Multi-source heat pump coupling system control method and system based on double-layer optimization

Abstract

The invention discloses a multi-source heat pump coupling system control method and system based on double-layer optimization. The method comprises the steps that refrigerating and heating working condition data of all heat pumps are taken; the multi-source heat pump comprises a ground source heat pump and / or an air source heat pump; upper-layer optimization is executed, the target load of each heat pump serves as a decision variable, and the target load of each heat pump in the first time scale is obtained through a first multi-target optimization algorithm; s2, performing lower-layer optimization, and determining a control instruction of each heat pump in a second time scale through a second multi-target optimization algorithm by taking the real-time operation state data of each heat pump as a decision variable so as to match the target load in the step S2; and each heat pump executes the control instruction. According to the method, decision making is carried out under different time scales through the double-layer optimization architecture, so that the problem that long-term decision making and short-term control are mixed in the same optimization problem, so that the calculation complexity is too high or the control effect is not stable is solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for building heating, ventilation and air conditioning systems, and specifically to a control method and system for a multi-source heat pump coupling system based on dual-layer optimization. Background Technology

[0002] As building energy consumption accounts for an increasingly larger share of total societal energy consumption, energy-saving optimization of HVAC systems has become crucial for building energy conservation. Ground source heat pump systems utilize the stable temperature of underground soil for heat exchange, offering high efficiency and energy savings, but suffer from high initial investment and soil heat accumulation issues. Air source heat pump systems, on the other hand, have lower investment costs and more flexible installation, but their performance is significantly affected by outdoor temperatures. Coupling these two systems can leverage their respective advantages and achieve complementary benefits. However, existing technologies for controlling ground source heat pump coupled with air source heat pump systems have the following problems:

[0003] 1. Inappropriate load sharing strategy Traditional control methods allocate the load between ground source heat pumps and air source heat pumps using fixed priorities or simple rules, such as a strategy of "prioritizing ground source heat pumps and starting air source heat pumps when insufficient." This approach fails to comprehensively consider multiple optimization objectives, including total power consumption, operating costs (time-of-use pricing), and carbon emissions, resulting in low system efficiency and poor economic performance. For example, during peak electricity price periods, even if the ground source heat pump has a high energy efficiency ratio, it may not be as economically viable as optimizing the coordinated operation of multiple heat sources through load allocation to reduce operating costs.

[0004] 2. Neglecting the soil heat accumulation effect Long-term operation of ground source heat pump systems can lead to a cumulative increase (during the cooling season) or decrease (during the heating season) in soil temperature, affecting system performance. Existing MPC (Model Predictive Control) methods typically assume a constant soil temperature or use oversimplified linear models, failing to accurately predict dynamic changes in soil temperature and long-term heat accumulation effects. This results in severe performance degradation of ground source heat pumps after long-term operation, and may even lead to uncontrolled soil temperature and system malfunction.

[0005] 3. Single-layer optimization struggles to accommodate multiple time scales. The optimization of building air conditioning systems involves multiple time scales: long-term strategic decisions (such as load allocation and soil thermal balance management, with time scales ranging from hours to days) and short-term tactical controls (such as equipment start-up and shutdown, and temperature and flow regulation, with time scales ranging from minutes). Mixing these decisions with different time scales in a single-layer optimization will lead to an excessively high dimensionality of the optimization problem (curse of dimensionality), an exponential increase in computational complexity, difficulty in solving within a finite time, and impact on the real-time performance of control.

[0006] 4. Lack of practical engineering implementation plans Existing research largely remains at the theoretical simulation stage, lacking system architecture design for practical engineering applications. It fails to consider engineering issues such as hardware deployment, communication latency, fault tolerance mechanisms, and device interfaces in real-world systems, making it difficult to apply and promote theoretical methods in real-world buildings.

[0007] Therefore, there is an urgent need for a control method for a ground source heat pump coupled with an air source heat pump system that can comprehensively consider multi-objective optimization, soil thermal balance, multi-timescale coordination, and has engineering feasibility. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a control method and system for a multi-source heat pump coupling system based on dual-layer optimization. It aims to address the common problems in existing ground source heat pump and air source heat pump coupling systems, such as the single load distribution strategy, difficulty in balancing energy consumption, operating costs and long-term system stability, and the difficulty of achieving coordinated optimization at different time scales using traditional control methods.

[0009] The technical solution adopted by this invention to solve its technical problem is: This invention provides a control method for a multi-source heat pump coupling system based on two-layer optimization, comprising the following steps: S1: Obtain cooling and heating operating data for each heat pump; the multi-source heat pump includes ground source heat pumps and / or air source heat pumps; S2: Perform upper-level optimization, where the target load of each heat pump is used as the decision variable, and the target load of each heat pump in the first time scale is obtained through the first multi-objective optimization algorithm; S3: Perform lower-level optimization, where the real-time operating status data of each heat pump is used as the decision variable, and the control command of each heat pump in the second time scale is determined by the second multi-objective optimization algorithm to match the target load of step S2; S4: Each heat pump executes the control command; Wherein, the first time scale is larger than the second time scale.

[0010] Preferably, in step S2, the objective function of the first multi-objective optimization algorithm is one or more of the following optimization objectives: total power consumption, operating cost, and carbon emissions during operation.

[0011] Preferably, in step S2, the constraints of the first multi-objective optimization algorithm are one or more of the following: load balance constraint, equipment capacity constraint, soil thermal balance constraint, and equipment start-stop frequency constraint, wherein the load balance constraint is that the total cooling or heating provided by the heat pump in the system should meet the building load demand within any time step; the equipment capacity constraint is that the output capacity of the heat pump meets its rated capacity limit; the soil thermal balance constraint is to prevent soil heat accumulation; and the equipment start-stop frequency constraint is to prevent frequent start-stop.

[0012] Preferably, the building load in the load balance constraint is: based on historical building load data and historical and / or predicted meteorological data, the building cooling and heating load at each time step in the future preset time domain is output using a building load prediction model.

[0013] Preferably, in step S3, the decision variables of the second multi-objective optimization algorithm include one or more of the following: heat pump temperature setpoint, heat pump flow rate setpoint, water pump speed ratio, and operating mode.

[0014] Preferably, in step S3, the objective function of the second multi-objective optimization algorithm is one or more of the following optimization objectives: minimizing the tracking error of the actual output load of the heat pump to the target load of step S2; minimizing the total power or total energy consumption of the system.

[0015] Preferably, in step S3, the constraints of the second multi-objective optimization algorithm are one or more of the following: temperature constraints, flow rate constraints, pump speed constraints, and equipment start-up and shutdown frequency constraints, wherein the temperature constraints include cooling season / heating season temperature constraints and ground source side return water temperature constraints.

[0016] Preferably, the first multi-objective optimization algorithm is the NSGA-II multi-objective genetic algorithm, and the second multi-objective optimization algorithm is the model predictive control (MPC) framework. In each optimization cycle, the system operating state in the future multiple time steps is predicted based on the system mathematical model, and the control variables are solved using the multi-objective particle swarm optimization algorithm (MOPSO).

[0017] The present invention also provides a control system for a multi-source heat pump coupling system based on two-layer optimization, for executing the above-mentioned control method for a multi-source heat pump coupling system based on two-layer optimization, comprising: The data acquisition module is configured to execute step S1; The upper-level optimization module is configured to execute step S2; The lower-level optimization module is configured to execute step S3; The control execution module is configured to execute step S4.

[0018] The present invention has the following beneficial effects: This invention divides the control problem of a multi-source heat pump coupled system into two levels: upper-level load distribution optimization and lower-level real-time control optimization. Decisions are made at different time scales, thus avoiding the problems of excessive computational complexity or unstable control performance caused by mixing long-term decisions and short-term control in the same optimization problem. Through this two-level optimization architecture, the upper-level optimization focuses on the overall operational strategy, while the lower-level optimization focuses on fine-grained control at the execution level, making the system operation more coordinated and orderly. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 Here is a diagram of the LSTM structure; Figure 2 A schematic diagram of the Dung Beetle Optimization Algorithm (DBO); Figure 3 This is a structural diagram of the DBO-LSTM-Attention building load prediction model. Figure 4 Flowchart of the two-layer optimized control framework; Figure 5 This is a diagram of the overall system framework. Figure 6 This is a diagram of the system's communication network topology. Figure 7 A comparison of the annual cumulative power consumption of each device under different strategies; Figure 8 This section compares the total operating costs under different strategies. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. 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.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] The following description, in conjunction with the accompanying drawings, details the specific scheme of a two-layer optimized control method for a multi-source heat pump coupling system provided by the present invention.

[0023] I. System Overall Structure and Two-Level Optimization Control Framework This invention proposes a two-layer optimization control method for multi-source heat pump coupling systems based on NSGA-II and MPC combined with MOPSO, applicable to building cooling and heating systems including ground source heat pumps and air source heat pumps. This system serves the building load and achieves comprehensive optimization of the building HVAC system in terms of energy consumption, operating costs, carbon emissions, and long-term operational stability by coordinating the operation modes of different heat source equipment.

[0024] like Figure 4As shown, this invention employs a two-layer optimization control framework, dividing system control into two levels—upper-layer optimization and lower-layer optimization—in terms of time scale and control function. The upper-layer optimization operates on a longer time scale and is primarily responsible for multi-objective load allocation based on load forecasting, time-of-use electricity pricing, and soil thermal conditions. The lower-layer optimization operates on a shorter time scale and is primarily responsible for real-time optimization of control variables such as heat pump supply water temperature, supply flow rate, and pump speed, while meeting the upper-layer target load.

[0025] At the engineering implementation level, such as Figure 5 As shown, this invention further constructs an edge-cloud collaborative control architecture consisting of a cloud layer, an edge layer, and a device layer. The cloud layer is responsible for computationally intensive prediction and upper-layer optimization tasks, the edge layer is responsible for lower-layer model prediction and control tasks with high real-time requirements, and the device layer is used to execute specific control commands and collect operating status data, thereby achieving a balance between control performance and engineering feasibility.

[0026] II. Construction of System Mathematical Model To achieve dual-layer optimization control, this invention first establishes a complete mathematical model of a multi-source heat pump coupling system. This model includes a ground source heat pump performance model, an air source heat pump performance model, a variable frequency water pump model, a buried pipe heat exchange model, a building load prediction model, and system operation constraints.

[0027] 2.1 Performance Model of Ground Source Heat Pump The ground source heat pump is modeled using the polynomial correction coefficient method, and its performance parameters change dynamically with operating conditions.

[0028] (1) Refrigeration condition: Formula for calculating the cooling capacity of a ground source heat pump: , Formula for calculating the cooling power consumption of a ground source heat pump: , Formula for calculating the coefficient of performance (COP) of a ground source heat pump: , in: The cooling capacity of the ground source heat pump unit is expressed in kW. The power consumption of the ground source heat pump unit is kW; The cooling capacity of the unit under rated operating conditions is expressed in kW. This is a correction factor for the cooling capacity of the heat pump unit under actual operating conditions; This refers to the partial load rate of the heat pump unit. The input power of the heat pump unit under rated cooling conditions is expressed in kW. This is the input power correction factor for the heat pump unit under full-load operation. This is the input power correction factor for the heat pump unit under partial load operation conditions.

[0029] Calculation of cooling capacity correction factor: , Calculation of full-load input power correction factor: , Calculation of partial load input power correction factor: , in: This is the ratio of the actual water flow rate on the evaporator side to the rated water flow rate. This is the ratio of the actual water flow rate on the condenser side to the rated water flow rate. This is the ratio of the actual outlet water temperature on the evaporator side to the rated outlet water temperature. This is the ratio of the actual return water temperature on the condenser side to the rated return water temperature.

[0030] a1-f2 are the fitting coefficients, which are obtained by fitting heat pump sample data or actual operating data.

[0031] Ground-source heat exchange (heat dissipation into the soil) , (2) Heating mode: Formula for calculating the heating capacity of a ground source heat pump: , Formula for calculating the heating power consumption of a ground source heat pump: , , in: Heating capacity of the ground source heat pump unit, kW; The power consumption of the ground source heat pump unit is kW; The heating capacity of the unit under rated operating conditions is expressed in kW. This is a correction factor for the heating capacity of the heat pump unit under actual operating conditions; This refers to the partial load rate of the heat pump unit. The input power of the heat pump unit under rated heating conditions, in kW; This is the input power correction factor for the heat pump unit under full-load operation. This is the input power correction factor for the heat pump unit under partial load operation conditions.

[0032] Calculation of correction factor for heating condition: , , , in: This is the ratio of the actual water flow rate on the condenser side to the rated water flow rate. This is the ratio of the actual water flow rate on the evaporator side to the rated water flow rate. This is the ratio of the actual outlet water temperature on the condenser side to the rated outlet water temperature. This is the ratio of the actual return water temperature on the evaporator side to the rated return water temperature. a4-f4 are the fitting coefficients, which are obtained by fitting heat pump sample data or actual operating data.

[0033] Ground-source heat exchange (heat extraction from the soil): , III. Performance Model of Air Source Heat Pump The operating performance of air source heat pump units is significantly affected by outdoor weather conditions, especially changes in outdoor air temperature, which directly impact their heating and cooling capacity and energy efficiency. Therefore, this invention employs a polynomial correction coefficient method to establish an air source heat pump performance model to accurately reflect its performance variation characteristics under different operating conditions.

[0034] 3.1 Heating Condition: Heating capacity of air source heat pump: , Coefficient of performance (COD) of air source heat pump: , in: The heating capacity of the unit under rated operating conditions is expressed in kW. This is a correction factor for the heating capacity of the air source heat pump unit under actual operating conditions. The correction factor for heating capacity during defrosting of the air source heat pump unit is set to 0.9. The rated thermal performance coefficient of the unit; This is the unit's COP correction factor; This is the correction factor for COP at partial load rate; The loss coefficient for frosting and defrosting is set to 0.96.

[0035] Heating capacity correction factor: , Heating efficiency correction factor: , , in: This is the ratio of the actual outdoor dry-bulb temperature to the rated outdoor dry-bulb temperature under thermal conditions. This is the ratio of the actual condenser outlet water temperature to the condenser outlet water temperature under rated operating conditions. This refers to the partial load rate of the air source heat pump unit. a1-e3 are the fitting coefficients, which are obtained by fitting heat pump sample data or actual operating data.

[0036] 3.2 Refrigeration Conditions: Air source heat pump cooling capacity: , Air source heat pump cooling correction factor: , in: The cooling capacity of the unit under rated operating conditions is expressed in kW. This is a correction factor for the cooling capacity of the air source heat pump unit under actual operating conditions. The rated cooling performance coefficient of the unit; This is the unit's EER correction factor; This is the correction factor for EER at partial load rate. Cooling capacity correction factor: , Cooling energy efficiency correction factor: , , in: It is the ratio of the actual outdoor dry-bulb temperature to the outdoor dry-bulb temperature under rated cooling conditions; This is the ratio of the actual evaporator outlet water temperature to the evaporator outlet water temperature under rated operating conditions. This refers to the partial load rate of the air source heat pump unit. a1-f3 are the fitting coefficients, which are obtained by fitting heat pump sample data or actual operating data.

[0037] IV. Variable Frequency Water Pump Model To accurately describe the energy consumption characteristics of the variable frequency water pump in the system, this invention uses a cubic polynomial relationship between pump power and speed ratio to model the variable frequency water pump. The power model expression is as follows: , , in: This is the speed ratio, ranging from 0 to 1; and These are the variable speed and the rated speed, respectively, in r / min; and These are the current frequencies at the variable speed and rated speed, respectively, in Hz; Rated power at rated speed, in kW; , , , These are the fitting coefficients, which are obtained by fitting the actual operating data of the water pump.

[0038] This model can accurately reflect the power variation law of the water pump under different speed conditions, and provides a mathematical basis for the fine optimization of the water pump speed in the lower-level model predictive control.

[0039] V. Buried Pipe Heat Exchange Model The vertical U-shaped buried pipe heat exchanger in the Transient System Simulation Program (TRNSYS) is modeled using type 557a. This module, based on a finite line source model, is used to simulate the heat exchange process between the vertical U-shaped buried pipe heat exchanger and the soil under different operating conditions.

[0040] In the model of a buried pipe heat exchanger, the relationship between the inlet and outlet temperatures of the buried pipe and the soil temperature can be expressed as: , in: and These are the outlet and inlet temperatures of the buried pipe heat exchanger, respectively, in °C. Soil temperature, °C; The damping coefficient; , For soil volume, m 3 ; The fluid flow rate is expressed in kg / h. Volumetric specific heat capacity, kJ / (kg·K); The length of the buried pipe is in meters (m). and , respectively, are the thermal conductivity of the soil and the buried pipe, in W / (m·K).

[0041] The above model can dynamically calculate the changes in the inlet and outlet water temperature of the buried pipe and the soil temperature field during the simulation process, providing reliable input for the soil thermal balance constraints in the upper-level optimization and the state prediction in the lower-level control.

[0042] VI. Building Load Forecasting Model To achieve the load forecasting function in the upper-level optimization, this invention employs the Dung Beetle Optimizer (DBO) algorithm. Figure 2 A building load prediction model, namely the DBO-LSTM-Attention model, is constructed by optimizing a Long Short-Term Memory (LSTM) network and combining it with an attention mechanism. Its overall structure is as follows: Figure 3 As shown.

[0043] The building load forecasting model is constructed based on a long short-term memory network optimized by the dung beetle optimization algorithm and combined with an attention mechanism. The model input includes historical load sequences, external meteorological parameters, time features, and calendar features. The time-series features are extracted through a gating structure, and the weights of key time steps are calculated by the attention mechanism to generate a context vector. The dung beetle optimization algorithm is used to search for the hyperparameters of the forecasting model to minimize the prediction error on the validation set, thereby obtaining the future load forecast results for upper-level optimization.

[0044] 6.1 Input Feature Construction The input characteristics of the building load forecasting model include the following four categories: 1) Historical load sequence: Hourly load data for the past 24 hours; 2) Outdoor meteorological parameters: outdoor temperature, humidity, and solar radiation intensity; 3) Time characteristics: hour, day of the week, whether it is a holiday; 4) Calendar features: months and seasons.

[0045] The aforementioned multidimensional features are used together as model inputs to characterize the temporal variation of building loads and their relationship with external environmental factors.

[0046] 6.2 LSTM Network Structure The core architecture of LSTM includes three types of gating units: the forget gate (f) is responsible for filtering and retaining key information in historical data while eliminating redundant information; the input gate (i) adjusts the degree to which new information is integrated into the cell state; and the output gate (o) controls the amount of information output at the current time step.

[0047] like Figure 1 As shown, the LSTM unit achieves effective modeling of long-term dependent information through a gating mechanism, which can overcome the problem of gradient vanishing or gradient explosion that traditional recurrent neural networks are prone to in long-term series prediction.

[0048] 6.3 Attention Mechanism An attention mechanism is introduced based on the output of the LSTM network to enhance the model's ability to focus on information from key time steps. The attention mechanism consists of several sequential steps, including attention score calculation, weight allocation, and context vector generation.

[0049] Specifically, the alignment model is based on the encoder hidden state h i and the output s of the previous decoder t-1 Calculate score e t,i The score quantifies the correlation between each component in the input sequence and the current output at position t. The score function a(x) can be implemented using a feedforward neural network. Subsequently, the attention weight α is obtained by applying softmax normalization to the score. t,i The final context vector c t The weighted sum of all encoder hidden states is calculated and input into the decoder at each time step, as shown in the formula: , , .

[0050] 6.4 DBO Optimization Process To further improve prediction accuracy, the dung beetle optimization algorithm is used to optimize the hyperparameters of the LSTM network. The hyperparameters to be optimized include: Learning rate: search range [0.0001, 0.01]; Hidden layer dimension: search range [32, 256]; Number of LSTM layers: search range [1, 3]; Dropout rate: search range [0.1, 0.5].

[0051] The optimization objective of the DBO algorithm is to minimize the root mean square error (RMSE) on the validation set.

[0052] DBO algorithm parameters: population size; maximum number of iterations; fitness function: RMSE.

[0053] The DBO-LSTM-Attention prediction model described above can be used to obtain hourly building heating and cooling load predictions for the next 24 hours, which can then serve as an important input for optimizing multi-objective load distribution in the upper layers.

[0054] VII. System Operating Constraints To ensure the safety, stability, and feasibility of the multi-source heat pump coupling system during optimized operation, system operation constraints are introduced into the upper-level optimization and lower-level model predictive control. These constraints include load balance constraints, equipment capacity constraints, equipment power consumption constraints, temperature constraints, flow constraints, soil heat balance constraints, and equipment start-up and shutdown frequency constraints.

[0055] 7.1 Load Balance Constraints At any given time step, the total cooling or heating capacity provided by the ground source heat pump and the air source heat pump in the system should meet the building load demand, and its load balance relationship is expressed as:

[0056] Wherein, the building load Q load Output from the building load forecasting model. Q GSHP Q represents the cooling or heating capacity of a ground source heat pump. ASHP This refers to the cooling or heating capacity of an air source heat pump.

[0057] 7.2 Equipment Capacity Constraints The output capacity of ground source heat pumps and air source heat pumps is limited by their rated capacity, under the following constraints: , , in, The rated cooling or heating capacity of a ground source heat pump is expressed in kW. The rated cooling or heating capacity of an air source heat pump is expressed in kW. and These are the minimum output capacities of ground source heat pumps and air source heat pumps, respectively, in kW. They are usually determined based on the heat pump performance curve and are taken as 30% of the rated capacity.

[0058] 7.3 Equipment Power Constraints The input power of a heat pump unit is limited by its rated power, and the constraints are as follows: , , in, The input power of the ground source heat pump under rated operating conditions is kW. The input power of the air source heat pump under rated operating conditions is expressed in kW. and These are the minimum input power (kW) for ground source heat pumps and air source heat pumps, respectively. They are usually determined based on the equipment performance parameters and are taken as 20% of the rated power.

[0059] 7.4 Temperature Constraints To ensure system safety and end-user comfort, the following constraints are set for the system's supply and return water temperatures: Load-side water supply temperature constraints: Cooling season: ,

[0060] Heating season: , Ground source side return water temperature constraint: ,

[0061] 7.5 Flow Constraints To ensure the hydraulic balance of the system and the safe operation of the equipment, the following constraints are set for the system flow rate: Flow constraints on the load side of ground source heat pumps: , Ground source heat pump ground source side flow constraint: , Air source heat pump flow constraints: , in, =Rated flow rate of ground source heat pump × 0.6; Rated flow rate of ground source heat pump; =Rated flow rate of ground source heat pump × 0.7; =Rated flow rate of ground source heat pump; =Air source heat pump rated flow rate × 0.5; = Rated flow rate of air source heat pump.

[0062] 7.6 Soil thermal balance constraints To suppress the soil heat accumulation effect during the long-term operation of the ground source heat pump system, a soil heat balance constraint is introduced, the expression of which is: , , in, The annual cumulative heat extracted from the soil by a ground source heat pump is expressed in kWh. The annual cumulative heat discharged into the soil by the ground source heat pump is expressed in kWh.

[0063] This constraint limits the imbalance in annual heat exchange on the ground source side, preventing long-term soil temperature drift from causing system performance degradation.

[0064] 7.7 Equipment start-up and shutdown frequency constraints: To prevent frequent start-stop operations from causing mechanical wear and reduced energy efficiency, the following constraints are set for the equipment's operating status: 1) Equipment status variables: , , in, This indicates the operating status of the i-th ground source heat pump at time t, with a value of 0 or 1, where 1 indicates the equipment is running and 0 indicates the equipment is stopped. This indicates the operating status of the i-th air source heat pump at time t, with a value of 0 or 1, where 1 indicates the equipment is running and 0 indicates the equipment is stopped. i represents the ground source heat pump unit number, with values ​​1, 2, ..., NGSHP; j represents the air source heat pump unit number, with values ​​1, 2, ..., NASHP.

[0065] 2) Minimum continuous running time constraint: , ,

[0066] in, =2 hours =1 hour, k is the time step index, representing the time from the current time t to t+T. min,on -1 is accumulated.

[0067] 3) Minimum downtime constraint: , , =1 hour =1 hour.

[0068] VIII. TRNSYS-MATLAB Co-simulation Architecture To achieve efficient coupled computation of model predictive control and complex system physical processes, this invention adopts a co-simulation architecture of TRNSYS and MATLAB, in which TRNSYS is responsible for modeling the system physical processes and MATLAB is responsible for solving the optimization algorithm. The two interact with each other through the Type155 module provided by TRNSYS.

[0069] 8.1 Working principle of Type155 module Type155 is an external program call interface module provided by TRNSYS, which communicates directly with MATLAB through the interface.

[0070] Data interaction mechanism: TRNSYS → MATLAB (input): System state variables at the current simulation moment; building load data; outdoor meteorological parameters; inlet and outlet temperatures of buried pipes; inlet and outlet water temperatures and flow rates of each unit.

[0071] MATLAB → TRNSYS (output): The optimized calculations yielded the following control variables: temperature and flow rate; pump speed; and equipment start / stop signals.

[0072] 8.2 Co-simulation model structure TRNSYS side model composition: Type557a: Buried pipe heat exchanger module; Inputs: Ground source influent temperature and flow rate; Outputs: Geological source water temperature and soil temperature field; Parameters: borehole depth, spacing, soil thermal properties; Type 927: Ground Source Heat Pump Performance Model; Inputs: inlet and outlet water temperatures on the load side, inlet and outlet water temperatures on the ground source side, and flow rates; Outputs: Cooling / heating capacity, power consumption, coefficient of performance (COP); Type 941: Performance model of air source heat pump; Inputs: outdoor temperature, inlet and outlet water temperatures on the load side, flow rate; Outputs: Cooling / heating capacity, power consumption, coefficient of performance (COP); Type 114: Variable frequency water pump model; Output: Speed ​​ratio; Type 155: MATLAB interface module; Connect the TRNSYS physical model with the MATLAB optimization algorithm; Sampling period: 5 minutes (consistent with the MPC optimization period).

[0073] IX. Co-simulation Workflow The overall operation process of TRNSYS-MATLAB co-simulation is divided into the initialization phase and the simulation operation phase.

[0074] 9.1 Initialization Phase Perform the following steps before starting the simulation: 1. Start the TRNSYS simulation and load all Type modules; 2. Type 155 starts MATLAB on first call; 3. Load the MPC controller program and optimization algorithm into MATLAB; 4. Initialize state variables and historical data cache.

[0075] 9.2 Simulation Run Phase During the simulation phase, the system performs cyclic calculations with a control cycle of 5 minutes. The specific process is as follows: At time k: Step 1: The TRNSYS simulation proceeds to time k; Step 2: Type 155 collects system status information, including: Obtain building load from Type 56; Soil temperature was obtained from Type 557a; Obtain heat pump performance parameters from Type 927 / 941; Obtain pump operating parameters from Type 114; Step 3: Type 155 calls MATLAB to pass the state vector and the disturbance vector; Step 4: MATLAB performs MPC optimization, specifically including: 4.1 Status feedback correction; 4.2 Predicting state evolution in the time domain (24 steps): Calculate the state vector by calling the mathematical model; 4.3 MOPSO Optimization Solution: Initialize 50 particles, iterate 100 times, and obtain the Pareto optimal frontier; 4.4 Selecting the optimal solution: Adaptively adjust the weights based on the tracking error and select the control sequence with the minimum weighted objective function; 4.5 Extraction of the first step control; Step 5: MATLAB returns control commands to Type 155; Step 6: Type 155 distribution control instructions, including: Send temperature and flow rate settings to Type 927 / 941; Send the speed ratio to Type114; Step 7: TRNSYS continues the simulation up to time k+1. Each Type module operates according to the new control instructions, and the system responds and generates a new state.

[0076] Step 8: Loop back to Step 2 and enter the next control cycle.

[0077] 10. Two-layer optimization control strategy like Figure 4 As shown, this invention adopts a two-layer optimization control strategy, which divides the optimization problem of a multi-source heat pump coupling system into two levels, upper-layer optimization and lower-layer optimization, according to the time scale and control objective. This avoids the dimensionality curse caused by the mixing of time scales in single-layer optimization, while taking into account both the long-term operation objective and short-term control accuracy of the system.

[0078] 10.1 Upper Layer Optimization: Load Distribution Based on NSGA-II The upper-level optimization employs the NSGA-II multi-objective genetic algorithm to make load allocation decisions on a relatively long time scale (2 hours). It is mainly used to determine the target load allocation scheme between ground source heat pumps and air source heat pumps in the future forecast time domain.

[0079] (1) Prediction time domain 2 hours; (2) Decision variables; For each time step in the prediction time domain, the decision variable is the target load of each heat pump unit: (3) Objective function: The upper-level optimization is a multi-objective optimization problem, which includes at least the following three optimization objectives: Objective 1: Total power consumption; , , The power consumption of a ground source heat pump is measured in kWh. The power consumption of an air source heat pump is kWh. The power consumption of the water pump is measured in kWh. The power consumption of the water pump on the load side of the ground source heat pump (its calculation method is described in "IV. Variable Frequency Water Pump Model" above). Formulas (different fitting coefficients are obtained by fitting different pump models with actual operating data or sample data), kWh; The power consumption of the ground source side water pump in a ground source heat pump is , in kWh. The power consumption of the air source heat pump circulating water pump is kWh.

[0080] Objective 2: Operating Costs The formula for calculating system operating costs is: , , , in, Operating costs, CNY; Electricity cost for operation, CNY; Maintenance costs, CNY; , , Time-of-use electricity pricing, CNY / (kWh); , , Electricity consumption during peak, off-peak, and normal periods, in kWh.

[0081] Objective 3: Carbon emissions during operation The formula for calculating carbon emissions during system operation is as follows: , , , , in, This refers to the carbon emissions during the system's operation. ; and The refrigerant charge amounts for ground source heat pumps and air source heat pumps are respectively, in kg; The annual refrigerant leakage rate is 5%. and These are, respectively, global warming potential and atmospheric degradation products. ; for Grid emission factor .

[0082] (4) Constraints Upper-level optimization must satisfy the constraints defined in the aforementioned system mathematical model, including load balance constraints, equipment capacity constraints, soil thermal balance constraints, and equipment start-up and shutdown frequency constraints.

[0083] (5) NSGA-II algorithm parameters The non-dominated sorting genetic algorithm NSGA-II was used to solve the above multi-objective optimization problem. The algorithm parameters were set as follows: population size; maximum number of iterations; crossover probability; mutation probability; distribution index. The parameter settings range were: population size: 50-100; maximum number of iterations: 100-500; crossover probability: 0.7-0.95; mutation probability: 0.01-0.1; distribution index: 10-30.

[0084] (6) Pareto optimal solution selection strategy The NSGA-II algorithm outputs the Pareto optimal front, which contains multiple non-dominated solutions. An executable solution is selected using the following decision rule: Rule 1: Based on electricity price period; Rule 2: Based on soil temperature status.

[0085] (7) Output; The final selected target loads for the ground source heat pump and air source heat pump are used as the upper-level optimization outputs and transmitted to the lower-level model predictive control.

[0086] 10.2 Lower-level optimization: Real-time control based on MPC+MOPSO The lower-level optimization adopts a model predictive control (MPC) framework combined with a multi-objective particle swarm optimization (MOPSO) algorithm to perform real-time control on a short time scale (5 minutes), and is realized through TRNSYS-MATLAB co-simulation.

[0087] (1) Optimize the period and time domain. Optimization cycle: Execute once every 5 minutes (sampling interval). Prediction time domain: 24 steps (2 hours) Control time domain: 6 steps (30 minutes).

[0088] (2) Decision variables For the control time step, the decision variables include: Temperature setpoints: Ground source heat pump load-side outlet water temperature, air source heat pump outlet water temperature; Flow rate setpoints: Ground source heat pump load-side flow rate, air source heat pump flow rate; Pump speed ratio: load-side water pump of ground source heat pump, ground-side water pump of ground source heat pump, and side water pump of air source heat pump; Operating modes: Ground source heat pump operating mode (0=off, 1=heating, 2=cooling); Air source heat pump cooling / heating signal (0 / 1).

[0089] (3) Prediction model and TRNSYS-MATLAB co-simulation.

[0090] Based on the mathematical model constructed in step one, the future state evolution is predicted. State variables include: inlet and outlet temperatures of each device, flow rate, soil temperature, etc.; control variables are control outputs; disturbance variables include: building load (from DBO-LSTM-Attention prediction) and outdoor temperature.

[0091] (4) State feedback correction In each optimization cycle, the prediction bias is corrected using the measured conditions.

[0092] (5) Bi-objective optimization function Objective 1: Load Tracking Target , The predicted load (kW) borne by the ground source heat pump; The set load undertaken by the ground source heat pump, in kW; The predicted load (kW) borne by the air source heat pump; The set load undertaken by the air source heat pump, in kW.

[0093] Objective 2: Total system power consumption , Total power consumption of the GSHP-ASHP system (6) Constraints The lower-level MPC must meet the temperature constraints, flow constraints, pump speed constraints (0.7-1), and equipment start-stop frequency constraints defined in step one.

[0094] (7) MOPSO algorithm parameters The parameters of the multi-objective particle swarm optimization algorithm MOPSO are set as follows: Particle count; maximum number of iterations; inertia weight; individual learning factor; social learning factor; external archive size (for storing Pareto optimal solutions). Parameter setting range: Particle count: 30-100; maximum number of iterations: 50-200; inertia weight: 0.4-0.9; individual learning factor: 1.5-2.5; social learning factor: 1.5-2.5; external archive size: 50-200.

[0095] (8) Pareto optimal solution selection strategy An adaptive weighting method is adopted to dynamically adjust the load tracking deviation based on real-time load tracking.

[0096] (9) Rolling optimization mechanism MPC employs a rolling time-domain optimization strategy, and its execution process includes: Step 1: At time k, solve the optimization problem to obtain the control sequence u*(k), u*(k+1), ..., u*(k+N). c-1 ); Step 2: Only execute the control u*(k) from step 1; Step 3: At time k+1, collect new measurement data and perform state feedback correction; Step 4: Scroll forward one step in the time domain and re-solve the optimization problem; Step 5: Execute repeatedly to achieve closed-loop feedback control.

[0097] XI. Control System Implementation Architecture (Edge-Cloud Collaborative Architecture) like Figure 5 As shown, in order to apply the two-layer optimization control strategy to practical engineering, a three-layer architecture of edge-cloud collaboration was designed. The entire system consists of a cloud layer, an edge layer, and a device layer, with each layer working together to achieve the optimization control function.

[0098] 11.1 System Architecture Design (1) Cloud layer The cloud layer is deployed on a cloud server, and its main functions include: performing upper-layer NSGA-II load distribution optimization (once a day, with an optimization cycle of 2 hours); running the DBO-LSTM-Attention load prediction model; storing historical running data and performing long-term performance analysis; and providing a web-based visual monitoring interface.

[0099] The software environment uses a Windows operating system, MATLAB for NSGA-II optimization, Python in conjunction with a deep learning framework for load forecasting, a MySQL database (an open-source relational database management system) for storing historical data, and a Message Queuing Telemetry Transport (MQTT) service for message communication.

[0100] (2) Edge layer The edge layer is deployed on the edge computing server in the building's computer room, responsible for real-time control and device communication. Its main functions include: running TRNSYS-MATLAB co-simulation, where TRNSYS handles physical system simulation, MATLAB handles MPC+MOPSO optimization, and the Type155 module enables data interaction; executing real-time control of the lower-level MPC layer, employing 3-minute rolling optimization; collecting device data and issuing control commands through the OPC Unified Architecture (OPC UA) or ModBus communication protocol; and local data caching, enabling offline operation.

[0101] The software environment uses a Windows operating system, a TRNSYS 18 simulation platform and related modules (Type 155 MATLAB interface, Type 557a buried pipe model, Type 927 ground source heat pump model, Type 941 air source heat pump model), MATLAB and optimization toolbox, OPC UA and ModBus communication protocol stack, and SQLite local database.

[0102] (3) Equipment layer The equipment layer includes a ground source heat pump controller, an air source heat pump controller, a variable frequency water pump driver, a temperature sensor (PT1000 type), a flow meter (electromagnetic flow meter), a pressure transmitter, and a data acquisition module DAQ (supporting ModBus, OPC UA and other protocols).

[0103] 11.2 Communication Network Design like Figure 6 As shown, communication between the device layer and the edge layer uses the OPC UA protocol or the ModBus TCP / RTU (TCP Transmission Control Protocol / RTU Remote Terminal Unit) protocol. The network topology is star-shaped, and all devices are connected to the edge server through industrial Ethernet switches, with a latency of less than 50ms.

[0104] The edge layer sends control commands to the equipment (in real time, with a response time of less than 1 second): the operating mode of the ground source heat pump, the load-side water supply temperature setpoint (7-12℃ in the cooling season and 40-45℃ in the heating season), and the load-side flow rate setpoint; the start / stop signal, cooling / heating mode, water supply temperature setpoint, and flow rate setpoint of the air source heat pump; and the speed ratio setpoint and start / stop signal of the water pump.

[0105] Data collected from the equipment (every 3 minutes): inlet and outlet temperatures on the load side of the ground source heat pump, inlet and outlet temperatures on the ground source side (the outlet temperature of the buried pipe should not exceed 33℃ in summer and should not be lower than 4℃ in winter), flow rate, power, operating status, and fault codes; inlet and outlet temperatures, flow rate, power, outdoor temperature, operating status, and fault codes of the air source heat pump; speed, power, flow rate, inlet and outlet pressure, and operating status of the water pump; environmental parameters including outdoor temperature, humidity, and soil temperature.

[0106] Communication between the edge layer and the cloud layer uses the MQTT protocol or application programming interface (API) via 4G / 5G or VPN connection, with a latency of less than 500ms.

[0107] Edge layer uplink data (2-hour cycle): timestamp, soil temperature, actual load, target load, tracking error, total power, equipment operating status, optimized performance indicators, and alarm information.

[0108] Cloud-based downlink data (once a day): timestamp, load allocation scheme (24 steps), target load for each heat pump, parameter updates, and emergency mode flag.

[0109] 11.3 System Deployment Steps Hardware installation: Install temperature sensors at the inlet and outlet of each heat pump load side and ground source side; install flow meters on the main pipeline; install pressure transmitters at the inlet and outlet of the water pump; connect the water pump frequency converter; install an outdoor weather station; lay industrial Ethernet cables and install switches; configure network addresses; install an edge server in the computer room and connect it to a UPS power supply.

[0110] Software Deployment: On the cloud, a Windows system is installed, along with MATLAB, Python, MySQL, and MQTT services. Scheduled tasks are configured to perform daily upper-layer optimizations, and a web monitoring interface is deployed. On the edge, a Windows system is installed, along with TRNSYS and MATLAB. Control programs are compiled, communication protocol libraries are installed, and automatic startup is configured. Connections to the device layer and the cloud are tested.

[0111] System debugging: Test equipment communication and verify the accuracy of data acquisition; collect at least one week of continuous operation data, calibrate the geothermal model parameters, and verify the accuracy of the prediction model (RMSE<5%); conduct closed-loop operation tests, monitor load tracking error (requirement <10%), verify the effect of multi-objective optimization, and test the fault tolerance mechanism.

[0112] XII. Engineering Application Examples Taking a middle school in Beijing as an example, with a building area of ​​24,500 square meters... 2 The system is configured with two ground source heat pump units (each with a rated cooling capacity of 1519kW and a heating capacity of 1557kW) and 120 underground pipe boreholes (each with a depth of 122m, a spacing of 4m, a soil thermal conductivity of 1.8W / (m·K), and a far-field temperature of 12.9°C), nine air source heat pump units (each with a rated cooling capacity of 150kW and a heating capacity of 170kW), and a matching variable frequency water pump system. The control system adopts an edge-cloud collaborative architecture. The local time-of-use electricity price is 1.2710 yuan / kWh during peak hours, 0.2849 yuan / kWh during off-peak hours, and 0.7523 yuan / kWh during normal hours.

[0113] A complete system simulation model was established in the TRNSYS environment, including the Type 56 building load module, the Type 557a underground pipe module, the Type 927 ground source heat pump module, the Type 941 air source heat pump module, and the Type 114 variable frequency water pump module. Co-simulation with MATLAB was achieved through the Type 155 module. Building load data was generated using typical meteorological year data, covering 8760 hours of hourly load throughout the year. The upper-level NSGA-II optimization algorithm was set with a population size of 100, 200 iterations, and an optimization cycle of 2 hours. The lower-level MPC+MOPSO algorithm was set with 24 steps in the prediction time domain, 6 steps in the control time domain, 50 particles, 100 iterations, and an optimization cycle of 5 minutes.

[0114] To verify the superiority of the method of this invention, a comparative simulation scheme was set up to conduct continuous simulation throughout the year: Strategy 1: Traditional Control Strategy (1) Ground source heat pump operation strategy When both units are started at the same time, heat pump unit 1 is started first. When the operating load of unit 1 reaches more than 75%, heat pump unit 2 is started. When both units are working at the same time and the load of both units is less than 30%, heat pump unit 2 is shut down.

[0115] (2) Air source heat pump operation strategy In cooling mode, the air source heat pump starts when the inlet water temperature of the local source side main pipeline is ≥31℃ and starts in stages according to the temperature. It stops operating when the temperature is <26℃. In heating mode, it starts when the inlet water temperature is <10℃ and operates in stages. It stops operating when the temperature is ≥15℃.

[0116] (3) Load-side water pump and ground-source water pump strategy of ground source heat pump In both cooling and heating modes, the inverter controls the temperature difference between the inlet and outlet water of the condenser and evaporator of the unit, maintaining a stable temperature difference of 5℃, with a minimum protection frequency of 35Hz.

[0117] (4) Control strategy for air source heat pump circulating water pump The system automatically adjusts its operation based on the temperature difference of the manifold, maintaining a stable temperature difference of 5℃, with a minimum protection frequency of 35Hz.

[0118] Strategy 2: Two-layer optimization strategy (method of this invention): The ground source heat pump coupled with the air source heat pump system adopts the dual-layer optimization control method of the present invention. The upper layer NSGA-II performs multi-objective load distribution, and the lower layer MPC&MOPSO performs real-time optimization control.

[0119] (1) Comparison of power consumption of each device Figure 7The cumulative power consumption of ground source heat pumps, air source heat pumps, and water pumps under different strategies is shown. Strategy 1: Ground source heat pump cumulative power consumption is 214407.9 kWh, air source heat pump cumulative power consumption is 137555.4 kWh, total water pump cumulative power consumption is 106175.2 kWh, and total system power consumption is 458138.5 kWh. Strategy 2: Ground source heat pump cumulative power consumption is 191497.3 kWh, air source heat pump cumulative power consumption is 144129.2 kWh, total water pump cumulative power consumption is 62899.2 kWh, and total system power consumption is 398525.6 kWh.

[0120] Comparative analysis shows that Strategy 2 reduces the total power consumption of the system by 59,612.9 kWh compared to Strategy 1, achieving an energy saving rate of 13.01%. The most significant reduction was in water pump power consumption, decreasing by 43,276 kWh, a decrease of 40.77%. This is mainly due to the fine-tuning of the water pump speed by the lower-level MPC, avoiding prolonged high-speed operation of the water pump as in traditional control. Ground source heat pump power consumption decreased by 22,910.6 kWh, a decrease of 10.68%, indicating that optimized control allows the ground source heat pump to operate at a more reasonable load rate. Air source heat pump power consumption increased slightly by 6,573.8 kWh, an increase of 4.78%. This is because the optimized strategy appropriately increased the use of air source heat pumps under certain operating conditions to maintain soil thermal balance and utilize off-peak electricity prices. However, due to the significant energy savings of the ground source heat pump and water pump, the system still achieved a significant overall energy saving effect.

[0121] (2) Comparison of operating costs Figure 8 The comparison of total operating costs under different strategies is shown. The total annual operating cost of Strategy 1 is 525,779.97 yuan, while the total annual operating cost of Strategy 2 is 496,443.50 yuan. Strategy 2 reduces the total system cost by 29,336.47 yuan compared to Strategy 1.

[0122] The cost reduction rate (5.58%) is significantly lower than the power consumption reduction rate (13.01%). This is because Strategy 2 fully utilizes the time-of-use electricity price signal through the upper-level optimization algorithm. Specifically, it prioritizes the scheduling of high-efficiency ground source heat pumps during peak electricity price periods to reduce high-priced electricity consumption; and appropriately increases the operating ratio of air source heat pumps during off-peak electricity price periods to reduce the weighted average electricity price. Although the total power consumption of the system decreases by 13.01%, the actual reduction in electricity expenses is relatively small due to the optimized adjustment of the electricity consumption time structure. From the perspective of the economic benefits over the entire life cycle, Strategy 2 saves approximately RMB 29,300 in operating costs annually. Calculated based on a system design life of 15-20 years, the cumulative economic benefits can reach RMB 440,000-590,000.

[0123] With an initial ground temperature set at 12.9℃, Strategy 1, after 8760 hours of continuous simulation throughout the year, resulted in a cumulative ground temperature increase of 0.16℃, eventually stabilizing at 13.06℃. Strategy 2, after the same 8760 hours of simulation, resulted in a cumulative ground temperature increase of 0.09℃, eventually stabilizing at 12.99℃. Strategy 2 showed a smaller annual ground temperature increase than Strategy 1, effectively mitigating the problem of accumulated soil cooling.

[0124] The soil temperature change rate of Strategy 2 is only 56.25% of that of Strategy 1, indicating that the soil thermal balance constraint introduced by the upper-level optimization in this invention effectively suppresses the cumulative effect of soil temperature. Although the soil temperature change rates of both strategies are small in the single-year simulation, if the system operates continuously for many years, the cumulative effect of soil temperature in Strategy 1 will intensify year by year, which may lead to a gradual decline in the performance of the ground source heat pump, or even the risk of ground temperature exceeding the limit and failing to operate normally. Strategy 2 maintains the annual soil thermal balance by dynamically adjusting the load distribution ratio of the ground source heat pump and the air source heat pump, ensuring the long-term stable operation of the system.

[0125] In summary, the dual-layer optimization control method for a ground-source heat pump coupled with an air-source heat pump system proposed in this invention utilizes an edge-cloud collaborative architecture to deploy upper-layer multi-objective optimization in the cloud and lower-layer real-time control at the edge layer. This achieves rational allocation of computing resources and demonstrates excellent performance in energy saving, economy, environmental protection, and system stability. This method combines optimization performance with engineering feasibility, possessing significant theoretical and practical application value, and providing an effective technical solution for the intelligent control of building HVAC systems.

[0126] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A control method for a multi-source heat pump coupling system based on two-layer optimization, characterized in that, Includes the following steps: S1: Obtain cooling and heating operating data for each heat pump; the multi-source heat pump includes ground source heat pumps and / or air source heat pumps; S2: Perform upper-level optimization, where the target load of each heat pump is used as the decision variable, and the target load of each heat pump in the first time scale is obtained through the first multi-objective optimization algorithm; S3: Perform lower-level optimization, where the real-time operating status data of each heat pump is used as the decision variable, and the control command of each heat pump in the second time scale is determined by the second multi-objective optimization algorithm to match the target load of step S2; S4: Each heat pump executes the control command; Wherein, the first time scale is larger than the second time scale.

2. The control method for a multi-source heat pump coupling system based on dual-layer optimization according to claim 1, characterized in that, In S2, the objective function of the first multi-objective optimization algorithm is one or more of the following optimization objectives: total power consumption, operating cost, and carbon emissions during operation.

3. The control method for a multi-source heat pump coupling system based on dual-layer optimization according to claim 1, characterized in that, In step S2, the constraints of the first multi-objective optimization algorithm are one or more of the following: load balance constraint, equipment capacity constraint, soil thermal balance constraint, and equipment start-stop frequency constraint. The load balance constraint is that the total cooling or heating provided by the heat pump in the system should meet the building load demand at any time step. The equipment capacity constraint is that the output capacity of the heat pump meets its rated capacity limit. The soil thermal balance constraint is to prevent the accumulation of soil heat. The equipment start-stop frequency constraint is to prevent frequent start-stop.

4. The control method for a multi-source heat pump coupling system based on dual-layer optimization according to claim 3, characterized in that, The building load in the load balance constraint is: based on historical building load data and historical and / or predicted meteorological data, the building cooling and heating load at each time step in the future preset time domain is output using the building load prediction model.

5. The control method for a multi-source heat pump coupling system based on dual-layer optimization according to claim 1, characterized in that, In step S3, the decision variables of the second multi-objective optimization algorithm include one or more of the following: heat pump temperature setpoint, heat pump flow rate setpoint, water pump speed ratio, and operating mode.

6. The control method for a multi-source heat pump coupling system based on dual-layer optimization according to claim 1, characterized in that, In step S3, the objective function of the second multi-objective optimization algorithm is one or more of the following optimization objectives: minimizing the tracking error of the actual output load of the heat pump to the target load of step S2; minimizing the total power or total energy consumption of the system.

7. The control method for a multi-source heat pump coupling system based on dual-layer optimization according to claim 1, characterized in that, In step S3, the constraints of the second multi-objective optimization algorithm are one or more of the following: temperature constraints, flow constraints, pump speed constraints, and equipment start-up and shutdown frequency constraints, wherein the temperature constraints include cooling season / heating season temperature constraints and ground source side return water temperature constraints.

8. The control method for a multi-source heat pump coupling system based on dual-layer optimization according to claim 1, characterized in that, The first multi-objective optimization algorithm is the NSGA-II multi-objective genetic algorithm, and the second multi-objective optimization algorithm is the model predictive control (MPC) framework. In each optimization cycle, the system operating state in the future multiple time steps is predicted based on the system mathematical model, and the control variables are solved using the multi-objective particle swarm optimization algorithm (MOPSO).

9. A control system for a multi-source heat pump coupling system based on two-layer optimization, used to execute the control method for a multi-source heat pump coupling system based on two-layer optimization as described in any one of claims 1-8, characterized in that, include: The data acquisition module is configured to execute step S1; The upper-level optimization module is configured to execute step S2; The lower-level optimization module is configured to execute step S3; The control execution module is configured to execute step S4.

Images