, and the iteration process continues until the attenuation factor converges or the maximum number of iterations is reached. Updates to the heat transfer prediction model are performed online without interrupting system operation. Modifications to hidden layer nodes are implemented through a model hot reload mechanism, and prediction accuracy is verified immediately after the logic update is generated.

[0046] Reconstructing the heat transfer efficiency curve on the condenser side based on the actual temperature distribution data of the hot water tank requires multi-dimensional temperature measurement. The actual temperature distribution data is collected from a three-dimensional network of temperature sensors arranged within the hot water tank, with the sensor network covering the tank volume in a grid pattern. The reconstruction of the heat transfer efficiency curve utilizes nonlinear regression technology. The regression model maps the temperature distribution data to an efficiency function of time, and the curve fit is evaluated using the coefficient of determination. Abrupt slope changes in the heat transfer efficiency curve are extracted to identify the critical state of the heat transfer process. These abrupt slope changes are detected by calculating the local extrema of the curve's first derivative, and extrema threshold filtering removes spurious changes caused by noise. Cross-stage coupling is established by correlating evaporator pressure fluctuation data from the cooling recovery stage. This evaporator pressure fluctuation data is obtained from a high-pressure side pressure sensor, and cross-correlation analysis is used to align the pressure data with the slope changes over time. The initial value of the heat exchange coefficient in the dynamic correlation model is adjusted through correlation analysis between the slope changes and the pressure fluctuation data. The Pearson correlation coefficient is calculated to quantify the linear correlation strength between variables, and the adjustment range of the initial value of the heat exchange coefficient is proportional to the correlation coefficient. The data processing workflow for reconstructing the heat transfer efficiency curve includes data cleaning, interpolation, and fitting. Data cleaning removes abnormal readings caused by sensor malfunctions, interpolation fills in missing data points, and the least squares method is selected as the fitting algorithm. The slope abrupt change point detection algorithm integrates multi-scale analysis, searching for abrupt change points at different time scales to improve robustness, and the detection results are stored as a timestamp list. The correlation analysis module accesses the historical database of the cooling recovery stage, extracts segments of pressure fluctuation data that match the time window of the slope abrupt change point, and calculates statistical correlation after standardizing the segment data. The initial value of the heat exchange coefficient of the dynamic correlation model is adjusted by updating the model parameter file. The parameter file version control records each change, and the model is reinitialized after adjustment.

[0047] The system continuously monitors cumulative error and updates the decay factor. It employs a dual-buffer design to store error data: one buffer for real-time calculation and the other for background analysis. When the decay factor is injected into the hidden layer nodes of the heat transfer prediction model, the node activation function bias is adjusted synchronously, with the adjustment amount derived from error feedback. The acquisition cycle of hot water tank temperature distribution data is synchronized with the hot water heating stage. Data upload is triggered at the end of the heating stage, and data compression algorithms reduce storage space usage. The heat transfer efficiency curve reconstruction uses a moving window technique, with the window size adaptively changing according to data density. Curve visualization is used for engineer diagnostics. The correlation between slope abrupt change points and pressure fluctuation data generates feature vectors. These feature vectors are input into a machine learning classifier to identify operating patterns, and the pattern recognition results assist in threshold setting. A closed-loop interaction between the model and the physical system is implemented. Events where the cumulative error exceeds the threshold are logged into a log file for offline analysis. The log file includes timestamps, error values, and system state snapshots. The iterative process of decay factor recalculation monitors convergence conditions, including error change rate being less than the threshold or the upper limit of the iteration count. Divergence triggers a backup algorithm. The heat transfer prediction model is updated and validated by comparing the mean absolute error between the predicted and actual values ​​before and after the update. If the error increases, the model is rolled back to the previous version. The accuracy of the reconstructed heat transfer efficiency curve depends on sensor layout optimization. The layout scheme is determined through computational fluid dynamics simulation, and the simulation model verifies the representativeness of the measurements. The sensitivity parameter for detecting slope abrupt changes is configurable. Sensitivity adjustment balances the detection rate and false alarm rate, and parameter optimization is based on historical data review. The initial value adjustment of the heat exchange coefficient in the dynamic correlation model introduces a confidence interval. The confidence interval is derived from the uncertainty of correlation analysis, and the adjusted value is the most probable value within the interval.

[0048] The entire correction and calibration process is automated, with system state machine management handling the transitions. State machine states include monitoring, calculation, updating, and verification. The cumulative error threshold is dynamically calculated based on the operating environment, including ambient temperature and workload; the threshold formula is based on empirical rules. The implementation of the attenuation factor injection heat transfer prediction model is achieved by calling the model service through an application programming interface (API), which encapsulates the details of neural network operations. The algorithm for reconstructing the actual temperature distribution data of the hot water tank processes multi-sensor data in parallel, with parallel computing accelerated by multi-core processors. A report is generated based on the correlation analysis results of abrupt changes in the slope of the heat transfer efficiency curve. The report content is used for system health monitoring, and the monitoring indicator trends predict maintenance needs. After adjusting the initial value of the heat exchange coefficient of the dynamic correlation model, a consistency check between the model output and sensor readings ensures the effectiveness of the adjustment; failure to do so triggers an alarm. Table 1 shows the correspondence between cumulative error and the recalculated attenuation factor; the data comes from system simulation operation.

[0049] Table 1: Relationship between Cumulative Error and Attenuation Factor

[0050] The table data explains that when the cumulative error exceeds a set threshold (e.g., 65kJ), the attenuation factor is recalculated. A decrease in the attenuation factor value indicates a downward adjustment of the predicted heat transfer efficiency, and a negative adjustment value for the hidden layer nodes indicates a reduction in weight. The table data is generated based on a specific system configuration, and the actual values ​​vary with operating conditions.

[0051] Example 5: Substituting the adjusted initial value of the heat exchange coefficient into the derivation process of the sensible heat recovery potential value initiates a parameter iteration loop. The adjusted initial value of the heat exchange coefficient is a calibration result obtained through correlation analysis between the slope abrupt change point and pressure fluctuation data. Substituting it means rerunning the calculation algorithm for the sensible heat recovery potential value using the new coefficient value. The derivation process of the sensible heat recovery potential value is based on thermodynamic principles and involves the specific heat capacity, flow rate, and temperature measurements of the cooling medium. The adjusted initial value of the heat exchange coefficient acts as a multiplier factor in the calculation process, affecting the final numerical output of the sensible heat recovery potential value. The calculation system acquires the adjusted initial value of the heat exchange coefficient in real time. The initial value of the heat exchange coefficient is passed to the sensible heat recovery potential value derivation module through a shared memory area. The mathematical operation steps within the module are immediately recalculated using the new coefficient, and the intermediate results generated during the derivation process are temporarily stored in a cache.

[0052] Eliminating high-frequency noise components in the initial value of the heat exchange coefficient using an adaptive filtering algorithm is a crucial step in data refinement. The adaptive filtering algorithm dynamically adjusts the filter coefficients based on the minimum mean square error criterion. High-frequency noise components mainly refer to rapid random fluctuations caused by sensor electrical interference or transient fluid turbulence. The filtering process is performed online. The filter receives a continuous stream of initial heat exchange coefficient data, suppresses high-frequency components through convolution operations, and retains low-frequency trend signals. The filter order and cutoff frequency are pre-configured based on the system sampling rate and noise characteristics. The heat exchange coefficient value after eliminating high-frequency noise components exhibits better smoothness and stability. The smoothed data is more suitable for characterizing the steady-state heat exchange characteristics of the system, providing reliable input for subsequent calculations. The computational path for reconstructing recoverable heat based on the filtered heat exchange coefficient is used to optimize the computational logic. Reconstruction means redefining the mathematical mapping relationship from the heat exchange coefficient to recoverable heat. The computational path includes a series of steps such as data preprocessing, parameter transformation, and result synthesis. The filtered heat exchange coefficient serves as the core input parameter in the reconstructed computational path. The algorithms in this path may involve interpolation, integration, or lookup table operations, aiming to more accurately convert heat exchange characteristics into energy quantification values. The refactoring process may adjust the connection order of computing units or introduce new auxiliary variables. These auxiliary variables are obtained from a real-time database. The output of the computing path is an updated value of recyclable heat, which replaces the old version of the calculation results. The core of the coupled iteration is to run the refactored recyclable heat computing path and the heat transfer prediction model (which generates updated time-series prediction values) in parallel. Parallel execution is achieved through a multi-threaded or distributed computing architecture. The two computing modules execute independently but share the same time base and data source. The refactored recyclable heat computing path runs as an independent process, outputting the latest recyclable heat value every second; the heat transfer prediction model (which generates updated time-series prediction values) runs as another process, continuously generating predicted values ​​for heat transfer efficiency. The parallel operating environment ensures that the computing resources of the two modules do not interfere with each other. The modules exchange status information through inter-process communication mechanisms, and the running status is monitored in real time to detect anomalies.

[0053] Intermediate parameters between the two modules are exchanged and a consistency check is performed at preset intervals to ensure system coordination. The preset interval is set according to the system's dynamic response time, for example, an exchange is performed every 5 control cycles. The exchanged intermediate parameters include the sensible heat recovery potential value and the filtered heat exchange coefficient in the recoverable heat calculation path, as well as the hidden layer state and attenuation factor in the heat transfer prediction model. The consistency check process compares the logical compatibility between the output parameters of the two modules. For example, it checks whether the recoverable heat value is within a reasonable range supported by the predicted heat transfer efficiency value and verifies the norm of the parameter differences calculated by the algorithm. When the difference exceeds the tolerance, the consistency check fails, the system records the check result and triggers a recalculation; when the difference is within the tolerance, the consistency check passes. When the consistency check passes, the current parameter set is locked as the final heat pump co-control parameter set to complete the optimization process. The locking operation means marking all current active parameters as read-only and fixing them to non-volatile memory. The final heat pump co-control parameter set is a structured data set containing all adjustable parameters such as heat exchange coefficient, attenuation factor, model weights, and calculation path configuration. Once locked, the parameter set is set as the operating reference for the closed-loop control of the heat pump system. The control system loads these parameters to perform cooling medium flow regulation and compressor power control until the next iteration cycle starts.

[0054] In a given operating scenario, the initial value of the adjusted heat exchange coefficient changes from 0.75 to 0.72. This new value is substituted into the formula for deriving the sensible heat recovery potential. The derivation process uses an instantaneous cooling medium flow rate of 10 liters / minute and a subcooling of 5 Kelvin. After calculation, the sensible heat recovery potential value changes from 8500 kJ to 8160 kJ. The adaptive filtering algorithm uses a 5th-order FIR filter, with filter coefficients adaptively updated based on the real-time noise spectrum. After filtering, the random fluctuation amplitude in the initial value of the heat exchange coefficient is reduced by 60%, stabilizing the value in the range of 0.718 to 0.722. Based on the filtered heat exchange coefficient of 0.72, the recoverable heat calculation path is reconstructed. The new path introduces a bias correction term based on historical data, learned from the past 24 hours of operating records. The reconstructed recoverable heat calculation value is adjusted to 8190 kJ. The reconstructed recoverable heat calculation path runs in parallel with the heat transfer prediction model, which has updated its time-series prediction value generation logic, using a new attenuation factor of 0.79. The two modules operate independently. The recoverable heat calculation path outputs 8190 kJ per second, and the heat transfer prediction model outputs an efficiency prediction value of 0.83 per second. The system is set to an exchange cycle of 30 seconds. Every 30 seconds, the recoverable heat calculation path sends its intermediate parameters to the heat transfer prediction model; the heat transfer prediction model sends its intermediate parameters to the recoverable heat calculation path. A consistency check algorithm calculates whether the product of the recoverable heat value and the efficiency prediction value matches the theoretical heat transfer amount, with a tolerance of ±10 kJ. The current difference of 2.3 kJ is within the tolerance range, and the consistency check passes. The system locks the current parameter set, including the heat exchange coefficient of 0.72, the attenuation factor of 0.79, the configuration version v2.1 of the recoverable heat calculation path, and the weight matrix of the heat transfer prediction model. The locked final heat pump co-control parameter set is written to the controller's flash memory, and the heat pump system performs subsequent control based on these parameters. Closed-loop control uses the recoverable heat value of 8190 kJ and the efficiency of 0.83 to calculate the cooling medium flow setpoint, and the compressor power is adjusted according to the difference in the required supplementary heat. The system continues to run, with the parameter set remaining unchanged until the next trigger condition is met, at which point a new iteration cycle begins.

[0055] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0056] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for heat pump-assisted cooling recovery and intelligent hot water temperature control, characterized in that, Includes the following steps: The flow rate parameters of the cooling medium and the temperature parameters of the hot water tank are detected during the operation of the heat pump system, and a dynamic correlation model between the flow rate of the cooling medium and the temperature of the hot water tank is established. Based on the operating mode of the heat pump system, the cooling recovery stage and the hot water heating stage are divided, and the heat exchange characteristic parameters of the two stages are extracted respectively. Based on the dynamic correlation model and heat exchange characteristic parameters, the recoverable heat in the cooling recovery stage and the additional heat required in the hot water heating stage are calculated. A heat transfer prediction model is trained using historical operating data of the heat pump system to predict the heat transfer efficiency from the cooling recovery stage to the hot water heating stage. Based on the recoverable heat, the heat to be replenished, and the heat transfer efficiency, cooling medium flow rate regulation commands and heat pump compressor power regulation commands are generated. Real-time monitoring of the execution deviation between the cooling medium flow regulation command and the heat pump compressor power regulation command, and dynamic correction of the output parameters of the heat transfer prediction model; At the end of the hot water heating stage, the actual temperature distribution data of the hot water tank is collected, the heat exchange coefficient of the dynamic correlation model is calibrated in reverse, and the calculation logic of recoverable heat in the cooling recovery stage is updated based on the calibrated heat exchange coefficient. The updated recyclable heat calculation logic is coupled and iterated with the heat transfer prediction model to output the final heat pump collaborative control parameter set. Based on the final heat pump collaborative control parameter set, the cooling recovery and hot water heating of the heat pump system are controlled in a closed loop.

2. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 1, characterized in that, The establishment of a dynamic correlation model between the cooling medium flow rate and the hot water tank temperature includes: Collect the instantaneous values ​​of the cooling medium flow rate and the corresponding temperature gradient changes of the hot water tank under different load rates of the heat pump system; Identify the nonlinear mapping relationship between the instantaneous value of the cooling medium flow rate and the change in the temperature gradient of the hot water tank; The terminal difference parameter between the heat pump evaporator and condenser is introduced as a compensation variable in the dynamic correlation model; By statistically analyzing the weight distribution of compensation variables through a sliding time window, a dynamic correlation model with weight correction is constructed.

3. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 2, characterized in that, The extraction of heat exchange characteristic parameters in the two stages includes: During the cooling recovery phase, the subcooling of the cooling medium on the evaporator side and the rate of decrease in saturation temperature are captured. During the hot water heating stage, the temperature rise lag time and stratification temperature difference of the hot water tank on the condenser side were recorded. The supercooling, saturation temperature decrease rate, temperature rise lag time, and stratified temperature difference are normalized into a heat exchange characteristic matrix.

4. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 3, characterized in that, The calculation of recoverable heat in the cooling recovery stage and the additional heat required in the hot water heating stage includes: Based on the subcooling and saturation temperature drop rate in the heat exchange characteristic matrix, the sensible heat recovery potential value of the evaporator is derived. By combining the temperature difference between the layers in the hot water tank and the temperature rise lag time, the latent heat demand gap of the condenser can be calculated. The sensible heat recovery potential value is converted into recoverable heat value through a dynamic correlation model, and the latent heat demand gap value is converted into heat that needs to be supplemented.

5. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 4, characterized in that, The training heat transfer prediction model includes: Extract heat overflow events during the cooling recovery phase and heat shortage events during the hot water heating phase from historical operation data; Establish time-series matching relationships between heat overflow events and heat shortage events; By employing a bidirectional long short-term memory network to learn the transfer decay characteristics of time series matching relationships, a time-series predicted value of heat transfer efficiency is generated.

6. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 5, characterized in that, The generation of cooling medium flow rate regulation commands and heat pump compressor power regulation commands includes: Multiply the recoverable heat by the time-series predicted value of the heat transfer efficiency to obtain the theoretical adjustment reference value of the cooling medium flow rate; Calculate the power compensation amount of the heat pump compressor based on the difference between the required heat supply and the theoretical adjustment benchmark value; Using the set value of the hot water tank temperature parameter as a constraint, boundary corrections are made to the theoretical adjustment reference value and the power compensation amount to generate adjustment commands.

7. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 6, characterized in that, The output parameters of the dynamically corrected heat transfer prediction model include: Compare the cumulative error between the theoretical execution value of the cooling medium flow rate regulation command and the actual sensor feedback value; When the cumulative error exceeds the threshold, the attenuation factor of the heat transfer efficiency is recalculated; The attenuation factor is injected into the hidden layer nodes of the heat transfer prediction model to update the generation logic of the time series prediction values.

8. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 7, characterized in that, The heat exchange coefficient of the reverse calibration dynamic correlation model includes: Based on the actual temperature distribution data of the hot water tank, the heat transfer efficiency curve on the condenser side is reconstructed. Extract the abrupt change points in the slope of the heat transfer efficiency curve and correlate them with evaporator pressure fluctuation data during the cooling recovery stage; By analyzing the correlation between the slope abrupt change point and pressure fluctuation data, the initial value of the heat exchange coefficient of the dynamic correlation model was adjusted.

9. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 8, characterized in that, The calculation logic for recoverable heat in the updated cooling recovery stage includes: The process of substituting the adjusted initial value of the heat exchange coefficient into the derivation of the sensible heat recovery potential value; An adaptive filtering algorithm is used to eliminate high-frequency noise components in the initial value of the heat exchange coefficient. The calculation path for reconstructing recoverable heat is based on the filtered heat exchange coefficient.

10. The method for heat pump synergistic cooling recovery and intelligent hot water temperature control according to claim 9, characterized in that, The coupling iteration process includes: The reconstructed recoverable heat calculation path and the heat transfer prediction model with the updated time-series prediction value generation logic are run in parallel. The intermediate parameters of the two are exchanged and a consistency check is performed at preset intervals; When the consistency check passes, the current parameter set is locked as the final heat pump collaborative control parameter set.

Images