Rotation speed prediction control method and system of air source heat pump fan and medium

Abstract

The invention discloses a rotating speed prediction control method and system for an air source heat pump fan and a medium, and belongs to the technical field of heat pump system energy-saving control, and the method comprises the following steps: S1, data collection: collecting historical rotating speed data of the air source heat pump fan and corresponding environmental parameter historical data; s2, an LSTM-Attention prediction model is constructed; s3, establishing a mathematical model of the air source heat pump system, wherein the mathematical model at least comprises a frequency conversion fan model, a circulating water pump model and a surface air cooler model; s4, taking a fan rotating speed prediction value output by the LSTM-Attention prediction model as a set value, and inputting the set value into a model prediction controller; and S5, control execution is carried out, the optimal fan control signal is applied to an actual fan of the air source heat pump, and closed-loop optimization control over the rotating speed is completed. According to the method, through an LSTM-Attention mechanism, the long-term dependency relationship in a multivariable time sequence can be effectively captured, fan rotating speed prediction control is carried out according to a model physical mechanism, the control precision is improved, and the precise energy-saving effect of the fan is achieved.

Description

Technical Field

[0001] This invention belongs to the field of energy-saving control technology for heat pump systems, specifically relating to a predictive control system and method for fan speed optimization of air source heat pump energy consumption based on LSTM-Attention mechanism. Background Technology

[0002] Air source heat pumps, as highly efficient and energy-saving heating / cooling equipment, are increasingly widely used. The fan is a key energy-consuming component in an air source heat pump system, and its speed control has a significant impact on the overall system energy efficiency. Existing air source heat pump fan control methods mostly employ simple PID control or proportional control based on a fixed temperature difference. These methods have several shortcomings: First, the speed setpoint usually cannot comprehensively consider dynamically changing environmental parameters, such as the effects of indoor and outdoor temperature differences, humidity, light intensity, and occupancy rates, resulting in the inability to provide the optimal speed setpoint for energy consumption under partial load or variable operating conditions. Second, existing control strategies generally do not incorporate the complex mathematical model of the air source heat pump itself into the control loop, leading to low control accuracy and difficulty in achieving precise energy consumption optimization.

[0003] While existing patents (such as CN115307349B, which discloses an air source heat pump defrosting control method and an air source heat pump) disclose defrosting control methods for air source heat pumps, their focus is on defrosting logic and actual structural control, without addressing speed optimization based on long-term historical data and multi-feature prediction. Another patent (such as CN115840917A, which discloses a motor temperature rise prediction method) discloses a general method for predicting motor temperature rise using an LSTM-Attention model, but it does not combine this method with the mathematical model of the specific physical system of an air source heat pump to solve the problem of predictive control of fan speed.

[0004] Therefore, there is an urgent need in this field for a technical solution that can accurately predict and optimize the control of the air source heat pump fan speed in order to maximize the overall energy efficiency of the system. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a predictive control system and method for fan speed optimization in air-source heat pumps based on an LSTM-Attention mechanism. This method integrates LSTM (Long Short-Term Memory) networks with the Attention mechanism to construct a system that simultaneously considers the long-term dependencies of multi-dimensional influencing factors such as temperature difference, humidity, light intensity, and occupancy rate, thereby improving the long-term prediction accuracy of fan speed. A precise mathematical model of the air-source heat pump system is established and embedded as an internal constraint into the predictive controller, achieving real-time, forward-looking control with the optimization objective of minimizing system energy consumption. This significantly improves the control accuracy of fan speed and system energy efficiency, thus solving the technical problems mentioned in the background.

[0006] The objective of this invention is achieved as follows: a method for predictive control of the fan speed of an air source heat pump, comprising the following steps: S1, data acquisition, acquiring historical speed data of the air source heat pump fan and corresponding historical environmental parameter data; S2, constructing an LSTM-Attention prediction model, using the historical speed data and historical environmental parameter data as input, and the predicted fan speed value at a future time as output; S3, establishing a mathematical model of the air source heat pump system, the mathematical model including at least a variable frequency fan model, a circulating water pump model, and a surface cooler model, used to describe the dynamic relationship between system energy consumption and fan speed; S4, using the predicted fan speed value output by the LSTM-Attention prediction model as a setpoint, inputting it to a model predictive controller; the model predictive controller, constrained by the mathematical model of the air source heat pump system and with the optimization objective of minimizing system energy consumption, calculates the optimal fan control signal; S5, control execution, applying the optimal fan control signal to the actual fan of the air source heat pump to complete closed-loop optimization control of the speed. The LSTM-Attention prediction model is used to predict the reasonable fan speed required in the future based on historical operating data and environmental data (feedforward). Then, this predicted value is not directly output, but is input as a set value to the model predictive controller (MPC). The MPC then combines a precise mathematical model of the air source heat pump system with the goal of minimizing total energy consumption, and calculates a truly optimal control command under the current system constraints (feedback optimization), thereby achieving precise and energy-saving closed-loop control.

[0007] By employing the LSTM-Attention mechanism, long-term dependencies in multivariate time series are effectively captured, and the Attention mechanism enhances the model's sensitivity to key environmental features (such as temperature difference and occupancy rate), significantly improving the prediction accuracy of fan speed under complex and variable operating conditions. Combining a precise mathematical model of the air-source heat pump system with model predictive control enables forward-looking optimization control based on the system's future state, truly minimizing global energy consumption while ensuring comfort. Because the mathematical model fully considers the physical characteristics and coupling relationships of various components within the system (fan, pump, and surface cooler), the predictive control output based on this model better matches the actual operating characteristics of the system, significantly improving fan speed control accuracy. This approach leverages the forward-looking advantages of data prediction while ensuring that control decisions conform to the inherent laws and constraints of the physical system, significantly reducing system energy consumption and improving control accuracy and stability while meeting comfort requirements.

[0008] Furthermore, the environmental parameters mentioned in step S1 include at least the indoor-outdoor temperature difference, humidity, light intensity, and occupancy rate. The indoor-outdoor temperature difference is the core driving factor of heat load; humidity affects latent heat load and perceived temperature; light intensity affects building heat gain; and occupancy rate directly determines internal heat generation and fresh air demand. These parameters together constitute a complete environmental profile affecting the fan speed decision. In implementation, historical data for these four types of parameters are obtained using temperature sensors, humidity sensors, light intensity sensors, and (e.g.,) access control systems / infrared sensors. By systematically selecting these four heterogeneous features strongly correlated with building heat load and occupancy activity, and using them together as inputs for fan speed prediction, the prediction model can more comprehensively "understand" the causes of environmental changes, rather than just the results. The model can distinguish whether the temperature increase is due to increased sunlight or increased occupancy, thus making a more reasonable speed prediction, avoiding misjudgments and energy waste that may result from inputting a single parameter (such as only temperature), and greatly improving the accuracy and generalization ability of the prediction model.

[0009] Furthermore, in step S2, the LSTM-Attention prediction model construction step involves normalizing the input historical rotation speed data and historical environmental parameter data. Different environmental parameters (such as temperature, humidity, and rotation speed) have vastly different numerical ranges and dimensions. Normalization maps all feature data to a uniform scale (e.g., the [0,1] interval), eliminating the influence of dimensions and preventing certain large-valued features from dominating model training, thereby accelerating model convergence and improving training stability. In practice, methods such as Min-Max normalization are used to perform linear transformations on the data for each feature dimension, ensuring that the LSTM-Attention model can be effectively and reliably trained in this application scenario. This improves the training efficiency and final performance of the LSTM-Attention prediction model, which is a crucial technical guarantee for the successful implementation of the entire invention.

[0010] Furthermore, the mathematical model of the air source heat pump system in step S3 also includes a variable air volume (VAV) terminal control system model, used to control the air volume by adjusting the fan speed and damper opening. Adding the VAV terminal model to the mathematical model expands the system model from a single fan, water circulation, and surface cooler to a complete air and water system. This model describes the dynamic process of controlling the air volume in each area by adjusting the fan speed (total air volume) and the opening of each terminal damper (air volume distribution). In the heat pump system mathematical model within the MPC, equations describing the coupling relationship between the air network resistance characteristics, VAV terminal damper characteristics, and fan performance curves are added. Integrating the control characteristics of the VAV terminal into the global optimization model means that the MPC's optimization calculation no longer considers the fan itself in isolation, but rather the collaborative work of the fan and the entire air system. The optimization objective is upgraded from "lowest fan energy consumption" to "lowest overall air system energy consumption." It achieves more macroscopic and precise system-level energy consumption optimization, avoiding deviations in optimization results caused by not considering the impact of terminal air valve adjustment on the air network resistance, making the control of fan speed more in line with the actual system operating conditions, and the energy-saving effect more significant.

[0011] A predictive control system for the fan speed of an air source heat pump includes: a data acquisition module for real-time acquisition of fan speed and environmental parameters; an LSTM-Attention prediction module connected to the data acquisition module for receiving historical data and outputting predicted fan speed values; a model predictive control module connected to the LSTM-Attention prediction module for receiving the predicted speed values; and an air source heat pump mathematical model module embedded in the model predictive control module as the basis model for its internal prediction and optimization. The model predictive control module, based on the air source heat pump mathematical model module, is used to solve for the optimal fan control signal and output it to the fan actuator of the air source heat pump. In implementation, these modules can be integrated into an embedded controller (such as a PLC or dedicated industrial computer) as software programs, and connected to field sensors and actuators through its I / O interface.

[0012] Furthermore, the LSTM-Attention prediction module comprises an input layer, at least two LSTM hidden layers, an Attention layer, and a fully connected output layer connected in sequence. The input layer receives multi-dimensional time-series data; the at least two LSTM hidden layers are used to extract deep features and complex dependencies of the data in the time dimension layer by layer; the Attention layer automatically learns and assigns different weights to historical data of different times and features to highlight key information; the fully connected output layer maps the weighted feature vector to the final predicted wind turbine speed. In implementation, a neural network with this specific structure can be built using frameworks such as TensorFlow and PyTorch, and trained using historical data. For the specific task of wind turbine speed prediction, the specific network architecture of "input-LSTM×2-Attention-output" can more effectively capture the long-term, nonlinear, and multi-factor coupling relationships affecting wind turbine speed, thus achieving higher prediction accuracy than a single LSTM layer or a simple network structure.

[0013] Furthermore, the variable frequency fan model in the air source heat pump mathematical model module is used to calculate the output power and air mass flow rate of the fan based on the fan's control signal, and to calculate the outlet air enthalpy based on the output power, air mass flow rate, inlet air enthalpy, fan motor efficiency, and heat loss rate. The surface cooler model in the air source heat pump mathematical model module is used to calculate the heat exchange between air and chilled water, the enthalpy of the outlet air, and the absolute humidity ratio based on inlet air state parameters, chilled water temperature, air mass flow rate, and air bypass coefficient; and to determine and calculate the condensate mass flow rate based on condensation on the surface cooler. In implementation, these precise calculation logics are implemented in the MPC's internal code based on thermodynamic and fluid dynamics formulas.

[0014] An air source heat pump device includes: a controller, the controller being configured to predict and control the rotational speed of a variable frequency fan according to the method described above.

[0015] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0016] The beneficial effects of this invention are as follows: By employing the LSTM-Attention mechanism, long-term dependencies in multivariate time series are effectively captured, and the Attention mechanism enhances the model's sensitivity to key environmental features (such as temperature difference and occupancy rate), significantly improving the prediction accuracy of fan speed under complex and variable operating conditions. Combining a precise mathematical model of the air-source heat pump system with model predictive control enables forward-looking optimization control based on the system's future state, truly minimizing global energy consumption while ensuring comfort. Because the mathematical model fully considers the physical characteristics and coupling relationships of various components within the system (fan, pump, and surface cooler), the predictive control output based on this model better matches the actual operating characteristics of the system, significantly improving the fan speed control accuracy. This invention leverages the forward-looking advantages of data prediction while ensuring that control decisions conform to the inherent laws and constraints of the physical system, significantly reducing system energy consumption and improving control accuracy and stability while meeting comfort requirements. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system configuration in this invention;

[0018] Figure 2 This is a schematic diagram of the method flow of the present invention;

[0019] Figure 3 This is a time distribution diagram of indoor thermal disturbance parameters considered in the air source heat pump model of this invention;

[0020] Figure 4 This is the case of indoor and outdoor temperature difference considered in the air source heat pump model of this invention;

[0021] Figure 5 This is the control effect achieved by the system and method in this invention;

[0022] Among them, 1 is the air source heat pump model, 2 is the predictive control model, and 3 is the prediction model under the LSTM-Attention mechanism. Detailed Implementation

[0023] The present invention will now be described in further detail with reference to the accompanying drawings. It should be noted that this is only for the purpose of more clearly illustrating and explaining the present invention.

[0024] like Figure 1 and 2As shown in the figure, this embodiment discloses a speed prediction control method for an air source heat pump fan, including the following steps: S1, data acquisition: acquiring historical speed data of the air source heat pump fan and corresponding historical environmental parameter data; S2, constructing an LSTM-Attention prediction model: using the historical speed data and historical environmental parameter data as input, and the predicted fan speed value at a future time as output; S3, establishing a mathematical model of the air source heat pump system: the mathematical model includes at least a variable frequency fan model, a circulating water pump model, and a surface cooler model, used to describe the dynamic relationship between system energy consumption and fan speed; S4, using the predicted fan speed value output by the LSTM-Attention prediction model as a setpoint, inputting it to the model prediction controller; the model prediction controller uses the mathematical model of the air source heat pump system as a constraint and the minimization of system energy consumption as the optimization objective to calculate the optimal fan control signal; S5, control execution: applying the optimal fan control signal to the actual fan of the air source heat pump to complete the closed-loop optimization control of the speed. The LSTM-Attention prediction model is used to predict the reasonable fan speed required in the future based on historical operating data and environmental data (feedforward). Then, this predicted value is not directly output, but is input as a set value to the model predictive controller (MPC). The MPC then combines a precise mathematical model of the air source heat pump system with the goal of minimizing total energy consumption, and calculates a truly optimal control command under the current system constraints (feedback optimization), thereby achieving precise and energy-saving closed-loop control.

[0025] The mathematical models for the variable frequency fan, circulating water pump, and surface cooler are as follows:

[0026] 1) Variable frequency fan model

[0027] Variable air volume (VAV) terminals in central air conditioning systems utilize variable frequency (VFD) fans to adjust the system's airflow, thereby regulating indoor temperature. The formulas for calculating fan output power, air mass flow rate through the fan, and outlet air enthalpy are as follows:

[0028] P fan =P rated (a0+a1γ+a2γ 2 +a3γ 3 +a4γ 4 +...) (1)

[0029] m air =m rated γ=3.6v rated ρ air (2)

[0030]

[0031] In the formula, γ is the control signal of the fan; P rated The rated power of the fan is kJ / h; η motor，f f loss，f For fan motor efficiency and heat loss rate; m air The mass flow rate of air passing through the fan is expressed in kg / h and m³. rated The maximum air mass flow rate through the fan is expressed in kg / h; h. air，in h air，out Enthalpy of the air at the inlet and outlet of the fan, in kJ / kg.

[0032] 2) Circulating water pump model

[0033] Both the load side and ground source side water circulation systems of the central air conditioning system use fixed-frequency pumps. Given the inlet water temperature and flow rate of the circulating water pumps, the outlet water temperature, efficiency, shaft power, and heat transferred from the pump to the circulating fluid can be calculated using the following formulas:

[0034]

[0035] Q f，p =P shaft (1-η pump )+(P rated -P shaft )f loss，p (7)

[0036] In the formula, η pump η motor，p η overall These are the pump's effective efficiency, the motor's efficiency, and the overall efficiency, respectively; m f，p Q is the flow rate of the circulating medium, kg / h; f，p Heat gained by the fluid, kJ / h; P rated P shaft These are the rated power and shaft power of the circulating pump, respectively, in kJ / h; T out，p T in,p Here, Q represents the inlet and outlet fluid temperatures of the pump, in °C; H represents the pump head, in m; and Q represents the outlet fluid temperature. f For fluid flow rate, m 3 / h;f loss,p C represents the heat loss rate of the water pump. p is the specific heat capacity of the fluid, kJ / (kg·K).

[0037] 3) Surface cooler model

[0038] The surface cooler is located within the air duct system of the air conditioning system and uses chilled water or other refrigerants to cool the air delivered through the duct. The relationship between the changes in inlet and outlet chilled water temperature, condensate mass flow rate, and the enthalpy change and absolute humidity ratio of the mixed air is as follows:

[0039]

[0040] m coil =m air (1-f bypass (w) air，out -w air，in (9)

[0041] h ai，mix =(1-f bypass )h air，out +f bypass h air，in (10)

[0042] w air，mix =1-f byypass )w air，out +f bypass w air，in (11)

[0043] Q f，s =m air (1-f bypass (h) air，in -h air，out )-m coil h coil (12)

[0044] In the formula, f bypass Q is the air bypass coefficient, with a value in the range [0, 1]. f，s Heat transferred to water by air exchange, kJ / h; h coil T represents the enthalpy of the condensate at the outlet of the surface cooler, expressed in kJ / kg. in，s T out，s The inlet and outlet chilled water temperatures of the surface cooler are in °C; m f，s m coil m air The mass flow rates of chilled water, condensate, and air at the inlet of the surface cooler are kg / h; C fluid Specific heat capacity of chilled water, kJ / kg·K; h air，in h air，out h air,mix The enthalpy of the air entering, leaving, and mixing with the surface cooler, expressed in kJ / kg; w air，in w air，out w air，mix The absolute humidity ratio of the inflowing air, the outflowing air, and the mixed air is expressed as kgH2O / kgAir.

[0045] This embodiment utilizes the LSTM-Attention mechanism to effectively capture long-term dependencies in multivariate time series data. The Attention mechanism enhances the model's sensitivity to key environmental features (such as temperature difference and occupancy rate), significantly improving the accuracy of fan speed prediction under complex and variable operating conditions. By combining a precise mathematical model of the air-source heat pump system with model predictive control, it achieves forward-looking optimization control based on the system's future state, truly minimizing global energy consumption while ensuring comfort. Because the mathematical model fully considers the physical characteristics and coupling relationships of various components within the system (fan, pump, and surface cooler), the predictive control output based on this model better matches the actual operating characteristics of the system, significantly improving fan speed control accuracy. It leverages the forward-looking advantages of data prediction while ensuring that control decisions conform to the inherent laws and constraints of the physical system, significantly reducing system energy consumption and improving control accuracy and stability while meeting comfort requirements.

[0046] The environmental parameters mentioned in step S1 include at least the indoor-outdoor temperature difference, humidity, light intensity, and occupancy rate. The indoor-outdoor temperature difference is the core driving factor of heat load; humidity affects latent heat load and perceived temperature; light intensity affects building heat gain; and occupancy rate directly determines internal heat generation and fresh air demand. These parameters together constitute a complete environmental profile affecting fan speed decisions. In implementation, historical data for these four types of parameters are acquired using temperature sensors, humidity sensors, light intensity sensors, and (e.g.,) access control systems / infrared sensors. By systematically selecting these four heterogeneous features strongly correlated with building heat load and occupancy activity, and using them together as inputs for fan speed prediction, the prediction model can more comprehensively "understand" the causes of environmental changes, not just the results. The model can distinguish whether the temperature increase is due to increased sunlight or increased occupancy, thus making a more reasonable speed prediction. This avoids misjudgments and energy waste that may result from inputting a single parameter (e.g., only temperature), greatly improving the accuracy and generalization ability of the prediction model.

[0047] In step S2, the LSTM-Attention prediction model construction process involves normalizing the input historical rotation speed data and historical environmental parameter data. Different environmental parameters (such as temperature, humidity, and rotation speed) have vastly different numerical ranges and dimensions. Normalization maps all feature data to a uniform scale (e.g., the [0,1] interval), eliminating the influence of dimensions and preventing certain features with large values ​​from dominating model training, thereby accelerating model convergence and improving training stability. In practice, methods such as Min-Max normalization are used to perform linear transformations on the data for each feature dimension, ensuring that the LSTM-Attention model can be effectively and reliably trained in this application scenario. This improves the training efficiency and final performance of the LSTM-Attention prediction model, which is a crucial technical guarantee for the successful implementation of the entire invention.

[0048] The mathematical model of the air source heat pump system in step S3 also includes a variable air volume (VAV) terminal control system model, used to control the air volume by adjusting the fan speed and damper opening. Adding the VAV terminal model to the mathematical model expands the system model from a single fan, water circulation, and surface cooler to a complete air and water system. This model describes the dynamic process of controlling the air volume in each area by adjusting the fan speed (total air volume) and the opening of each terminal damper (air volume distribution). In the heat pump system mathematical model within the MPC, equations describing the coupling relationship between the air network resistance characteristics, VAV terminal damper characteristics, and fan performance curves are added. Integrating the control characteristics of the VAV terminal into the global optimization model means that the MPC's optimization calculation no longer considers the fan itself in isolation, but rather the coordinated work of the fan and the entire air system. The optimization objective is upgraded from "lowest fan energy consumption" to "lowest overall air system energy consumption." It achieves more macroscopic and precise system-level energy consumption optimization, avoiding deviations in optimization results caused by not considering the impact of terminal air valve adjustment on the air network resistance, making the control of fan speed more in line with the actual system operating conditions, and the energy-saving effect more significant.

[0049] Variable air volume (VAV) air conditioning systems control the air volume by adjusting the fan speed and the opening of the air valves in the terminal devices, thereby achieving the purpose of indoor temperature regulation.

[0050] The air volume of the variable air volume terminal of the fan system is calculated as follows:

[0051]

[0052] In the formula, L is the air supply volume, m 3 / h;Q s Q q Sensible heat load and total heat load, respectively, in kW and t. in t sThese are the room temperature and the supply air temperature, respectively, in °C; ρ a air density, kg / m³ 3 c is the specific heat capacity of air at constant pressure, kJ / (kg·℃); h in h out The enthalpy values ​​are the indoor air state point enthalpy and the supply air state point enthalpy, in kJ / kg.

[0053] A speed prediction and control system for an air source heat pump fan, comprising:

[0054] The data acquisition module is used to collect fan speed and environmental parameters in real time.

[0055] The LSTM-Attention prediction module is connected to the data acquisition module and is used to receive historical data and output the predicted value of the wind turbine speed.

[0056] The model prediction control module is connected to the LSTM-Attention prediction module and is used to receive the predicted rotational speed value;

[0057] An air source heat pump mathematical model module is embedded in the model predictive control module as the basic model for its internal prediction and optimization. The model predictive control module, based on the air source heat pump mathematical model module, is used to solve for the optimal fan control signal and output it to the fan actuator of the air source heat pump.

[0058] In practice, the above modules can be integrated into an embedded controller (such as a PLC or a dedicated industrial computer) as software programs, and connected to field sensors and actuators through its I / O interface.

[0059] The air source heat pump model, the predictive control model, and the predictive model under the LSTM-Attention mechanism provide a reference value for the fan speed in the air source heat pump model. The predictive control model is used to control the fan speed in the air source heat pump model.

[0060] The air source heat pump model includes a 1A variable frequency fan, a 1B air duct, a 1C surface cooler, a 1E cooling tower, a 1F circulating water pump, a 1G bypass valve, a 1H compressor, a 1I evaporator, a 1J condenser, and a 1K expansion valve. The 1H compressor compresses the low-temperature, low-pressure gaseous refrigerant into a high-temperature, high-pressure state, and delivers it to the 1D cooling tower connected to the 1J condenser to release heat. Subsequently, the refrigerant is depressurized and cooled by a throttling device, enters the 1I evaporator to absorb heat from the outside air and vaporizes, and finally returns to the 1H compressor to complete the cycle. The 1F circulating water pump circulates water through the 1C surface cooler and the 1J condenser, while the 1A variable frequency fan drives air through the 1B air duct to achieve heat transfer.

[0061] The LSTM-Attention prediction module comprises an input layer, at least two LSTM hidden layers, an attention layer, and a fully connected output layer, connected sequentially. The input layer receives multi-dimensional time-series data; the at least two LSTM hidden layers extract deep features and complex dependencies in the time dimension layer by layer; the attention layer automatically learns and assigns different weights to historical data of different times and features, highlighting key information; the fully connected output layer maps the weighted feature vector to the final predicted wind turbine speed. In implementation, a neural network with this specific structure can be built using frameworks such as TensorFlow and PyTorch, and trained using historical data. For the specific task of wind turbine speed prediction, this specific network architecture of "input-LSTM×2-Attention-output" can more effectively capture the long-term, nonlinear, and multi-factor coupling relationships affecting wind turbine speed, thus achieving higher prediction accuracy than a single LSTM layer or a simple network structure.

[0062] The variable frequency fan model in the air source heat pump mathematical model module is used to calculate the output power and air mass flow rate of the fan based on the fan's control signal, and to calculate the outlet air enthalpy based on the output power, air mass flow rate, inlet air enthalpy, fan motor efficiency, and heat loss rate. The surface cooler model in the same module is used to calculate the heat exchange between air and chilled water, the enthalpy of the outlet air, and the absolute humidity ratio based on inlet air state parameters, chilled water temperature, air mass flow rate, and air bypass coefficient; and to determine and calculate the condensate mass flow rate based on condensation on the surface cooler. In implementation, these precise calculations are carried out in the MPC's internal code using thermodynamic and fluid dynamics formulas.

[0063] An air source heat pump device includes: a controller, the controller being configured to predict and control the rotational speed of a variable frequency fan according to the method described above.

[0064] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0065] In this embodiment, step 1 involves collecting historical fan speed data under the influence of environmental factors such as different indoor and outdoor temperature differences, humidity, light intensity, and occupancy rates within a physical space.

[0066] Step 2: Use the data from the previous step as the input layer to establish a wind turbine speed prediction model under the LSTM-Attention mechanism.

[0067] Step 3: Establish a wind turbine speed prediction model.

[0068] Step 4: Establish a predictive control model for an air source heat pump based on the LSTM-Attention mechanism. The mathematical model of the air source heat pump collects the fan speed of the air source heat pump.

[0069] Step 5: Combine the data collected in Step 4 with the predicted speed value under the LSTM-Attention mechanism in Step 2 to complete the predictive control of the wind turbine speed.

[0070] 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 concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method and system for speed prediction control of an air source heat pump fan, and a medium, characterized in that, Includes the following steps: S1. Data acquisition: Collect historical speed data of the air source heat pump fan, as well as corresponding historical environmental parameter data; S2. Construct an LSTM-Attention prediction model, using the historical rotational speed data and historical environmental parameter data as inputs, and the predicted wind turbine rotational speed at future moments as outputs; S3. Establish a mathematical model of the air source heat pump system. The mathematical model includes at least a variable frequency fan model, a circulating water pump model, and a surface cooler model to describe the dynamic relationship between system energy consumption and fan speed. S4. The predicted fan speed output by the LSTM-Attention prediction model is used as a set value and input to the model prediction controller. The model prediction controller uses the mathematical model of the air source heat pump system as a constraint and the minimization of system energy consumption as the optimization objective to calculate the optimal fan control signal. S5. Control execution: Apply the optimal fan control signal to the actual fan of the air source heat pump to complete the closed-loop optimization control of the speed.

2. The rotational speed prediction control method of an air source heat pump blower according to claim 1, characterized by, The environmental parameters mentioned in step S1 include at least the indoor-outdoor temperature difference, humidity, light intensity, and the number of people in the room.

3. The speed prediction and control method, system, and medium for an air source heat pump fan according to claim 1, characterized in that, In step S2, the LSTM-Attention prediction model is constructed by normalizing the input historical rotational speed data and historical environmental parameter data.

4. The speed prediction and control method, system, and medium for an air source heat pump fan according to claim 1, characterized in that, The mathematical model of the air source heat pump system in step S3 also includes a variable air volume terminal control system model, which is used to control the air supply volume by adjusting the fan speed and the opening of the air valve.

5. A speed prediction and control system for an air source heat pump fan, characterized in that, include: The data acquisition module is used to collect fan speed and environmental parameters in real time. The LSTM-Attention prediction module is connected to the data acquisition module and is used to receive historical data and output the predicted value of the wind turbine speed. The model prediction control module is connected to the LSTM-Attention prediction module and is used to receive the predicted rotational speed value; An air source heat pump mathematical model module is embedded in the model prediction and control module as the basic model for its internal prediction and optimization. The model prediction control module is based on the air source heat pump mathematical model module, which is used to solve for the optimal fan control signal and output it to the fan actuator of the air source heat pump.

6. The speed prediction and control system for an air source heat pump fan according to claim 5, characterized in that, The LSTM-Attention prediction module includes an input layer, at least two LSTM hidden layers, an attention layer, and a fully connected output layer connected in sequence.

7. The speed prediction and control system for an air source heat pump fan according to claim 5, characterized in that, The variable frequency fan model in the mathematical model module of the air source heat pump is used to calculate the output power and the mass flow rate of the air passing through it according to the control signal of the fan, and to calculate the outlet air enthalpy based on the output power, mass flow rate of the air, inlet air enthalpy, fan motor efficiency and heat loss rate.

8. The speed prediction and control system for an air source heat pump fan according to claim 5, characterized in that, The surface cooler model in the air source heat pump mathematical model module is used to calculate the heat exchange between air and chilled water, the enthalpy of the outlet air and the absolute humidity ratio based on the inlet air state parameters, chilled water temperature, air mass flow rate and air bypass coefficient; and to determine and calculate the condensate mass flow rate based on the condensation on the surface cooler.

9. An air source heat pump device, characterized in that, include: A controller for predicting and controlling the rotational speed of a variable frequency fan according to the method of any one of claims 1-4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-4.

Images