Constant-temperature and constant-pressure hot water system based on heat pump and intelligent control method thereof

Abstract

The invention discloses a constant-temperature and constant-pressure hot water system based on a heat pump and an intelligent control method of the constant-temperature and constant-pressure hot water system, and belongs to the technical field of heat generation devices.The intelligent control method comprises the steps that multi-element sensing information containing performance degradation data is collected in real time, and a state observation space fusing the equipment health state and instantaneous working condition characteristics is constructed; the space is input into a self-adaptive decoupling control model with the online learning capability, a water pump rotating speed and water mixing valve opening degree instruction which are cooperatively optimized is generated, and dynamic decoupling of a pressure loop and a temperature loop is achieved; and finally, performing closed-loop parameter self-correction on the model based on the control residual error and the performance degradation data. Slow-varying factors such as equipment aging are regarded as observable and compensable total disturbance, dependence on an accurate fixed model is not needed, high-precision and high-stability constant-temperature and constant-pressure water supply of the system is adaptively maintained in the whole life cycle, the energy efficiency is effectively improved, and abrasion of an executing mechanism is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of heat generating device, in particular to a constant temperature and pressure hot water system based on heat pump and an intelligent control method thereof. BACKGROUND

[0002] In the field of modern civil and industrial buildings, it has become a core requirement for building water supply and drainage design to provide a stable and energy-saving hot water supply system. The constant temperature and pressure hot water system based on heat pump has been widely used due to its energy conversion efficiency and low carbon characteristics. The system is usually composed of key components such as heat pump unit, heat preservation water tank, variable frequency water pump group, electric regulating valve and intelligent controller, and realizes the stability of water supply through automatic control means. Its basic working logic is: the pressure sensor is used to monitor the pressure change of the pipe network, and the frequency converter is used to adjust the water pump speed in real time to maintain constant pressure water supply; at the same time, the temperature sensor is used to monitor the outlet water temperature, and the heat pump running power or the water valve opening degree is adjusted to ensure the constant temperature of the end water.

[0003] However, the existing constant temperature and pressure hot water system based on heat pump generally faces the challenges of regulation stability and long-term reliability in actual operation. The mainstream scheme currently adopts a double-loop independent PID control architecture, that is, a closed-loop feedback of pressure and temperature is established respectively. Although some schemes introduce flow meter feedforward compensation, decoupling control algorithm or physical decoupling tank structure, there are still significant defects in dealing with complex dynamic conditions. Since pressure regulation belongs to a fast response process of milliseconds, and temperature regulation is limited by heat transfer and fluid delivery delay, which is a slow process of minutes, the significant heterogeneity in time scale, combined with the strong coupling characteristics of pressure and temperature in the physical level, often leads to continuous cross-variable oscillation of the system when the water consumption fluctuates sharply. In addition, the existing technology relies heavily on accurate and static mathematical models, and lacks online perception and adaptive correction capability for performance degradation in the whole life cycle of the system (such as heat exchanger fouling, water pump efficiency decline, actuator wear, etc.). This leads to a serious mismatch of controller parameters when the system enters the aging period, which not only reduces the user's water comfort, but also reduces the energy efficiency ratio due to frequent power fluctuations, shortening the service life of the equipment. SUMMARY

[0004] The present application overcomes the shortcomings of the prior art and provides a constant temperature and pressure hot water system based on heat pump and an intelligent control method thereof.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is as follows: a constant temperature and pressure hot water intelligent control method based on heat pump, comprising:

[0006] S1, receiving real-time multi-sensor data and pre-processing; wherein the multi-sensor data includes pressure, temperature, flow and performance degradation data;

[0007] S2, based on the multi-sensor data, constructing a state observation space that integrates device performance degradation indicators and instantaneous working condition characteristics;

[0008] S3, inputting the state observation space into an adaptive decoupling control model to generate a first control instruction for the variable frequency water pump speed and a second control instruction for the water mixing valve opening degree;

[0009] S4, adjusting the drive frequency of the variable frequency water pump according to the first control instruction to change its output flow, and adjusting the cold and hot water mixing ratio of the water mixing valve according to the second control instruction to correct the mixed water temperature;

[0010] S5, evaluating the residual error between the current control effect and the preset target, and combining the performance degradation data to perform closed-loop parameter self-correction on the adaptive decoupling control model.

[0011] In a preferred embodiment of the present application, step S1 comprises:

[0012] The obtained performance degradation data includes abnormal temperature rise trend of compressor operating current, associated offset of variable frequency water pump motor speed and torque at the same frequency, and heat balance residual error between heat exchanger inlet and outlet temperature difference in the heat preservation water tank and heat pump input power;

[0013] Time synchronization and alignment are performed on all sensor signals;

[0014] The original signal is nonlinearly smoothed by using sliding window median filtering technology, and the statistical criteria are used to distinguish and remove outliers;

[0015] The cleaned data is standardized mapped to generate dimensionless feature operators.

[0016] In a preferred embodiment of the present application, the state observation space includes dimensions of water pressure deviation, mixed water temperature deviation, instantaneous flow rate, compressor efficiency ratio trend, water pump efficiency attenuation index and heat exchanger fouling index; wherein the water pressure deviation is defined as the residual error between the preset target pressure value and the current measured pressure value; the mixed water temperature deviation is defined as the residual error between the preset target temperature value and the current measured temperature value; and the instantaneous flow rate is calculated by first-order difference calculation on the flowmeter reading.

[0017] In a preferred embodiment of the present application, the adaptive decoupling control model adopts a deep neural network architecture, the input layer of which receives a seven-dimensional state observation vector containing the performance degradation indicators; the deep neural network includes multiple fully connected layers, and a multi-head attention mechanism module is introduced in the middle layer to dynamically calculate and weight the influence degree of each state dimension on the pressure response and temperature adjustment.

[0018] In a preferred embodiment of the present application, the process of generating the first control instruction and the second control instruction includes: the output layer of the deep neural network is processed by two parallel instruction mapping branches; wherein the pressure control branch combines the instantaneous flow rate change rate to map the target rotating speed of the variable frequency water pump in real time, and generates the first control instruction; the temperature control branch combines the heat exchanger fouling index and the energy efficiency trend to map the target opening degree of the water mixing regulating valve under the premise of compensating for the large inertia lag of heat transfer, and generates the second control instruction.

[0019] In a preferred embodiment of the present application, the adaptive decoupling control model is closed-loop parameter self-corrected, including:

[0020] The pressure control residual and the temperature control residual in the current time window are calculated;

[0021] When the residual continuously exceeds the preset threshold, a model fine-tuning program is triggered;

[0022] The parameters of the adaptive decoupling control model are locally corrected by using an online learning algorithm in combination with the performance degradation data and the working condition residual;

[0023] Wherein, the objective function of model optimization simultaneously considers the control error, the total energy consumption of the system, and the change amplitude of the control instruction.

[0024] In a preferred embodiment of the present application, the closed-loop parameter self-correction adopts a gradient increment learning strategy based on a time window, specifically including:

[0025] The root mean square value of the control error is calculated with a fixed time length as a statistical window;

[0026] If the root mean square values of the errors of continuous multiple statistical windows exceed the threshold, fine-tuning is started;

[0027] The fine-tuning adopts a small-batch gradient descent algorithm, and introduces a weight decay term, and only updates the related parameters of the output layer and the attention module in the adaptive decoupling control model.

[0028] The present application provides a constant temperature and constant pressure hot water system based on a heat pump, comprising:

[0029] A data fusion module is configured to receive and synchronously process multi-source sensing signals from the pressure sensor, the temperature sensor, the flow meter and the performance degradation monitoring unit, and generate a standardized state observation vector;

[0030] An adaptive decoupling control module is configured to input the state observation vector into a pre-trained two-variable collaborative optimization network, and output a water pump rotating speed instruction and a water mixing valve opening degree instruction;

[0031] An execution driving module is configured to convert the water pump rotating speed instruction into a frequency control signal of a frequency converter, and convert the water mixing valve opening degree instruction into a position control signal of an electric actuator;

[0032] A performance degradation evaluation module is configured to calculate a compressor energy efficiency trend, a water pump efficiency attenuation index and a heat exchanger fouling index based on device operation parameters in real time, and generate a device health state report;

[0033] A model online updating module is configured to periodically fine-tune parameters of the two-variable collaborative optimization network according to control residuals and device health states, and maintain long-term stability of control performance.

[0034] The present application provides an electronic device, comprising:

[0035] at least one processor; and

[0036] a memory connected to the at least one processor in communication; wherein,

[0037] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the intelligent control method of the constant temperature and constant pressure hot water system based on the heat pump.

[0038] The present application provides a computer readable storage medium, the computer readable storage medium stores computer instructions, the computer instructions are used to make the processor execute to realize the intelligent control method of the constant temperature and constant pressure hot water system based on the heat pump.

[0039] The present application solves the defects in the background art, and has the following beneficial effects:

[0040] (1) This invention constructs a state observation space that integrates the performance degradation index of the equipment and introduces an active disturbance rejection architecture with cross-variable predictive capabilities. It treats physical coupling and performance degradation as a total disturbance and performs dynamic compensation, enabling the controller to see the hardware health status and instantaneous operating conditions in real time. It directly realizes the deep collaboration between pressure response and temperature regulation on heterogeneous time scales, avoiding the frequent oscillation of minute-level temperature loops induced by millisecond-level pressure fluctuations. Compared with the problem of mutual disturbance and loss of control caused by independent dual-loop PID in the prior art, it further ensures high-precision constant temperature and pressure water supply quality under complex dynamic water load and improves the robustness of the system in sudden change scenarios.

[0041] (2) This invention uses a bivariate collaborative optimization network and online learning mechanism in the adaptive decoupling control model to explicitly incorporate equipment performance degradation data into the control decision. It dynamically adjusts the control weights based on health dimensions such as the compressor energy efficiency ratio trend and the water pump efficiency decay index, directly overcoming the control performance degradation caused by slow-changing factors such as heat exchanger scaling or water pump wear. Compared with the limitations of existing technologies that rely on fixed decoupling matrices and become interference sources after feedforward compensation failure, this invention further enables the system to maintain stable energy efficiency without human intervention throughout its entire life cycle, reducing the risk of energy consumption increase and adjustment misalignment caused by equipment aging.

[0042] (3) This invention establishes a closed-loop self-correction mechanism based on gradient incremental learning and introduces a multi-objective loss function that balances stability quality and operating energy consumption. It uses control residuals combined with degradation feature data to fine-tune the output layer of the neural network online, thereby realizing the self-updating capability of the control algorithm as it evolves with actual operating conditions. Compared with the existing technology, which is unable to distinguish the source of disturbance and forces the actuator to over-respond, this invention further extends the service life of key hardware and reduces maintenance costs. At the same time, it avoids overshoot caused by mutual interference between fast and slow loops through predictive hedging adjustment, thereby enhancing long-term operational reliability. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart of a preferred embodiment of the present invention;

[0045] Figure 2 This is a system physical structure diagram of a preferred embodiment of the present invention;

[0046] Figure 3This is a control model architecture diagram of a preferred embodiment of the present invention. Detailed Implementation

[0047] 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.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0049] Application Overview:

[0050] This application primarily addresses the performance degradation problem of hot water supply systems in civil and industrial buildings under complex dynamic operating conditions and long-term aging scenarios. Hot water systems exhibit a characteristic of control separation but physical coupling: pressure regulation aims for millisecond-level dynamic response, while temperature changes involve heat exchange and minute-level processes with large inertial delays. Since the pressure loop is a fast process at the millisecond level, while the temperature loop is a slow process at the minute level, forcibly combining the two in the same PID logic will induce unnecessary frequent oscillations and mechanical wear in the temperature regulation mechanism, leading to extreme system instability. Therefore, existing technologies generally adopt an independent dual-loop PID architecture, but essentially treat pressure and temperature as isolated variables that do not interfere with each other; this causes mutual oscillations in the two loops when the water load changes abruptly, making it difficult to maintain high-precision water supply quality.

[0051] Existing technologies address these issues through decoupling algorithms and feedforward compensation. However, this approach heavily relies on the precise physical parameters of the equipment in its initial state. In actual operation, heat exchanger scaling increases thermal resistance, and pump wear causes characteristic curve shifts. These slowly varying factors gradually render the originally fixed decoupling matrix or compensation gain ineffective, even evolving into new sources of system interference. Using conventional control techniques in a time-varying, variable-parameter system like a hot water system will prevent the system from automatically adjusting its control strategy based on equipment degradation, leading to excessive wear and tear on actuators during high-frequency oscillations.

[0052] This application constructs a state observation space that integrates equipment performance degradation indicators and introduces an active disturbance rejection architecture with cross-variable predictive capabilities. It treats physical coupling and performance degradation as a total disturbance and performs dynamic compensation, thereby achieving deep synergy between pressure response and temperature regulation without the need for an accurate model, ensuring the energy efficiency stability of the system throughout its entire life cycle.

[0053] Example 1:

[0054] Exemplary method:

[0055] like Figure 1 As shown, a smart control method for constant temperature and pressure hot water based on a heat pump includes the following steps:

[0056] S1. Receive real-time multi-sensor data and preprocess it; the multi-sensor data includes: pressure, temperature, flow rate and performance degradation data;

[0057] S2. Based on multi-sensor data, construct a state observation space that integrates equipment performance degradation indicators and instantaneous operating condition characteristics;

[0058] S3. Input the state observation space into the adaptive decoupling control model to generate the first control command for the speed of the variable frequency water pump and the second control command for the opening of the mixing valve.

[0059] S4. Adjust the drive frequency of the variable frequency water pump according to the first control command to change its output flow rate; adjust the hot and cold water mixing ratio of the mixing regulating valve according to the second control command to correct the temperature of the mixed water.

[0060] S5. Evaluate the residual between the current control effect and the preset target, and combine the performance degradation data to perform closed-loop parameter self-calibration on the adaptive decoupling control model.

[0061] Step S1, as the starting point of the control process, is responsible for collecting and preprocessing multi-source signals. Through in-depth cleaning and synchronous processing of multi-source heterogeneous sensor signals, complex physical field changes are transformed into high-fidelity feature streams that can be directly called by the decision-making model.

[0062] In step S1, real-time multi-sensor data is received;

[0063] Among them, pressure data refers to the pressure value fed back in real time by the pressure sensor installed at the beginning of the water supply network, which is used to characterize the instantaneous fluctuation of water consumption at the user end; temperature data refers to the real-time temperature signal of the mixed hot water collected by the temperature sensor set at the output end of the mixing valve, which is used to reflect the final thermodynamic state after heat exchange and flow mixing; flow data refers to the instantaneous water flow of the system main pipe obtained by the flow meter, which is used to capture the characteristics of load step changes.

[0064] Performance degradation data refers to a set of features that can quantitatively describe the deviation of system hardware performance from the initial baseline state. Specifically, it includes: abnormal temperature rise trend of compressor operating current, correlation deviation of variable frequency water pump motor speed and torque at the same frequency, and thermal balance residual between the inlet and outlet temperature difference of heat exchanger in the insulated water tank and the input power of heat pump.

[0065] All sensor signals are transmitted to the central controller via an industrial fieldbus (such as Modbus RTU or CANopen), with the sampling frequency uniformly set to 1Hz to ensure the alignment of each signal in the time dimension.

[0066] To address electromagnetic interference and impulse noise in the original signal, a sliding window midpoint filtering technique is employed for nonlinear smoothing. Furthermore, the 3σ criterion is used to identify and remove outliers acquired instantaneously, ensuring the authenticity of the input data. Finally, the preprocessed data undergoes a standardization mapping process, transforming physical quantities with vastly different dimensions into dimensionless feature operators for subsequent use in constructing the state observation space.

[0067] By synchronizing the time, the system can eliminate the phase difference caused by asynchronous data, thus reserving sufficient computational margin for the controller before millisecond-level pressure fluctuations cause minute-level temperature disturbances; and by explicitly preprocessing performance degradation data, the system has the ability to identify hardware performance drift.

[0068] In real-world operating conditions, heat exchanger scaling leads to increased thermal resistance over time, while pump efficiency continuously declines due to impeller wear. With existing technology, the controller can only detect deviations in outlet water temperature or pressure, but cannot determine whether these deviations are temporary disturbances caused by fluctuations in user water consumption or systemic control performance degradation due to equipment aging. Attempts to maintain system performance by blindly increasing proportional gain or feedforward compensation are often unsustainable in scenarios like hot water systems where fast and slow loops are strongly coupled, easily inducing high-frequency oscillations in the actuators.

[0069] Step S2 transforms the discrete physical quantities of S1 into feature vectors, and constructs a state observation space that integrates equipment performance degradation indicators and instantaneous operating condition characteristics.

[0070] Among them, the state observation space refers to a mathematical vector space composed of multiple independent dimensions describing the dynamic characteristics of the system. It not only includes the deviation dimension that characterizes the control target, but also explicitly integrates the health dimension that reflects the physical characteristics of the equipment. Specifically, it includes: water pressure deviation, mixed water temperature deviation, instantaneous flow rate change rate, compressor energy efficiency ratio trend, water pump efficiency decay index and heat exchanger fouling index.

[0071] Water supply pressure deviation and mixed water temperature deviation are defined as the residuals between the preset target value and the current measured value, respectively; instantaneous flow rate change rate is used to characterize the severity of load fluctuations at the user end; compressor energy efficiency ratio trend refers to the evolution trajectory of the energy conversion efficiency of the heat pump unit over time during the heat generation process; water pump efficiency decay index is used to quantify the degree of loss of mechanical performance of the pump unit; heat exchanger fouling index refers to the increase in heat exchange resistance caused by scale deposition.

[0072] In the specific process of constructing the state observation space, the central controller performs secondary processing on the standardized data output from step S1 through a mathematical transformation model.

[0073] First, considering the characteristics of instantaneous operating conditions, the water supply pressure deviation is defined as the set pressure. Compared with the measured pressure The difference, i.e. The temperature deviation of the mixed water is defined as the set temperature. Compared with the measured temperature The difference, i.e. Instantaneous flow rate change It is obtained by performing a first-order difference calculation on the flow meter reading, i.e. This allows us to capture dynamic trends in users' water usage behavior.

[0074] The system has entered the in-depth construction phase of performance degradation indicators. For constructing the compressor energy efficiency ratio trend, the system uses a unit heating capacity power consumption model for back-calculation, and the calculation formula is as follows: ,in, The mass flow rate through the heat pump side, $c_p$ is the specific heat capacity of water at constant pressure. These are the real-time water temperatures at the inlet and outlet of the heat pump unit, respectively. This represents the real-time electrical power of the compressor. This indicator reflects the current energy conversion capability of the heat pump unit.

[0075] For the water pump efficiency degradation index Its construction process is based on the ratio of motor input power to fluid output power: ,in, For fluid density, It is the acceleration due to gravity. For instantaneous flow, This is the current head of the water pump. This refers to the real-time input power of the motor, obtained through the speed-torque monitoring module. A continuous decrease in this indicator directly reflects the wear of the pump impeller and the aging of the bearings.

[0076] For heat exchanger fouling index This method uses the heat balance equation to deduce the current effective heat transfer coefficient. , ,in, This represents the actual heat power transferred by the heat pump to the water tank. The effective heat exchange area of ​​the heat exchanger. The temperature difference is the logarithmic mean. By... Compared with the reference heat transfer coefficient of the equipment in its new state By comparing the results, we can obtain the dirt index. Finally, the system introduces an ambient temperature compensation factor. The theoretical output of the heat pump is corrected by concatenating all the above dimensions into a feature vector. .

[0077] By transforming the aging factors, which were originally hidden in nonlinear fluctuations, into an independent mathematical dimension, the system can perform attribution analysis of control deviations at the data level. The controller can then detect that pressure fluctuations are caused by the user opening and closing valves. The sudden change was caused by insufficient power from the water pump at the same speed due to decreased efficiency.

[0078] In actual operation, the physical characteristics of hot water systems evolve dynamically with time and the environment. Due to the lack of explicit characterization of performance degradation indicators, existing technologies cannot distinguish the true source of deviations when facing equipment aging, and can only compensate for lag through passive feedback adjustment. Step S2 of this application upgrades the system from traditional single feedback control to full-state predictive control based on physical characteristics by constructing an observation space that can provide real-time insight into the health status of the hardware.

[0079] Step S3 is used to solve the failure problem of traditional dual-loop control in equipment aging scenarios as pointed out in the background technology. It transforms the static and fixed control logic into an adaptive decoupling control mechanism that can sense the health status of the equipment and dynamically adjust the control strategy.

[0080] In step S3, the state observation space is input into the preset adaptive decoupling control model to generate a first control command for the speed of the variable frequency water pump and a second control command for the opening of the mixing valve.

[0081] The adaptive decoupling control model consists of a bivariate co-optimization network. Historical operating data and performance degradation labels are introduced during the training phase, and the control parameter weights are dynamically adjusted through an online learning mechanism.

[0082] The adaptive decoupling control model refers to an intelligent algorithm model based on a deep learning architecture that can automatically adjust control weights according to the input state to eliminate mutual interference between multiple variables; the first control command refers to the physical electrical signal used to control the driving frequency of the variable frequency water pump, which directly determines the flow and pressure of the water supply network; the second control command refers to the adjustment signal used to drive the electric actuator of the mixing valve, which corrects the outlet water temperature by changing the mixing ratio of hot and cold water.

[0083] Bivariate collaborative optimization network refers to the logical architecture within the model used to balance two controlled objectives, pressure and temperature, and to achieve collaborative work across heterogeneous time scales; online learning mechanism refers to the system's ability to dynamically update model parameters based on real-time feedback errors during operation.

[0084] likeFigure 3 As shown, in step S3, the adaptive decoupling control model adopts a deep neural network structure. Its input layer receives the seven-dimensional state observation vector generated in step S2, and the hidden layer contains three fully connected layers with 64, 128, and 64 neurons in each layer, respectively. A multi-head attention mechanism module is introduced after the second layer to automatically weight the importance of different state dimensions. The output layer contains two independent branches, which respectively correspond to the target rotational speed of the strain frequency pump. Target opening degree of the mixing valve .

[0085] In the offline phase, the deep neural network is pre-trained using a simulation dataset that includes various aging scenarios, such as different degrees of scaling, pump efficiency degradation levels, and ambient temperature ranges, to ensure that it has prior knowledge of the slowly changing characteristics of the system.

[0086] During the online operation phase, an incremental learning algorithm is used to continuously integrate newly collected operational data to prevent catastrophic amnesia.

[0087] Within each control cycle, the model maps commands through two parallel output branches: the pressure control branch quickly generates the pump speed command based on the instantaneous flow rate change, and the first control command... The signal is sent to the frequency converter to adjust the drive frequency of the water pump motor; the temperature control branch, taking into account the heat exchange efficiency trend and the large inertial lag in heat transfer, generates precise valve opening commands, the second control command. The signal is converted into a 4-20mA analog signal or Modbus digital command to drive the electric actuator of the mixing valve to adjust the hot and cold water mixing ratio.

[0088] By explicitly incorporating equipment health status into control decisions, the model can moderately increase the pump speed to maintain flow when pump efficiency declines, and increase heat pump power or adjust the mixing ratio in advance when heat exchanger scaling occurs, thereby suppressing control deviations caused by equipment aging at the source.

[0089] The aforementioned adaptive decoupling control model construction method, through end-to-end mapping of a neural network, allows the system to predict potential disturbances to the temperature loop during pressure control command generation via weighted correlation, and to preemptively mitigate these disturbances in the valve opening command. The reason for using an adaptive decoupling control model instead of traditional PID or static decoupling algorithms in step S3 is that existing technologies typically employ independent PID controllers. When faced with strong coupling interference caused by sudden changes in water consumption, the two loops often experience severe mutual interference oscillations due to inconsistent adjustment paces. Even if some existing technologies employ feedforward decoupling or diagonal matrix decoupling, these decouplers often rely on precise mathematical models from the initial equipment stage. Once the system enters its aging phase, with scaling or pump wear, the original decoupling matrix may completely fail due to physical parameter mismatch, or even evolve into a new source of interference, leading to frequent vibrations in the actuator.

[0090] Step S4, based on the control commands generated in step S3, achieves precise control of the physical system through the actuator.

[0091] A variable frequency water pump is a power device that adjusts the motor speed by changing the power supply frequency, thereby changing the pump head and flow rate. The drive frequency refers to the AC frequency signal output by the frequency converter that directly controls the motor rotation speed. The output flow rate refers to the volume of water flowing through the water supply network per unit time. A mixing valve is a three-way regulating mechanism that can dynamically adjust the mixing ratio of high-temperature hot water and low-temperature return water. The hot and cold water mixing ratio refers to the flow ratio of the two fluids entering the regulating valve in the mixing chamber, which directly determines the final water supply temperature.

[0092] The inlet of the variable frequency water pump is connected to the outlet of the insulated water tank through the second pipe, and its outlet is connected to the first input end of the mixing valve through the third pipe; the second input end of the mixing valve is connected to the end of the return water network through the fourth pipe to receive the low-temperature return water after use by the user; the output end of the mixing valve is connected to the beginning end of the water supply network to provide the user with constant temperature and pressure hot water.

[0093] When the first control command is received Then, the frequency converter adjusts the AC frequency output to the water pump motor according to the preset speed-frequency mapping relationship, driving the water pump motor speed to change step or smoothly, thereby establishing the target pressure at the beginning of the water supply network.

[0094] When the second control command is received Then, the electric actuator of the mixing regulating valve drives the valve core to rotate, changing the flow cross-sectional area ratio of the cold (return water) and hot (from the water pump) water flows, thereby precisely controlling the outlet water temperature after mixing.

[0095] This execution process enables coordinated regulation of the hydraulic and thermal loops, avoiding oscillations and overshoot caused by strong coupling of dependent variables in traditional dual-loop control.

[0096] For example, when a sudden large-scale water usage at the user end causes a sharp drop in pressure, the rapid increase in the pump speed will instantly increase the flow rate through the heat pump and disturb the water temperature. However, since the mixing valve adjustment action in step S4 has been pre-compensated according to the decoupling algorithm, this physical hydraulic impact is transformed into a small fluctuation within the thermal equilibrium range, thereby ensuring the dual stability of the outlet water temperature and pressure.

[0097] In step S5, after the adjustment action of the physical actuator is completed, the system needs to provide quantitative feedback on the achievement of the control target and perform self-evolution at the algorithm level to address the performance degradation caused by long-term operation of the equipment, which is the closed-loop guarantee link of the intelligent control method of the present invention.

[0098] The current control effect refers to the physical quantities of water supply pressure and outlet temperature fed back in real time by pressure and temperature sensors; the preset target refers to the reference pressure and reference temperature values ​​set by the user; the residual refers to the deviation between the control effect and the preset target, usually expressed as the root mean square error (RMSE); closed-loop parameter self-calibration refers to the process by which the system automatically adjusts the weights and biases of neurons within the adaptive decoupling control model based on the operating results; online fine-tuning logic refers to the incremental updates of local model parameters without interrupting the normal operation of the system, so that they continuously match the current physical environment.

[0099] Specifically, first calculate the pressure residual within the current window. With temperature residual When the residual value continuously exceeds the preset performance threshold, the system triggers the model fine-tuning program. At this time, the controller performs correlation analysis with the performance degradation data obtained in step S1, including heat exchanger fouling index and water pump efficiency, and uses an online learning algorithm based on gradient descent to locally correct the control gain of the neural network output layer.

[0100] During fine-tuning, the feature extraction portion of the network's bottom layer (the first two fully connected layers) is frozen, and only the output layer and attention weights are locally updated to reduce computational overhead. The data used for updating is the runtime log from the most recent 24 hours, including state observation vectors, control commands, and actual measurement results.

[0101] Meanwhile, the multi-objective loss function dynamically adjusts the weights during training: the loss function is defined as follows: ,in, These represent the squared error terms for pressure and temperature, respectively, and are used to measure the stability of water supply quality. The total energy consumption of the system is quantified by the sum of the heat pump power and water pump power collected in real time, which characterizes the economic efficiency of the system operation. The L2 regularization term, used to control the rate of change of the command, is used to constrain the adjustment range of the pump speed and valve opening, and to prevent the actuator from experiencing high-frequency vibration. These are the weighting coefficients for each item.

[0102] For example, when the heat exchanger fouling index is >0.2, or the water pump efficiency is below 80%, Automatically increase by 20% to suppress the rise in ineffective energy consumption.

[0103] This invention constructs a state observation space that integrates equipment performance degradation information, adopts an adaptive decoupling control model with online learning capabilities, and designs a closed-loop self-correction mechanism based on control residuals and health status. Under the premise of eliminating physical decoupling tanks and abandoning feedforward compensation that relies on precise models, it effectively overcomes the performance degradation problem of traditional dual-loop PID control under system aging, environmental drift, and complex water disturbances. This enables the system to maintain high-precision constant temperature and pressure water supply capabilities throughout its entire life cycle, while optimizing long-term operating energy efficiency.

[0104] Example 2:

[0105] In this embodiment, based on the steps of Embodiment 1, the improvements to steps S2 and S5 are further explained in detail.

[0106] In step S2, the specific implementation of the heat exchange efficiency estimation module includes: utilizing the temperature sensor at the inlet of the insulated water tank. Outlet temperature sensor Flow meter reading and real-time electrical power of heat pump units Through the steady-state thermal balance equation Calculate the current heat exchange power, where The heat pump cycle efficiency is used to deduce the effective heat transfer coefficient. ,in, For the heat exchange area, The logarithmic mean temperature difference; Compared with the initial installation calibration By comparison, the dirt index is obtained. This is then input into the control model as one dimension of the state observation vector.

[0107] To address the closed-loop self-calibration mechanism of the control strategy, this embodiment introduces a gradient incremental learning strategy based on a time window. At the method level, the central controller calculates the root mean square value of the control error in a statistical window of 10 minutes. , Represents pressure or temperature. This represents the number of sampling points within the window. If If two consecutive windows exceed the threshold, fine-tuning is initiated. Fine-tuning uses Mini-batch Stochastic Gradient Descent (Mini-batchSGD) with a learning rate of 10. −4 Only the output layer weights and the query / key projection matrix of the attention module are updated. To prevent overfitting, an L2 weight decay term with a coefficient of 10 is introduced. −5 After fine-tuning, the new model parameters are loaded into the central controller's memory and take effect immediately.

[0108] Furthermore, the data acquisition and processing of the performance degradation monitoring unit is completed through an independent embedded coprocessor, which communicates with the main controller via an SPI bus to ensure that the monitoring task does not affect the real-time performance of the main control loop. The coprocessor periodically packages and uploads compressor current, pump speed-torque pairs, and heat exchange efficiency estimates, and the main controller updates the status observation vector accordingly.

[0109] This embodiment further enhances the system's adaptability to the equipment aging process through accurate online estimation of heat exchange efficiency and an efficient model fine-tuning mechanism, ensuring that the control performance remains stable over an operating cycle of several years.

[0110] Exemplary device:

[0111] like Figure 2 As shown, a constant temperature and pressure hot water intelligent control system based on a heat pump includes:

[0112] The data fusion module is used to receive and synchronously process multi-source sensor signals from pressure sensors, temperature sensors, flow meters, and performance degradation monitoring units to generate standardized state observation vectors.

[0113] The adaptive decoupling control module is used to input the state observation vector into a pre-trained bivariate collaborative optimization network and output pump speed command and mixing valve opening command.

[0114] The drive module is used to convert the water pump speed command into the frequency control signal of the frequency converter, and to convert the mixing valve opening command into the position control signal of the electric actuator.

[0115] The performance degradation assessment module is used to calculate the compressor energy efficiency trend, water pump efficiency decay index and heat exchanger fouling index in real time based on equipment operating parameters, and generate an equipment health status report.

[0116] The online model update module is used to periodically fine-tune the parameters of the bivariate co-optimization network based on the control residuals and equipment health status, so as to maintain the long-term stability of control performance.

[0117] Optionally, the data fusion module includes: a signal acquisition unit that receives raw data from each sensor via an industrial fieldbus interface; a time synchronization unit that aligns the timestamps of all signals based on a hardware clock; an anomaly detection unit that cleans the data using the 3σ criterion and sliding window midpoint filtering; and a feature construction unit that calculates pressure deviation, temperature deviation, flow rate change rate, and various performance degradation indicators according to preset formulas, and splices them into a seven-dimensional state vector.

[0118] Optional, the adaptive decoupling control module includes: a neural network inference engine, deployed on the ARM Cortex-A series processor of the central controller, supporting TensorFlow Lite format models; an attention weight calculation submodule, used to dynamically evaluate the importance of each state dimension; and an instruction generation submodule, which maps the network output to control signals that conform to the actuator interface specification.

[0119] Optional, the performance degradation assessment module includes: an energy efficiency calculation unit that calculates the COP trend based on heat pump power and water temperature difference; a water pump efficiency analysis unit that evaluates water pump performance by combining flow rate, head, and motor input power; a fouling index calculation unit that back-calculates the effective heat transfer coefficient of the heat exchanger and calculates the degradation rate through the heat balance equation; and a health status report generation unit that periodically outputs equipment maintenance recommendations.

[0120] Optionally, the online model update module includes: a residual statistics unit, which calculates the RMSE of the control error; a fine-tuning trigger judgment unit, which compares the RMSE with a threshold; a parameter freezing unit, which locks the weights of the underlying network; a local optimization unit, which updates only the output layer and the attention module; and a model validation unit, which ensures that the performance of the model on the validation set does not deteriorate after fine-tuning.

[0121] This invention provides an intelligent control device for a heat pump-based constant temperature and pressure hot water system. It achieves standardized processing of multi-source heterogeneous sensor signals through a data fusion module; utilizes a deep neural network to achieve coordinated optimization of hydraulic and thermal loops; a performance degradation assessment module quantifies the aging degree of equipment online; and a model online update module ensures that the control strategy dynamically evolves with the system state. This device forms a complete closed loop from state perception and control decision-making to strategy self-correction, effectively solving the performance degradation problem of traditional hot water systems during long-term operation and significantly improving system reliability, energy efficiency, and user experience consistency.

[0122] Example 4:

[0123] An electronic device includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute an intelligent control method for a heat pump-based constant temperature and pressure hot water system.

[0124] The processor is an ARM Cortex-A72 quad-core processor with a clock speed of 1.8GHz, and integrates the NEON SIMD instruction set to accelerate neural network inference; the memory includes 4GB LPDDR4 RAM and 16GB eMMC flash memory, where the flash memory stores the pre-trained control model and system firmware, and the RAM is used for runtime data caching and model loading.

[0125] The electronic equipment also includes communication interfaces that support RS485, CAN bus and Wi-Fi 5 for connecting field sensors, actuators and remote monitoring platforms; the input / output interfaces include 4 analog inputs (0-10V / 4-20mA), 2 analog outputs and 8 digital I / Os to meet industrial control requirements.

[0126] In some embodiments, the computer program is tangibly contained in eMMC flash memory, loaded into RAM at startup, and executed by the processor to implement the intelligent control method of the present invention.

[0127] The system described in this paper can be implemented in computing systems including back-end servers, edge computing gateways, or embedded controllers, and can exchange data with other devices via a local area network or the Internet, supporting remote diagnostics and model updates.

[0128] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A smart control method for constant temperature and pressure hot water based on a heat pump, characterized in that, include: S1. Receive real-time multi-sensor data and preprocess it; the multi-sensor data includes: pressure, temperature, flow rate and performance degradation data; S2. Based on multi-sensor data, construct a state observation space that integrates equipment performance degradation indicators and instantaneous operating condition characteristics; S3. Input the state observation space into the adaptive decoupling control model to generate the first control command for the speed of the variable frequency water pump and the second control command for the opening of the mixing valve. S4. Adjust the drive frequency of the variable frequency water pump according to the first control command to change its output flow rate; adjust the hot and cold water mixing ratio of the mixing regulating valve according to the second control command to correct the temperature of the mixed water. S5. Evaluate the residual between the current control effect and the preset target, and combine the performance degradation data to perform closed-loop parameter self-calibration on the adaptive decoupling control model.

2. The intelligent control method for constant temperature and pressure hot water based on a heat pump according to claim 1, characterized in that: Step S1 includes: The acquired performance degradation data includes the abnormal temperature rise trend of the compressor operating current, the correlation shift between the speed and torque of the variable frequency water pump motor at the same frequency, and the thermal balance residual between the inlet and outlet temperature difference of the heat exchanger in the insulated water tank and the input power of the heat pump. Synchronize and align all sensor signals in time; The original signal is nonlinearly smoothed using a sliding window midpoint filtering technique, and outliers are identified and removed using preset statistical criteria. The cleaned data is standardized and mapped to generate dimensionless feature operators.

3. The intelligent control method for constant temperature and pressure hot water based on a heat pump according to claim 1, characterized in that: The state observation space includes the following dimensions: water pressure deviation, mixed water temperature deviation, instantaneous flow rate change rate, compressor energy efficiency ratio trend, water pump efficiency decay index, and heat exchanger fouling index; wherein, the water pressure deviation is defined as the residual between the preset target pressure value and the current measured pressure value; the mixed water temperature deviation is defined as the residual between the preset target temperature value and the current measured temperature value; and the instantaneous flow rate change rate is obtained by performing a first-order difference calculation on the flow meter reading.

4. The intelligent control method for constant temperature and pressure hot water based on a heat pump according to claim 1, characterized in that: The adaptive decoupling control model adopts a deep neural network architecture, whose input layer receives a seven-dimensional state observation vector including the performance degradation index; the deep neural network includes multiple fully connected layers, and a multi-head attention mechanism module is introduced in the middle layer to dynamically calculate and weight the influence of each state dimension on the pressure response and temperature regulation.

5. The intelligent control method for constant temperature and pressure hot water based on a heat pump according to claim 1, characterized in that: The process of generating the first control command and the second control command includes: the output layer of the deep neural network is processed through two parallel command mapping branches; wherein, the pressure control branch combines the instantaneous flow rate change rate to map the target speed of the variable frequency water pump in real time to generate the first control command; the temperature control branch combines the heat exchanger fouling index and energy efficiency trend to map the target opening degree of the mixing valve under the premise of compensating for the large inertial lag of heat transfer, and generates the second control command.

6. The intelligent control method for constant temperature and pressure hot water based on a heat pump according to claim 1, characterized in that: Self-calibration of closed-loop parameters for the adaptive decoupling control model includes: Calculate the pressure control residual and temperature control residual within the current time window; When the residual continuously exceeds a preset threshold, the model fine-tuning procedure is triggered; Using an online learning algorithm, combined with the performance degradation data and operating condition residuals, the parameters of the adaptive decoupling control model are locally corrected; The objective function for model optimization simultaneously considers control error, total system energy consumption, and the magnitude of changes in control commands.

7. The intelligent control method for constant temperature and pressure hot water based on a heat pump according to claim 6, characterized in that: The closed-loop parameter self-calibration employs a gradient increment learning strategy based on a time window, specifically including: Using a fixed duration as a statistical window, calculate the root mean square value of the control error; If the root mean square error of multiple consecutive statistical windows exceeds the threshold, fine-tuning is initiated. The fine-tuning employs a mini-batch gradient descent algorithm and introduces a weight decay term, updating only the relevant parameters of the output layer and attention module in the adaptive decoupling control model.

8. A constant temperature and pressure hot water system based on a heat pump, comprising the intelligent control method for constant temperature and pressure hot water based on a heat pump as described in any one of claims 1-7, characterized in that, include: The data fusion module is used to receive and synchronously process multi-source sensor signals from pressure sensors, temperature sensors, flow meters, and performance degradation monitoring units to generate standardized state observation vectors. The adaptive decoupling control module is used to input the state observation vector into a pre-trained bivariate collaborative optimization network and output pump speed command and mixing valve opening command. The drive module is used to convert the water pump speed command into the frequency control signal of the frequency converter, and to convert the mixing valve opening command into the position control signal of the electric actuator. The performance degradation assessment module is used to calculate the compressor energy efficiency trend, water pump efficiency decay index and heat exchanger fouling index in real time based on equipment operating parameters, and generate an equipment health status report. The online model update module is used to periodically fine-tune the parameters of the bivariate co-optimization network based on the control residuals and equipment health status, so as to maintain the long-term stability of control performance.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the intelligent control method for the heat pump-based constant temperature and pressure hot water system according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the intelligent control method for a heat pump-based constant temperature and pressure hot water system as described in any one of claims 1-7.

Images