A high-efficiency heat exchange air source heat pump system and its control method

Abstract

This invention discloses a high-efficiency heat exchange air source heat pump system and its control method, belonging to the field of air source heat pump heat exchange control technology. The system includes a heat pump hardware body and an embedded intelligent control software system. The software system constructs a perception-decision-execution-iteration closed-loop architecture, integrating load forecasting, intelligent heat exchange regulation, phase change collaborative control, fault diagnosis, and data interaction modules. Through multi-algorithm fusion, it achieves accurate heat exchange load forecasting, dynamic parameter adjustment, and fault protection under all operating conditions, solving problems such as control lag and poor coordination in traditional systems. This invention significantly improves load forecasting accuracy and reduces regulation lag through innovative software system architecture and multi-module collaboration, achieving precise coordination between phase change components and the main unit, accurate and efficient fault diagnosis, supporting continuous system optimization, significantly improving heat exchange efficiency and operational stability under all operating conditions, reducing energy consumption and maintenance costs, and possessing broad application prospects.

Description

Technical Field

[0001] This invention relates to the field of air source heat pump heat exchange control technology, and in particular to a high-efficiency heat exchange air source heat pump system and its control method. Background Technology

[0002] Air source heat pumps, as highly efficient and energy-saving heating and energy supply equipment, are widely used in residential heating, commercial hot water supply, and industrial waste heat recovery due to their advantages of being clean, environmentally friendly, and having low energy consumption. They have become one of the key technologies for achieving the "dual carbon" goal. However, with users' increasing demands for heating stability, energy efficiency, and intelligent features, the technical shortcomings of traditional air source heat pump systems are becoming increasingly apparent. In particular, deficiencies in software control have become a core bottleneck restricting their performance upgrades.

[0003] Existing air source heat pump systems largely rely on optimizing traditional hardware structures to improve heat exchange efficiency, with software control serving only as an auxiliary function, lacking systematic architectural design and algorithmic innovation. In terms of load forecasting, traditional systems often use a single ambient temperature parameter for simple prediction, failing to consider multi-dimensional influencing factors such as humidity, user water usage habits, and historical energy consumption. This results in prediction errors generally exceeding 10%, failing to provide accurate basis for parameter adjustment, leading to indiscriminate compressor start-stop and delayed adjustment of throttling components, causing fluctuations in heat exchange efficiency and energy waste. In the parameter control stage, most systems use fixed-parameter PID control algorithms. Faced with complex operating conditions such as sudden changes in ambient temperature and fluctuations in heat exchange load, they struggle to quickly and dynamically correct control parameters, resulting in large deviations in heat exchanger inlet and outlet temperatures, low refrigerant flow matching, and limited improvement in the system's COP (coefficient of performance).

[0004] While phase change assisted heat exchange technology has been applied to air source heat pumps, existing systems often rely on a single ambient temperature threshold to trigger start-stop of the phase change components. This lack of precise coordination logic with the heat pump's main operating status frequently leads to mismatches between the heat charge / discharge sequence and the heat exchange load. This not only fails to fully utilize the energy compensation function of the phase change material but also increases ineffective energy consumption. In terms of fault diagnosis, traditional systems only monitor single parameter exceedances without extracting deeper features such as parameter fluctuation trends and cross-correlation, resulting in high false alarm rates, inaccurate fault type identification, and an inability to promptly warn of hidden faults such as pipe blockage and phase change material failure, thus affecting system stability. Furthermore, existing systems have weak data interaction capabilities and lack deep integration with cloud platforms, hindering online upgrades of control algorithms and dynamic calibration of control parameters. This makes continuous performance optimization difficult and makes it hard to adapt to diverse application scenarios.

[0005] These software-level technical deficiencies result in traditional air source heat pump systems exhibiting low heat exchange efficiency, high energy consumption, and poor stability under complex operating conditions, failing to meet the demands of modern heating and energy supply for intelligence, efficiency, and reliability. Therefore, there is an urgent need for a high-efficiency heat exchange air source heat pump system based on software system innovation, which can solve many of the pain points of existing technologies through architecture optimization and algorithm upgrades. Summary of the Invention

[0006] This invention proposes a high-efficiency heat exchange air source heat pump system and its control method to solve the problems mentioned in the prior art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a high-efficiency heat exchange air source heat pump system, comprising the following modules: The core innovation of heat pumps focuses on embedded intelligent control software systems, with the hardware itself serving only as an execution carrier; this software system achieves multi-module collaboration, precise prediction, and dynamic regulation. The software system adopts a closed-loop architecture of "perception, decision-making, execution, and iteration", which is divided into a perception layer, a decision-making layer, an execution layer, and an interaction layer. Each layer completes seamless data flow through an industrial-grade bus protocol. The decision-making layer integrates five core functional modules, sharing information and working collaboratively based on a unified data bus: the load forecasting module adopts an improved LSTM neural network algorithm, inputs 6 types of feature parameters to build a multi-dimensional model, and outputs the heat exchange load forecast value 1-3 hours in advance; The intelligent heat exchange regulation module adopts an adaptive PID-fuzzy control fusion algorithm, collects four types of real-time parameters to dynamically correct the PID coefficient, adjusts the compressor frequency and the opening of the electronic throttling component, and establishes a dynamic matching mechanism to maintain the optimal heat exchange efficiency. The phase change collaborative control module is equipped with a scenario-based decision-making algorithm, which comprehensively defines the start-stop threshold and timing of the phase change component by integrating three types of parameters, and works in collaboration with the host. The fault diagnosis module integrates multi-feature fusion and fuzzy inference algorithms to build a fault feature library of 12 types, automatically generate graded handling suggestions and trigger alarms; the data interaction and iterative optimization module is compatible with multiple protocols such as MQTT, Modbus, and OPCUA, completes bidirectional data transmission, and iteratively optimizes algorithms and parameters based on big data from the cloud platform.

[0008] Furthermore, the improved LSTM neural network algorithm of the load prediction module reduces the risk of gradient vanishing by introducing layer normalization technology. The training process adopts an adaptive learning rate optimizer, which automatically halves the loss value when the loss value decreases by less than 0.5% for five consecutive iterations. The training sample size is no less than 10,000 groups, and the prediction duration supports three levels: 1 hour, 2 hours, and 3 hours.

[0009] Furthermore, the adaptive PID-fuzzy control fusion algorithm of the intelligent heat exchange adjustment module contains a 5×5×5 three-dimensional fuzzy matrix in the fuzzy rule base. The input quantities are load deviation and deviation change rate, and the output quantities are PID parameter correction quantities. The load deviation is divided into 5 fuzzy subsets: negative large, negative small, zero, positive small, and positive large. The deviation change rate is also divided into 5 fuzzy subsets. The smooth correction of parameters is completed through fuzzy inference.

[0010] Furthermore, the scenario-based decision-making algorithm of the phase change collaborative control module presets three core operating scenarios, each with an independent phase change component start / stop threshold and collaborative logic: In the low temperature and high load scenario, the phase change component and the host operate synchronously, with a charge / discharge switching cycle of 15 minutes; In the normal temperature and medium load scenario, the phase change component is only started at the initial stage of host startup and automatically shuts down after 30 minutes; In the high temperature and low load scenario, the phase change component maintains a heat storage state and only starts to release heat when the heat exchange efficiency suddenly drops.

[0011] Furthermore, the multi-feature fusion algorithm of the fault diagnosis module extracts deep features including parameter over-threshold duration, parameter fluctuation standard deviation, and correlation coefficient between different parameters. The parameter over-threshold duration threshold is set to 30s, and the fluctuation standard deviation threshold is set according to the parameter type.

[0012] Furthermore, the data interaction and iterative optimization module supports both local and cloud storage modes. Local storage can store nearly 90 days of operating data, while the cloud platform supports remote upgrades of algorithm models and batch calibration of control parameters. The upgrade process does not affect the normal operation of the system, and differential upgrade technology is used to reduce the amount of data transmission.

[0013] Furthermore, the intelligent control software system based on claim 1 includes the following steps: Step 1: The perception layer collects seven types of real-time data through sensors, including ambient temperature, humidity, system high and low pressure, heat exchanger inlet and outlet temperatures, refrigerant flow rate, water tank temperature, and user water usage status. The data is then filtered and uploaded to the decision layer. Step 2: The load forecasting module calls the improved LSTM neural network algorithm, inputs historical data and real-time data, and outputs the heat exchange load forecast value 1-3 hours later; Step 3: The intelligent heat exchange regulation module generates control commands for compressor frequency and throttling component opening based on the load forecast value and real-time load deviation through an adaptive PID-fuzzy fusion algorithm. Step 4: The phase change collaborative control module controls the start-up, shutdown, and heat charging / discharging status of the phase change components based on the load forecast, real-time heat exchange efficiency, and ambient temperature through a scenario-based decision-making algorithm. Step 5: The fault diagnosis module monitors for data anomalies in real time and triggers the fault protection mechanism. Step six: The data interaction module uploads the running data, receives the algorithm upgrade instructions from the cloud platform, and completes the iterative optimization of system performance.

[0014] Furthermore, the parameter control logic in step three is as follows: when the load forecast value is higher than 120% of the real-time load, the compressor frequency is increased to 80% of the rated value 5 minutes in advance, and the opening of the throttling component is pre-adjusted to the appropriate position; when the load forecast value is lower than 80% of the real-time load, the compressor frequency is gradually reduced, and the heat storage time of the phase change component is extended; when the deviation between the load forecast value and the real-time load is within ±10%, the parameters are kept stable.

[0015] Furthermore, the phase change coordination logic in step four is as follows: In low-temperature, high-load scenarios, the phase change component first stores heat for 10 minutes, and then releases heat synchronously with the host; in normal-temperature, medium-load scenarios, the phase change component only releases heat at the initial stage of host startup, and switches to heat storage after the water temperature in the storage tank rises to 80% of the target value; in high-temperature, low-load scenarios, the phase change component maintains heat storage state, and automatically starts releasing heat when the heat exchange efficiency is lower than 80% of the optimal value, and the duration of heat release is dynamically adjusted according to the load deviation.

[0016] Furthermore, the fault protection logic in step five is as follows: when an abnormal parameter is detected, first-level protection is executed; if the abnormality is not eliminated, second-level protection is executed; if it is still not eliminated, third-level protection is executed; after the fault is cleared, the system executes a restart preheating program to gradually restore the parameters to the normal range.

[0017] Compared with existing technologies, the beneficial effects of this invention are: This invention constructs a closed-loop intelligent control software system of "perception-decision-execution-iteration", with algorithm innovation and module collaboration as the core, and realizes intelligent, precise and efficient heat exchange control of air source heat pump, which has significant advantages over existing technologies.

[0018] Regarding load forecasting accuracy, this invention employs an improved LSTM neural network algorithm, integrating six types of multi-dimensional feature parameters. The model is optimized through an attention mechanism and layer normalization technology, and the prediction deviation is minimized using the RMSE loss function. This keeps the heat exchange load forecasting error within ±3%, far exceeding the accuracy of traditional single-parameter forecasting. This provides accurate predictions for subsequent parameter adjustments, preventing energy waste caused by blind operation from the outset. In terms of parameter control response speed and adaptability, an innovative adaptive PID-fuzzy control fusion algorithm is adopted. PID parameters are dynamically corrected using fuzzy rules, constructing a dynamic matching mechanism of "predicted load - real-time parameters - control commands." This ensures that compressor frequency and throttling component opening adjustments are shock-free, and heat exchange efficiency fluctuations do not exceed ±2%. Even under complex operating conditions such as sudden changes in ambient temperature and load fluctuations, the optimal heat exchange state can still be maintained. The system COP value is 15%-25% higher than that of traditional fixed parameter control.

[0019] In terms of phase change coordinated control, by employing scenario-based decision-making algorithms and start-stop criterion formulas, and comprehensively considering load forecasts, real-time heat exchange efficiency, and ambient temperature change rates, weight coefficients are dynamically allocated to achieve precise coordination between the phase change component's heat charging and discharging sequence and the heat pump unit's operating status. This avoids energy waste, with particularly significant improvements in heat exchange efficiency at low temperatures. Regarding fault diagnosis and protection, multi-feature fusion and fuzzy inference algorithms are integrated to extract deep features of parameters and quantify them using feature fusion formulas. The fault diagnosis accuracy is no less than 95%, accurately identifying 12 common faults and generating tiered handling suggestions, effectively reducing fault misjudgment rates and downtime risks, and improving system operational stability.

[0020] Furthermore, this invention supports multi-protocol data interaction and local-cloud dual-mode storage. It achieves online upgrades of algorithm models and dynamic calibration of control parameters through big data analysis on a cloud platform. Differential upgrade technology reduces data transmission volume, ensuring continuous optimization of system performance and adaptability to diverse application scenarios. Overall, through innovative software architecture and algorithm optimization, this invention completely solves the problems of inaccurate load prediction, lagging parameter adjustment, poor phase change coordination, and inaccurate fault diagnosis in traditional systems. It significantly improves the heat exchange efficiency, operational stability, and intelligence level of air source heat pumps, and significantly reduces energy consumption and maintenance costs. It provides reliable technical support for energy conservation and carbon reduction in the heating and energy supply sector, and has broad application prospects and promotional value. Attached Figure Description

[0021] Figure 1 This is a schematic block diagram of a high-efficiency heat exchange air source heat pump system proposed in this invention; Figure 2 This is a schematic block diagram of a control method for a high-efficiency heat exchange air source heat pump system proposed in this invention. Figure 3 A comparison chart of load prediction errors under different ambient temperatures; Figure 4 This is a graph comparing the system's COP value with ambient temperature. Figure 5 A comparison chart of phase change energy utilization under different operating conditions; Figure 6 A comparison chart of diagnostic accuracy for different fault types. Detailed Implementation

[0022] 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.

[0023] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0024] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0025] Reference Figures 1 to 6 A high-efficiency heat exchange air source heat pump system includes the following modules: The heat pump hardware body and embedded intelligent control software system are described. The heat pump hardware body serves only as the execution carrier. The core innovation is focused on the architecture design and algorithm optimization of the intelligent control software system. Through multi-module collaboration, accurate prediction and dynamic regulation at the software level, the system solves the problems of traditional air source heat pumps, such as heat exchange efficiency being greatly affected by environmental fluctuations, parameter adjustment lag, and poor coordination of phase change components, and achieves efficient and stable operation under all operating conditions.

[0026] The intelligent control software system adopts a closed-loop architecture of "perception-decision-execution-iteration", which is divided into a perception layer, a decision layer, an execution layer and an interaction layer. Each layer achieves seamless data flow through an industrial-grade bus protocol. The perception layer is responsible for collecting multi-dimensional operating data, the decision layer generates control commands based on innovative algorithms, the execution layer drives hardware to perform actions, and the interaction layer realizes data uploading and command reception. The four-layer architecture works together to ensure the speed and accuracy of control response.

[0027] The decision-making layer integrates five core functional modules, each based on a unified data bus to achieve information sharing and collaborative work: The load forecasting module adopts an improved LSTM neural network algorithm, which differs from traditional single-parameter forecasting. It inputs six types of feature parameters: historical ambient temperature (T_env_hist), humidity (H_hist), water tank temperature change curve (ΔT_water_hist), user water usage time distribution (T_use), and historical energy consumption data of the heat exchange system (E_hist). Through an attention mechanism, it strengthens the weighting of key influencing factors, constructing a multi-dimensional load forecasting model. The root mean square error (RMSE) is used as the model training loss function, as shown in the formula: ,in To predict the sample size, This represents the actual heat exchange load value. This formula is used to predict load values ​​for the model and minimize prediction deviation, outputting accurate heat exchange load prediction values ​​1-3 hours in advance with prediction errors controlled within ±3%. This provides a basis for subsequent parameter adjustments and avoids inefficiency caused by blind operation.

[0028] The intelligent heat exchange regulation module innovatively adopts an adaptive PID-fuzzy control fusion algorithm, overcoming the lag defect of traditional fixed-parameter PID regulation. It collects four real-time parameters: heat exchanger inlet and outlet temperatures (T_in, T_out), system high and low pressures (P_high, P_low), refrigerant flow rate (Q_ref), and water temperature gradient in the storage tank (ΔT_water). The module dynamically corrects the PID proportional coefficient using fuzzy rules. ), integral coefficient ( ) and differential coefficients ( The corrected formula is: ,in , , These are the initial parameters for the PID controller. , , This is the parameter correction amount for the fuzzy inference output. Based on the load forecast value ( ) and real-time load ( deviation The compressor operating frequency (30-60Hz continuously adjustable) and the opening of the electronic throttling component (adjustment accuracy up to ±0.5%) are dynamically adjusted to establish a dynamic matching mechanism of "predicted load - real-time parameters - control commands" to ensure that the heat exchange efficiency is always maintained in the optimal range.

[0029] The phase change coordinated control module is equipped with a scenario-based decision-making algorithm. Through software logic, it accurately defines the start-stop thresholds and heat charging / discharging sequence of the phase change auxiliary components. It is not simply triggered based on ambient temperature, but rather comprehensively considers the load forecast value. ), real-time heat exchange efficiency ( ), Ambient temperature change rate (Δ The start / stop criterion formula for three types of parameters ( / Δt) is as follows: ,in , , Weighting coefficients ( + + =1), For the system rated load, For optimal heat exchange efficiency. When ≥0.8 and ambient temperature When the temperature is <5℃, start the phase change module heat storage 3-5 minutes in advance; when ≤0.5 or <0.8× When the phase change heat release compensation is activated, it forms a precise coordination with the operating status of the heat pump host, avoiding energy waste caused by the asynchronous operation of the phase change component and the host.

[0030] The fault diagnosis module integrates multi-feature fusion and fuzzy inference algorithms, which not only monitors single parameter anomalies but also extracts parameter fluctuation amplitudes. ), changing trend ( Cross-correlation () Based on deep features such as pipeline blockage, sensor malfunction, compressor overload, and phase change material failure, a fault feature library is constructed, covering 12 common faults including pipeline blockage, sensor failure, compressor overload, and phase change material failure. A feature fusion formula is used: ,in , , For feature weights ( + + =1), achieving accurate fault type identification through a fuzzy inference model with a diagnostic accuracy of no less than 95%, while automatically generating graded handling suggestions and supporting local audible and visual alarms and remote notification linkage; the data interaction and iterative optimization module supports multiple protocols such as MQTT, Modbus, and OPCUA, realizing bidirectional data transmission with the cloud platform and user terminals, regularly uploading system operation data and heat exchange efficiency curves, and realizing online upgrades of algorithm models and dynamic calibration of control parameters based on big data analysis on the cloud platform, continuously optimizing system operation performance.

[0031] In this invention, the improved LSTM neural network algorithm of the load prediction module reduces the risk of gradient vanishing by introducing layer normalization technology. The layer normalization formula is as follows: Where x is the output of the LSTM hidden layer, This is the batch average. For batch variance, To prevent tiny values ​​where the denominator is zero (take 1e-6). , These are learnable parameters. The training process uses an adaptive learning rate optimizer (Adam), with an initial learning rate value of... Set to 0.001, when the loss value decreases by less than 0.5% for five consecutive iterations, according to the formula... Automatic halving, training sample size no less than 10,000 groups, prediction duration supports three options: 1 hour, 2 hours, and 3 hours. Users can configure it themselves through the interactive layer to meet the prediction needs of different scenarios.

[0032] In this invention, the adaptive PID-fuzzy control fusion algorithm of the intelligent heat exchange regulation module has a fuzzy rule base containing a 5×5×5 three-dimensional fuzzy matrix. The input quantities are load deviation (E) and deviation change rate (EC=ΔE / Δt), and the output quantity is the PID parameter correction quantity (…). , , The load deviation is divided into five fuzzy subsets: negative large (NB), negative small (NS), zero (ZO), positive small (PS), and positive large (PB). The deviation change rate is also divided into five fuzzy subsets. The fuzzy membership function adopts a triangular distribution and is determined using fuzzy inference formulas. ( For fuzzy rule weights, (For the output of the i-th rule) to achieve smooth parameter correction, ensuring no impact on compressor frequency and throttling component opening adjustment, and heat exchange efficiency fluctuation not exceeding ±2%. The default value for the initial PID parameters is... =0.8、 =0.1、 =0.3, which allows users to customize and modify it through the interaction layer.

[0033] In this invention, the scenario-based decision-making algorithm of the phase change collaborative control module presets three core operating scenarios (low temperature high load, normal temperature medium load, and high temperature low load), and the weight coefficients are dynamically allocated according to the scenario: low temperature high load scenario ( )Down, =0.4、 =0.3、 =0.3, the phase change component operates synchronously with the main unit, and the charge / discharge switching cycle is calculated according to the formula. Calculation (unit: minutes); Medium load scenario at normal temperature ( )Down, =0.2、 =0.5、 =0.3, the phase change component is activated only during the initial startup of the host, and the duration is long-pressed. Calculation (unit: minutes); High temperature and low load scenario ( )Down, =0.1、 =0.3、 =0.6, the phase change component remains in a heat storage state, only when <0.7× Heat release is initiated at that time.

[0034] In this invention, the deep feature parameters extracted by the multi-feature fusion algorithm of the fault diagnosis module are defined as: the duration of parameter exceeding the threshold. (Unit: s), Standard deviation of parameter fluctuation (Stress-related) Temperature ), Pearson correlation coefficient r between different parameters. Feature weights are assigned according to parameter type: =0.4、 =0.3、 =0.3, the fault judgment threshold is dynamically adjusted according to the formula Fth=0.6+σmaxmax(σ)×0.3 (σ_max is the maximum allowable fluctuation value of the parameter). When F≥F_th and two or more types of surface parameter anomalies are triggered, it is judged as a fault state. For phase change material failure faults, an additional phase change latent heat decay coefficient is introduced. ( For the current latent heat, (for initial latent heat), when A special alarm is triggered when the value is less than 0.8.

[0035] In this invention, the data interaction and iterative optimization module supports both local and cloud storage modes. Local storage can hold nearly 90 days of operational data, while the cloud platform uses the least squares method to calibrate control parameters. The calibration formula is as follows: ,in These are the parameters after calibration. These are the parameters before calibration. To calibrate the sample size, This refers to parameter adjustment amounts. The upgrade process employs differential upgrade technology, and the upgrade package size is calculated according to the formula. compression( This is the full package size. The percentage of redundant data is set at 0.6-0.8, the upgrade time is no more than 5 minutes, and it does not affect the normal operation of the system.

[0036] In this invention, the intelligent control software system based on claim 1 includes the following steps: Step 1, the sensing layer collects ambient temperature data through sensors (…). The system collects seven types of real-time data: humidity (H), system high and low pressure (P_high, P_low), heat exchanger inlet and outlet temperatures (T_in, T_out), refrigerant flow rate (Q_ref), water tank temperature (T_water), and user water usage status (S_use). The sampling frequency is 1Hz, and the data is filtered using a moving average formula. ( (The filter window size is set to 5) After processing the data, it is uploaded to the decision layer; Step 2, the load forecasting module calls the improved LSTM neural network algorithm, inputs historical data and filtered real-time data, optimizes the model through the RMSE loss function, and outputs the heat exchange load forecast value 1-3 hours later. Step 3: The intelligent heat exchange regulation module calculates the load deviation. ( For real-time load (calculated from Q_ref and ΔT_water), ΔKp, ΔKi, and ΔKd are output through fuzzy inference, and Kp, Ki, and Kd are updated by substituting them into the PID parameter correction formula to generate control commands for compressor frequency and throttling component opening; Step four, the phase change collaborative control module substitutes the scenario-based weighting coefficients for calculation. Value, based on and Step 5: Determine whether to start the phase change component and the charge / discharge mode; Step 6: The fault diagnosis module calculates the F value and F_th, and triggers graded fault protection when F≥F_th; Step 7: The data interaction module uploads the running data and core calculated values ​​such as F and S, receives the parameter calibration instructions from the cloud platform, substitutes them into the calibration formula to update the control parameters, and realizes iterative optimization of system performance.

[0037] In this invention, the parameter adjustment logic in step three is as follows: when >1.2 At that time, press the formula 5 minutes in advance. Increase the initial frequency of the compressor ( (for the rated frequency), the opening degree of the throttling component is as follows: Pre-adjustment ( (for optimal opening); when <0.8 At that time, according to the formula Gradually reduce the compressor frequency while extending the heat storage time of the phase change component; when At the same time, the parameters are kept stable, and only micro-calls are made through the fuzzy PID algorithm to ensure a balance between heat exchange efficiency and energy consumption.

[0038] In this invention, the phase change synergy logic in step four is refined as follows: Under low temperature and high load scenarios, the phase change component first stores heat for 10 minutes, and then switches between charging and discharging states according to the T_cycle formula; under normal temperature and medium load scenarios, the phase change component performs real-time calculations after startup. ,when ≥0.9× And continue for 10 minutes, press Extending operating time; under high temperature and low load scenarios, the phase change component's heat storage process is achieved through formulas. ( For the heat storage power of phase change components, To monitor the heat storage capacity (for the duration of heat storage), when... ≥0.8 Heat storage stops only when Heat release is activated when the temperature drops by more than 20%.

[0039] In this invention, the fault protection logic in step five is as follows: For Level 1 protection (F_th ≤ F < 0.9), the PID correction formula is substituted to increase the weight of ΔKi (Ki correction coefficient increased by 50%), and monitoring continues for 30 seconds; for Level 2 protection (0.9 ≤ F < 1.0), the following steps are performed: Reduce compressor frequency and shut down phase change components; in the third-level protection scenario (F≥1.0), control the compressor to stop, close the throttling component, and activate audible and visual alarms and remote notification; after troubleshooting, the system executes a restart preheating procedure, according to... Gradually restore the frequency (t_warm is the warm-up time in minutes) to avoid impact damage to the hardware.

[0040] The following two examples further illustrate the specific implementation of this system: Example 1: Application of residential heating in northern winters This embodiment is applied to a centralized heating scenario for a high-rise residential building in a northern city, with a heating area of ​​1000 square meters. The ambient temperature range during the heating season is -25℃ to 10℃, with large diurnal temperature differences and significant fluctuations in heat exchange load. Problems such as low heat exchange efficiency, lag in parameter adjustment, and poor coordination of phase change components under low-temperature conditions need to be addressed. This system achieves efficient and stable heating under all operating conditions through the collaborative operation of multiple modules of an embedded intelligent control software system, fully covering all technical features of the claims. The specific implementation process is as follows.

[0041] During system deployment, the heat pump hardware includes a compressor, a finned tube-microchannel composite heat exchanger, an electronic throttling device, a 500L insulated water tank, and phase change auxiliary components. The intelligent control software system is pre-installed in an embedded controller and communicates with the sensors and actuators of the hardware via an industrial bus. The sensing layer deploys seven types of sensors: an ambient temperature sensor installed on the air inlet side of the heat exchanger; a humidity sensor positioned adjacent to the ambient temperature sensor; high and low pressure sensors installed on the compressor inlet and outlet pipes respectively; inlet and outlet temperature sensors embedded in the heat exchanger pipes; a refrigerant flow sensor connected in series on the outlet side of the throttling device; three water tank temperature sensors evenly distributed along the height of the tank; and a user water usage status sensor installed on the water tank outlet pipe. All sensors are sampled at a frequency of 1Hz, and the data is processed by moving average filtering before being uploaded to the decision layer.

[0042] When the load forecasting module runs, it receives six types of characteristic parameters: historical ambient temperature and humidity data for the past 30 days, water temperature variation curves in the storage tank, water usage distribution of residential users over the past month (concentrated between 7-8 AM and 6-8 PM), and historical energy consumption data of the heat exchange system, totaling 12,000 training samples. The model is trained using an improved LSTM neural network algorithm, incorporating layer normalization to reduce the risk of gradient vanishing. An Adam adaptive learning rate optimizer is employed, with an initial learning rate of 0.001, which is automatically halved when the loss rate decreases below 0.5% after five consecutive iterations. The model outputs predicted heat exchange load values ​​two hours in advance, and the prediction error is controlled within ±3% under different ambient temperatures during the heating season, providing a precise basis for subsequent parameter adjustments.

[0043] The intelligent heat exchange regulation module initiates an adaptive PID-fuzzy control fusion algorithm, collecting four types of parameters in real time: heat exchanger inlet and outlet temperatures, system high and low pressures, refrigerant flow rate, and water temperature gradient in the storage tank. The fuzzy rule base is a 5×5×5 three-dimensional fuzzy matrix, with load deviation and deviation change rate divided into five fuzzy subsets. The PID parameters are dynamically corrected through fuzzy inference. When the ambient temperature is -15℃ and the predicted load is 90% of the rated load, the initial PID parameters are Kp=0.8, Ki=0.1, and Kd=0.3. After fuzzy inference, the output parameters are ΔKp=0.2, ΔKi=0.05, and ΔKd=0.1. After correction, Kp=1.0, Ki=0.15, and Kd=0.4. The compressor operating frequency is adjusted to 55Hz, and the electronic throttling device opening is adjusted to 80%, ensuring optimal heat exchanger efficiency.

[0044] The phase change collaborative control module assigns weighting coefficients k1=0.4, k2=0.3, and k3=0.3 according to low-temperature, high-load scenarios. When the ambient temperature is below -10℃ and the predicted load is higher than 80% of the rated load, the phase change component heat storage is activated 4 minutes in advance. When the heat exchange efficiency is lower than 80% of the optimal value, phase change heat release compensation is activated. The charge / discharge switching cycle is dynamically calculated based on the predicted load value. When the predicted load is 90% of the rated load, the switching cycle is 18 minutes, ensuring precise matching between phase change energy and heat exchange load and avoiding energy waste.

[0045] The fault diagnosis module extracts deep features of parameters in real time, including the duration of parameter exceeding the threshold, the standard deviation of fluctuation, and the cross-correlation degree. When the system's high-pressure sensor detects that the pressure is higher than 3.2 MPa for 30 seconds, and the standard deviation of pressure fluctuation is 0.15 MPa, with a correlation coefficient of 0.85 with the refrigerant flow rate, the F-value after feature fusion reaches 0.92, exceeding the fault judgment threshold of 0.88. The system determines that it is a pipeline blockage fault and immediately executes secondary protection: reducing the compressor frequency to 50% of the rated value, shutting down the phase change component, activating local audible and visual alarms, and sending a remote notification to the property operation and maintenance terminal through the data interaction module, along with the operating data for 10 minutes before and after the fault.

[0046] The data interaction and iterative optimization module supports three protocols: MQTT, Modbus, and OPCUA. It locally stores nearly 90 days of operational data, including sensor readings, control commands, heat exchange efficiency, and fault records, in JSON format. Operational data and heat exchange efficiency curves are uploaded to the cloud platform weekly. The cloud platform calibrates the control parameters using the least squares method. After calibration, the PID parameters are better suited to the load fluctuation characteristics of residential heating, and the system's heat exchange efficiency is continuously optimized.

[0047] Table 1. Performance Comparison of Invention and Traditional Systems in Residential Heating Scenarios in Northern China Table 1 shows data from 10 sets of continuous operation test results during the heating season of this embodiment, covering different ambient temperatures from -25℃ to 10℃. Traditional systems use single-parameter load prediction and fixed PID control, with phase change components starting and stopping based on fixed temperature thresholds, and fault diagnosis only monitoring surface parameters. This invention, through a multi-feature fusion load prediction model, reduces the prediction error from ±11.5% of the traditional system to ±2.8%, laying the foundation for precise control; adaptive PID-fuzzy fusion control increases the system COP value from 3.1 to 3.8, reducing energy consumption by 20.6%; scenario-based phase change collaborative logic improves phase change energy utilization by 44%, fully leveraging the low-temperature compensation effect of phase change materials; the multi-feature fusion fault diagnosis algorithm increases accuracy from 75.3% to 96.2%, effectively reducing false shutdowns; heating temperature fluctuations are narrowed from ±1.8℃ to ±0.5℃, significantly improving user experience and fully verifying the technical advantages of this invention in low-temperature, high-load heating scenarios.

[0048] Example 2: Application of hot water supply in commercial buildings in southern China This embodiment applies to a centralized hot water supply scenario in a commercial office building in southern China. The daily hot water demand is 800L, with peak usage concentrated between 12-1 PM and 7-9 PM. The ambient temperature ranges from 5°C to 35°C, and the system primarily operates under low to medium load conditions. The system needs to address issues such as inaccurate load forecasting, high energy consumption under high temperature and low load conditions, and a high rate of false fault diagnosis. This system achieves efficient hot water supply through precise regulation and iterative optimization of an intelligent control software system, fully implementing all the technical solutions in the claims. The specific implementation process is as follows.

[0049] In terms of system configuration, the heat pump hardware includes a compressor, a high-efficiency composite heat exchanger, a stepping electronic expansion valve, an 800L insulated water tank, and phase change auxiliary components. The intelligent control software system integrates five core modules. Sensors at the sensing layer are deployed as required: ambient temperature and humidity sensors are installed on the equipment platform on the building roof; system pressure, temperature, and flow sensors are deployed at corresponding pipeline locations; three water temperature sensors in the water tank are arranged along the height; and a water usage status sensor is installed on the main outlet pipe. Sensor data, after filtering, is uploaded to the decision-making layer in real time via an industrial bus to ensure the stability and accuracy of data transmission.

[0050] When the load forecasting module is running, it takes six types of characteristic parameters as input: historical ambient temperature and humidity data for the past 45 days, water temperature change curves in the storage tank, water usage distribution during the past two months in the office building (peak times at midday and evening), and historical energy consumption data of the heat exchange system, totaling 15,000 training samples. An improved LSTM neural network model outputs the predicted heat exchange load value one hour in advance. Considering the concentrated water usage in commercial buildings, the prediction model strengthens the weight of the water usage distribution parameters through an attention mechanism, keeping the prediction error within ±2.5%, thus providing support for parameter pre-adjustment during peak hours.

[0051] The intelligent heat exchange regulation module employs an adaptive PID-fuzzy control fusion algorithm to collect four types of operating parameters in real time. When the ambient temperature is 28℃ and the predicted load is 25% of the rated load (low load scenario), the initial PID parameters are Kp=0.6, Ki=0.08, and Kd=0.2. After fuzzy inference, the outputs are ΔKp=-0.1, ΔKi=-0.02, and ΔKd=0.05. After correction, Kp=0.5, Ki=0.06, and Kd=0.25. The compressor operating frequency is adjusted to 35Hz, and the electronic throttling device opening is adjusted to 30%, avoiding ineffective energy consumption under low load. Five minutes before the peak water usage period, when the predicted load rises to 75% of the rated load, the system pre-adjusts the compressor frequency to 50Hz and the throttling device opening to 65%, ensuring a stable hot water supply temperature during peak hours.

[0052] The phase change collaborative control module assigns weighting coefficients k1=0.1, k2=0.3, and k3=0.6 according to high temperature and low load scenarios. The phase change components remain in heat storage mode, and only start heat release when the heat exchange efficiency suddenly drops by more than 20%. When the ambient temperature is 32℃ and the load is 30% of the rated load, if the heat exchange efficiency is lower than 70% of the optimal value, the system immediately starts phase change heat release. After 15 minutes, the heat exchange efficiency recovers to 85% of the optimal value, and then the phase change components switch back to heat storage mode to ensure efficient energy utilization.

[0053] The fault diagnosis module monitors the deep characteristics of parameters in real time. When the water temperature sensor in the storage tank detects no change in water temperature for 30 seconds and the correlation coefficient with the heat exchanger outlet temperature is below 0.3, and the F-value after feature fusion reaches 0.95, the system determines it to be a sensor malfunction and executes Level 1 protection: adjusting the PID parameters to increase the weight of the integral coefficient and continuously monitoring for 30 seconds. If the anomaly is not eliminated, Level 2 protection is executed and an alarm message is sent. Maintenance personnel quickly replace the sensor according to the fault handling suggestions provided by the system, and the system returns to normal operation after a restart.

[0054] The data interaction and iterative optimization module enables bidirectional data transmission between the local machine and the cloud platform. The local machine stores nearly 90 days of operating data, while the cloud platform calibrates the control parameters every 10 days. The algorithm model is upgraded online through differential upgrade technology, with the upgrade time controlled within 4 minutes, so as not to affect the normal supply of hot water.

[0055] Table 2 Performance Comparison of Invention and Traditional Systems in Hot Water Supply Scenarios of Commercial Buildings Table 2 shows data from 15 sets of continuous operation test results in this embodiment, covering different ambient temperatures from 5℃ to 35℃ and different load scenarios. Traditional systems lack accurate load prediction and dynamic parameter adjustment capabilities, resulting in high energy consumption, high fault misjudgment rate, and delayed hot water supply response under low load. This invention reduces the prediction error from ±13.2% of traditional systems to ±2.5% through a multi-feature load prediction model, providing support for energy-saving operation under low load; adaptive PID-fuzzy fusion control reduces low load energy consumption by 22.3%, significantly improving energy-saving effects; the multi-feature fusion fault diagnosis algorithm reduces the misjudgment rate from 24.7% to 3.8%, reducing unnecessary downtime; the pre-adjustment mechanism shortens the hot water supply response time from 8.5 seconds to 3.2 seconds, meeting the centralized water demand of commercial buildings; cloud platform parameter calibration further improves the system COP by 8.5%, achieving continuous performance optimization, fully verifying the practicality and advancement of this invention in medium-high temperature and medium-low load hot water supply scenarios.

[0056] Reference Figure 3 This diagram visually demonstrates the high-precision advantage of the load prediction module in this invention. Traditional systems rely solely on a single ambient temperature parameter for load prediction, neglecting crucial factors such as humidity and user water usage habits. Prediction errors typically exceed 10%, leading to frequent compressor start-ups and shutdowns, delayed adjustments in throttling components, and energy waste. This invention utilizes an improved LSTM neural network algorithm, integrating six types of multi-dimensional feature parameters and introducing layer normalization and adaptive learning rate optimization models. This controls the prediction error under different ambient temperatures to within ±3%, providing accurate predictive basis for subsequent PID parameter correction and phase change component start-up and shutdown. This avoids blind adjustments from the outset, ensuring the system always operates in an optimal load-matching state.

[0057] Reference Figure 4This diagram clearly demonstrates the core value of the intelligent heat exchange regulation module of this invention. Traditional systems use fixed-parameter PID control, which cannot dynamically correct parameters in response to changes in ambient temperature. This results in a sharp drop in heat exchange efficiency at low temperatures, with a COP of only 2.5 at -20℃. This invention innovatively employs an adaptive PID-fuzzy control fusion algorithm, dynamically correcting PID parameters through fuzzy rules. It adjusts the compressor frequency and throttling device opening in real time based on load deviation and the rate of change of deviation, achieving a significant improvement in COP across the entire temperature range. Especially in the core heating scenario of -10℃ in northern China, the COP increases from 3.0 to 3.8, reducing energy consumption by 21%, thus solving the industry pain point of low-temperature inefficiency in traditional systems.

[0058] Reference Figure 5 This figure highlights the scenario-based decision-making advantages of the phase change collaborative control module of this invention. Traditional systems trigger the start and stop of phase change components solely based on a single ambient temperature threshold, resulting in a mismatch between the heat charging and discharging sequence and the heat exchange load. Energy utilization is generally below 70%, leading to energy waste of the phase change material. This invention, through a scenario-based decision-making algorithm, dynamically allocates weight coefficients according to different operating conditions, comprehensively considering load forecasts, real-time heat exchange efficiency, and the rate of change of ambient temperature to determine the start and stop timing. This enables precise coordination between the heat charging and discharging of the phase change components and the main unit's operation, improving energy utilization by more than 20% under all operating conditions, reaching 89% under low-temperature, high-load scenarios. This fully leverages the energy compensation role of the phase change material, further enhancing the overall energy efficiency of the system.

[0059] Reference Figure 6 This diagram visually illustrates the accuracy advantage of the fault diagnosis module in this invention. Traditional systems only monitor single parameter exceeding thresholds, failing to extract deeper features such as parameter fluctuation amplitude and cross-correlation, resulting in a high false alarm rate. The diagnostic accuracy for different fault types is below 80%, easily leading to ineffective shutdowns or missed faults. This invention integrates multi-feature fusion and fuzzy inference algorithms, constructing a feature library covering 12 fault categories. It quantifies the degree of anomaly through feature fusion formulas, achieving a diagnostic accuracy of over 94% for different fault types, with an overall accuracy improvement of 23%. This significantly reduces the false alarm rate, shortens fault handling time, and enhances system operational stability.

[0060] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A high-efficiency heat exchange air source heat pump system, characterized in that, Includes the following modules: The core innovation of heat pumps focuses on embedded intelligent control software systems, with the hardware itself serving only as an execution carrier; this software system achieves multi-module collaboration, precise prediction, and dynamic regulation. The software system adopts a closed-loop architecture, which is divided into a perception layer, a decision layer, an execution layer and an interaction layer. Each layer completes seamless data flow through an industrial-grade bus protocol. The decision-making layer integrates five core functional modules, sharing information and working collaboratively based on a unified data bus: the load forecasting module adopts an improved LSTM neural network algorithm, inputting six types of feature parameters to construct a multi-dimensional model; The intelligent heat exchange regulation module adopts an adaptive PID-fuzzy control fusion algorithm, collects four types of real-time parameters to dynamically correct the PID coefficient, adjusts the compressor frequency and the opening of the electronic throttling component, and establishes a dynamic matching mechanism to maintain the optimal heat exchange efficiency. The phase change collaborative control module is equipped with a scenario-based decision-making algorithm, which comprehensively defines the start-stop threshold and timing of the phase change component by integrating three types of parameters, and works in collaboration with the host. The fault diagnosis module integrates multi-feature fusion and fuzzy inference algorithms to build a fault feature library of 12 types, automatically generate graded handling suggestions and trigger alarms; the data interaction and iterative optimization module is compatible with multiple protocols such as MQTT, Modbus, and OPCUA, completes bidirectional data transmission, and iteratively optimizes algorithms and parameters based on big data from the cloud platform.

2. The high-efficiency heat exchange air source heat pump system according to claim 1, characterized in that, The improved LSTM neural network algorithm of the load prediction module reduces the risk of gradient vanishing by introducing layer normalization technology. The training process uses an adaptive learning rate optimizer, which automatically halves the learning rate when the loss value decreases by less than 0.5% for five consecutive iterations. The training sample size is no less than 10,000 groups, and the prediction duration supports three levels: 1 hour, 2 hours, and 3 hours.

3. The high-efficiency heat exchange air source heat pump system according to claim 1, characterized in that, The adaptive PID-fuzzy control fusion algorithm of the intelligent heat exchange regulation module has a fuzzy rule base containing a 5×5×5 three-dimensional fuzzy matrix. The input quantities are load deviation and deviation change rate, and the output quantities are PID parameter correction quantities. The load deviation is divided into 5 fuzzy subsets: negative large, negative small, zero, positive small, and positive large. The deviation change rate is also divided into 5 fuzzy subsets. The smooth correction of parameters is completed through fuzzy inference.

4. The high-efficiency heat exchange air source heat pump system according to claim 1, characterized in that, The scenario-based decision-making algorithm of the phase change collaborative control module presets three core operating scenarios, each with an independent phase change component start / stop threshold and collaborative logic: In the low temperature and high load scenario, the phase change component and the host operate synchronously, with a charge / discharge switching cycle of 15 minutes; In the normal temperature and medium load scenario, the phase change component is only started at the initial stage of host startup and automatically shuts down after 30 minutes; In the high temperature and low load scenario, the phase change component remains in a heat storage state and only starts to release heat when the heat exchange efficiency suddenly drops.

5. The high-efficiency heat exchange air source heat pump system according to claim 1, characterized in that, The multi-feature fusion algorithm of the fault diagnosis module extracts deep features including parameter over-threshold duration, parameter fluctuation standard deviation, and correlation coefficient between different parameters. The parameter over-threshold duration threshold is set to 30s, and the fluctuation standard deviation threshold is set according to the parameter type.

6. The high-efficiency heat exchange air source heat pump system according to claim 1, characterized in that, The data interaction and iterative optimization module supports both local and cloud storage modes. Local storage can store nearly 90 days of operating data, while the cloud platform supports remote upgrades of algorithm models and batch calibration of control parameters. The upgrade process does not affect the normal operation of the system, and differential upgrade technology is used to reduce the amount of data transmission.

7. A control method for a high-efficiency heat exchange air source heat pump system, characterized in that, The intelligent control software system according to claim 1 includes the following steps: Step 1: The perception layer collects seven types of real-time data through sensors, including ambient temperature, humidity, system high and low pressure, heat exchanger inlet and outlet temperatures, refrigerant flow rate, water tank temperature, and user water usage status. The data is then filtered and uploaded to the decision layer. Step 2: The load forecasting module calls the improved LSTM neural network algorithm, inputs historical data and real-time data, and outputs the heat exchange load forecast value 1-3 hours later; Step 3: The intelligent heat exchange regulation module generates control commands for compressor frequency and throttling component opening based on the load forecast value and real-time load deviation through an adaptive PID-fuzzy fusion algorithm. Step 4: The phase change collaborative control module controls the start-up, shutdown, and heat charging / discharging status of the phase change components based on the load forecast, real-time heat exchange efficiency, and ambient temperature through a scenario-based decision-making algorithm. Step 5: The fault diagnosis module monitors for data anomalies in real time and triggers the fault protection mechanism. Step six: The data interaction module uploads the running data, receives the algorithm upgrade instructions from the cloud platform, and completes the iterative optimization of system performance.

8. The control method according to claim 7, characterized in that, The parameter control logic in step three is as follows: when the load forecast value is higher than 120% of the real-time load, the compressor frequency is increased to 80% of the rated value 5 minutes in advance, and the opening of the throttling component is pre-adjusted to the appropriate position; when the load forecast value is lower than 80% of the real-time load, the compressor frequency is gradually reduced, and the heat storage time of the phase change component is extended; when the deviation between the load forecast value and the real-time load is within ±10%, the parameters are kept stable.

9. The control method according to claim 7, characterized in that, The phase change coordination logic in step four is as follows: In low temperature and high load scenarios, the phase change component first stores heat for 10 minutes, and then releases heat synchronously with the host; in normal temperature and medium load scenarios, the phase change component only releases heat at the initial stage of host startup, and switches to heat storage after the water temperature in the storage tank rises to 80% of the target value; in high temperature and low load scenarios, the phase change component maintains the heat storage state, and automatically starts to release heat when the heat exchange efficiency is lower than 80% of the optimal value, and the heat release duration is dynamically adjusted according to the load deviation.

10. The control method according to claim 7, characterized in that, The fault protection logic in step five is as follows: when an abnormal parameter is detected, first-level protection is executed; if the abnormality is not eliminated, second-level protection is executed; if it is still not eliminated, third-level protection is executed; after the fault is eliminated, the system executes a restart and warm-up procedure to gradually restore the parameters to the normal range.

Images