Energy-saving management and control method and system for cold and heat source system based on particle swarm embedded decoupling

Abstract

This invention discloses an energy-saving control method and system for cold and heat source systems based on particle swarm optimization and nested decoupling, relating to the field of building energy conservation. It solves the problem of solving the coupling parameters of building cold and heat source systems. The key technical points are: establishing an equipment group model of the building cold and heat source system, including chillers, ground source heat pumps, chilled water pumps, ground source circulating pumps, cooling towers, and buried pipes; mapping and matching the associated coupling variables with the same physical meaning among the equipment based on the equipment group model to establish a cold and heat source system group model, including chiller systems and ground source heat pump systems; and using a particle swarm optimization algorithm to optimize and solve the global control parameters of the system based on the cold and heat source system group model, outputting the optimal values ​​of the global control parameters; thus achieving optimal energy-saving control of the building cold and heat source system as well as global collaborative optimization and accurate modeling of the building cold and heat source system.

Description

Technical Field

[0001] This invention relates to the field of building energy conservation, and more specifically, to a method and system for energy-saving management of cold and heat source systems based on particle swarm optimization and nested decoupling. Background Technology

[0002] Building heating and cooling systems are an important component of building energy consumption, accounting for up to 40% of total building energy consumption. Reducing the energy consumption of building heating and cooling systems is of great significance for energy conservation and carbon reduction in the building sector.

[0003] With the development of IoT technology, the management and control of building heating and cooling systems has gradually shifted from traditional manual operation and maintenance to automation. Public buildings often install management and control equipment and platforms to remotely start and stop, and interlock start and stop functions for building heating and cooling systems. Some management and control platforms can achieve energy savings by adjusting system operating parameters. However, current parameter adjustment methods are mostly manual adjustments by maintenance personnel based on their own experience, resulting in low levels of intelligence and a lack of scientific rigor. It is difficult to determine the optimal operating parameters based on actual needs, and the system still has significant potential for energy savings.

[0004] In existing technologies, artificial intelligence optimization algorithms are increasingly being applied to the management and control of building heating and cooling source systems. Heuristic algorithms such as particle swarm optimization, in the iterative process of solving for optimal control parameter values, construct energy consumption calculation functions for key equipment as the optimization objective function. The equipment energy consumption calculation model includes control parameters and system state parameters. However, building heating and cooling source systems have high complexity and strong coupling. System state parameters are related to the control parameters of each device, and there are closed-loop relationships between devices, with parameters acting as inputs and outputs to each other. Current practical management and control processes generally optimize each device independently, making it impossible to achieve globally optimal system control.

[0005] Therefore, there is an urgent need for a method that can solve the above-mentioned technical problems. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide an energy-saving control method and system for cold and heat source systems based on particle swarm optimization and nested decoupling. This achieves optimal energy-saving control of the building's cold and heat source system as well as global collaborative optimization and accurate modeling of the system. It fundamentally solves the technical problem of traditional methods that require independent optimization of each device and cannot perform accurate modeling and collaborative control from the perspective of the entire system.

[0007] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0008] Firstly, an energy-saving management method for cold and heat source systems based on particle swarm optimization and nested decoupling is provided, including the following steps:

[0009] S1. Establish a model of the equipment group of the building's cold and heat source system. The equipment group model includes a chiller unit, a ground source heat pump, a chilled water pump, a ground source circulating pump, a cooling tower model, and buried pipes.

[0010] S2. Based on the equipment group model, map and match the associated coupling variables with the same physical meaning among the equipment to establish a cold and heat source system group model, which includes a chiller system and a ground source heat pump system.

[0011] S3. Based on the cold and heat source system group model, the particle swarm optimization algorithm is used to optimize and solve the global control parameters of the system, and the optimal value of the global control parameters of the system is output.

[0012] Furthermore, in step S3, the particle swarm optimization algorithm includes an outer particle swarm optimization process and an inner particle swarm optimization process;

[0013] The outer particle swarm optimization process is used to find the optimal values ​​of the global system control parameters that minimize the total energy consumption of the system.

[0014] The inner-layer particle swarm optimization process is used to decouple and solve the internal coupling variables of the system for the given control parameters of the outer-layer particle swarm.

[0015] Furthermore, the outer particle swarm optimization process includes the following steps:

[0016] S3a. Input system condition variables, initialize the position vector and velocity vector of the outer particle swarm, and determine the search range and boundary conditions of the position vector;

[0017] S3b, Update the velocity vector and position vector of the outer particle swarm;

[0018] S3c. For each particle at the updated position, call the inner particle swarm decoupling solution process to solve the system state parameters of each particle after the outer particle swarm updates its position.

[0019] S3d: Calculate the objective function for the updated positions of outer particles using the system swarm model and the system state parameters output by the inner particle swarm, and update the individual extreme values ​​and the global optimal solution;

[0020] S3e, repeat steps S3b to S3d until the convergence condition is met or the maximum number of iterations is reached, and output the optimal value of the global control parameter.

[0021] Furthermore, the inner-layer particle swarm optimization process includes the following steps:

[0022] S3c1. Input the system condition variables and the position vector of the outer particle swarm, initialize the position vector and velocity vector of the inner particle swarm, the position vector of the inner particle swarm is a coupling parameter, including the condenser inlet water temperature and condenser outlet water temperature, the ground source outlet water temperature and the ground source inlet water temperature;

[0023] S3c2. Compare the input position vector of the outer particle swarm with the position of the optimal solution of the outer particle in the previous optimization step to obtain the comparison result. Based on the comparison result, query the preset search space boundary adjustment rules to determine the search direction adjustment strategy of the inner particle swarm coupling parameters. Adjust the search space boundary of the corresponding position vector of the inner particle swarm according to the adjustment strategy.

[0024] S3c3, Perform inner-layer particle swarm search and update the position vector and velocity vector of the inner-layer particle swarm;

[0025] S3c4. Input the updated inner particle position vector into the coupled device model, perform cross-prediction, and obtain the cross-prediction value.

[0026] S3c5. Establish a decoupled solution evaluation objective function, and use the current position of the inner particle and the cross prediction value to solve the objective function of each particle in the inner particle swarm.

[0027] S3c6. Repeat steps S3c3 to S3c5 until the convergence condition is met or the maximum number of iterations is reached, and output the solved system operating parameters.

[0028] Furthermore, the search space boundary adjustment rules include:

[0029] For the chiller system, the search direction of the condenser inlet water temperature and condenser outlet water temperature is adjusted according to the changing trend of the cooling water pump speed ratio and the cooling tower fan speed ratio.

[0030] For ground source heat pump systems, the search direction for ground source outlet water temperature is adjusted based on the changing trend of the ground source circulating pump speed ratio.

[0031] Furthermore, the total energy consumption of the system is the sum of the power of the main equipment of the cold and heat source system, including the main unit's cooling power, the main unit's heating power, the power of the chilled water pump or load water pump, the power of the cooling water pump or ground source circulation pump, and the power of the cooling tower.

[0032] Furthermore, the calculation process of the decoupled solution to the evaluation objective function is as follows:

[0033] Get the current particle position value;

[0034] The position value, along with the condition variables, the outer particle position, and the control parameters, is input into another device model coupled with it to calculate the predicted value.

[0035] The position values ​​of the particle hypothesis are compared with the predicted values ​​obtained through cross-prediction, and the squared deviations of the chiller system and the ground source heat pump system are calculated respectively.

[0036] The squared deviations of the chiller system and the ground source heat pump system are added together, and then the square root is taken to obtain the objective function value of the particle.

[0037] Furthermore, the method for adjusting the search space boundary of the corresponding position vector of the inner particle swarm also includes the following steps:

[0038] The pre-trained reinforcement learning agent is invoked, and the state feature vector is composed of the comparison result between the position vector of the outer particle swarm and the position of the optimal solution in the previous optimization step, the current state of the inner particle swarm, and the system condition variables. This vector is then input into the reinforcement learning agent.

[0039] The reinforcement learning agent outputs the corresponding boundary adjustment action based on the state feature vector;

[0040] Based on the boundary adjustment action, the search space boundary of the inner particle swarm position vector is dynamically adjusted.

[0041] Furthermore, the step of the reinforcement learning agent outputting the corresponding boundary adjustment action based on the state feature vector includes the following steps:

[0042] The policy network of the reinforcement learning agent performs multi-layer nonlinear transformation on the input state feature vector to output an action probability distribution. Each action in the action probability distribution corresponds to a predefined atomic operation that translates or scales the search space boundary.

[0043] Based on the action probability distribution, a final boundary is selected to adjust the action by sampling or selecting the maximum probability value;

[0044] The boundary adjustment action is executed to generate specific boundary adjustment instructions, which include the direction and magnitude of the adjustment.

[0045] Secondly, an energy-saving management and control system for a cold and heat source system based on particle swarm optimization and nested decoupling is provided. This system is used to implement the energy-saving management and control method for a cold and heat source system based on particle swarm optimization and nested decoupling as described in any one of the first aspects, including:

[0046] The equipment group module is used to establish an equipment group model of the building's cold and heat source system. The equipment group model includes chiller units and ground source heat pumps, ground source heat pumps, chilled water pumps, cooling water pumps and ground source circulating pumps, cooling tower models, and underground pipes.

[0047] The system group module is used to map and match the associated coupling variables with the same physical meaning among the devices based on the device group model, and to establish a cold and heat source system group model, which includes a chiller system and a ground source heat pump system.

[0048] The parameter analysis and decision module is used to optimize and solve the global control parameters of the system based on the cold and heat source system group model using a particle swarm optimization algorithm, and output the optimal values ​​of the global control parameters of the system.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. This invention establishes a data-driven equipment group model of a building's heating and cooling source system, maps and matches the associated coupling variables with the same physical meaning among the equipment, constructs a heating and cooling source system group model, and uses a particle swarm optimization algorithm to optimize and solve the global control parameters of the system, outputting the optimal values ​​of the global control parameters. It integrates the advantages of precise single-equipment models to solve the problem of difficult solution of coupling parameters of building heating and cooling source systems. Taking the total energy consumption of the main equipment such as the system's cooling and heating units, water pumps, and cooling towers as the target, it performs fast and accurate global optimization and solution of control parameters such as the set temperature of the cooling and heating units, the speed ratio of the water pumps, and the speed ratio of the cooling towers. This achieves the overall optimal energy-saving control of the building's heating and cooling source system and the global collaborative optimization and accurate modeling of the building's heating and cooling source system. It fundamentally solves the technical problem of traditional methods that optimize each equipment independently and cannot accurately model and coordinately manage the system from the perspective of the whole.

[0051] 2. This invention uses an outer-layer particle swarm optimization process to automatically and quickly solve for the optimal value of the global control parameters that minimizes the total energy consumption of the system by taking the total energy consumption of the main equipment such as refrigeration and heating units, water pumps, and cooling towers as the unified optimization target. This overcomes the blindness and limitations of relying on manual experience to adjust parameters, thereby achieving the highest energy efficiency of the system as a whole and significantly improving the search efficiency and energy-saving effect of the system's global optimal control parameters.

[0052] 3. This invention utilizes an inner-layer particle swarm optimization process specifically designed to solve highly coupled state parameters within a system (such as condenser inlet / outlet water temperatures). By employing cross-prediction and decoupling evaluation objective functions, it breaks the deadlock of closed-loop correlations between devices, ensuring that the actual operating state within the system can be accurately and quickly calculated under any given control parameters. This provides a reliable basis for calculating objective functions for outer-layer optimization, enabling accurate and rapid solutions to coupled parameters and effectively solving the problem of rapid and high-precision decoupling solutions for strongly coupled variables within the system.

[0053] 4. This invention employs a dynamic search space boundary adjustment mechanism. This mechanism can dynamically sense and intelligently adjust the search space and direction of the inner coupling parameters based on the changing trend of the outer control parameters and the real-time state of the system. This enables the search process to have adaptive capabilities and can more accurately locate the optimal solution, thereby significantly reducing the number of iterations required for the inner particle swarm and comprehensively improving the solution efficiency and robustness of the overall optimization algorithm. Attached Figure Description

[0054] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0055] Figure 1 This is a flowchart from Embodiment 1 of the present invention;

[0056] Figure 2 This is the chiller unit system group model and variable relationship diagram in Embodiment 1 of the present invention;

[0057] Figure 3 This is a group model and variable relationship diagram of the ground source heat pump system in Embodiment 1 of the present invention;

[0058] Figure 4 This is a flowchart of the particle swarm nested control parameter optimization solution algorithm framework in Embodiment 1 of the present invention;

[0059] Figure 5 This is a flowchart of the calculation of the inner-layer particle swarm objective function in Embodiment 1 of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0061] Example 1: This example provides an energy-saving management method for cold and heat source systems based on particle swarm optimization and nested decoupling, targeting energy-saving management of building cold and heat source systems. Figure 1 As shown, it includes the following steps:

[0062] S1. Establish a model of the equipment group of the building's cold and heat source system. The equipment group model includes chiller units, ground source heat pumps, chilled water pumps, ground source circulating pumps, cooling tower models, and underground pipes.

[0063] S2. Based on the equipment group model, the associated coupling variables with the same physical meaning among the equipment are mapped and matched to establish a cold and heat source system group model, which includes the chiller system and the ground source heat pump system.

[0064] S3. Based on the cold and heat source system group model, the particle swarm optimization algorithm is used to optimize and solve the global control parameters of the system, and output the optimal values ​​of the global control parameters of the system.

[0065] In step S1, the device cluster model adopts a data-driven neural network model. The neural network model training process is as follows:

[0066] The training data source is the actual operating parameters and energy consumption parameters of the building's heating and cooling system. The sample size is the operating parameters of the cooling and heating seasons for one consecutive year. The neural network swarm model uses the ReLU function as the activation function and the root mean square error as the error function. The maximum number of iterations is 1000. The R2 evaluation model is used, and the R2 can reach above 0.95. The prediction relative error is within 10%.

[0067] The population size and number of iterations are not fixed in practice; the specific settings are as follows:

[0068] (1) External particle swarm population size 50, maximum number of iterations 20, inertia factor 0.6, c1=c2=2,

[0069] (2) The internal particle swarm population size is 50, the maximum number of iterations is 50, the inertia factor is 0.6, c1=c2=2, and the parameters are selected based on the results of multiple training and parameter tuning.

[0070] In step S1, the inputs and outputs of each device are defined as follows, where For the input feature vector, This is the output vector.

[0071] (1) Chiller units and ground source heat pumps (cooling operation):

[0072] ;

[0073] in, This represents the unit load rate, with a value range of [0.3, 1]. This refers to the water flow rate of the condenser; the range of values ​​will be determined based on the actual project requirements. The condenser inlet water temperature, with a value range of [15, 35]. Set the outlet water temperature value, with a range of [6, 9]. This refers to the cooling capacity of the main unit; the range of values ​​should be determined based on the actual project requirements. —Condensate outlet water temperature, with a value range of [17, 40].

[0074] (2) Ground source heat pump (heating mode):

[0075] ;

[0076] in, This represents the unit load rate, with a value range of [0.3, 1]. This refers to the condenser water flow rate; the range of values ​​will be determined based on the actual project requirements. The evaporator inlet water temperature, with a value range of [10, 20]. Set the outlet water temperature value, with a range of [40, 50]. The value range for heating capacity is determined based on the actual project requirements. The evaporator outlet water temperature ranges from [15, 25].

[0077] (3) Chilled water pump (load pump):

[0078] ;

[0079] in, The system load factor has a value range of [0.2, 1]. This refers to the number of water pumps; the range of values ​​will be determined based on the actual project requirements. This is the rated speed ratio, with a value range of [0,1]. This refers to the power of the water pump; the range of values ​​should be determined based on the actual project requirements. This represents the water pump flow rate; the range of values ​​will be determined based on the actual project requirements.

[0080] (4) Cooling water pump, ground source circulation pump:

[0081] ;

[0082] in, This refers to the number of water pumps; the range of values ​​will be determined based on the actual project requirements. This is the rated speed ratio, with a value range of [0,1]. This refers to the power of the water pump; the range of values ​​should be determined based on the actual project requirements. This represents the water pump flow rate; the range of values ​​will be determined based on the actual project requirements.

[0083] (5) Cooling tower model:

[0084] ;

[0085] in, The outdoor wet-bulb temperature, with a value range of [10, 40]. The rated speed ratio of the fan is [0,1]. The value range for cooling water flow rate is determined based on the actual project requirements. The condensate outlet water temperature, with a value range of [17, 40]. The value range for the cooling tower power should be determined based on the actual project requirements. The condensate inlet water temperature is [15, 35].

[0086] (6) Buried pipe:

[0087] ;

[0088] in, The soil temperature of the geothermal well, with a value range of [10, 25]. The value range for the circulating flow rate is determined based on the actual project requirements. The temperature of the ground source water is taken as [10, 40]. The temperature of the ground-source water is [10, 35].

[0089] In step S2, the variables of the cold and heat source system group model are divided into system input variables, internal correlation and coupling variables, and system output variables. The chiller unit system group model and variable relationships are as follows: Figure 2 As shown, the ground source heat pump system group model and variable relationships are as follows: Figure 3 As shown.

[0090] In step S3, the nested particle swarm optimization algorithm includes an outer particle swarm optimization process and an inner particle swarm optimization process. The outer particle swarm optimization process is used to find the optimal value of the global control parameters that minimizes the total energy consumption of the system. The inner particle swarm optimization process is used to decouple and solve for the internal coupling variables of the system based on the control parameters given by the outer particle swarm. The flowchart of the nested particle swarm control parameter optimization algorithm is shown below. Figure 4 As shown.

[0091] In step S3, the outer particle swarm optimization process includes the following steps: S3a, input system condition variables, initialize the position vector and velocity vector of the outer particle swarm, and determine the search range and boundary conditions of the position vector;

[0092] S3b, Update the velocity and position vectors of the outer particle swarm;

[0093] S3c. For each particle at the updated position, call the inner particle swarm decoupling solution process to solve the system state parameters of each particle after the outer particle swarm updates its position.

[0094] S3d: Calculate the objective function for the updated positions of outer particles using the system swarm model and the system state parameters output by the inner particle swarm, and update the individual extreme values ​​and the global optimal solution.

[0095] S3e, repeat steps S3b to S3d until the convergence condition is met or the maximum number of iterations is reached, and output the optimal value of the global control parameters.

[0096] The condition variable is { , , , , The outer particle position vector is the combination of control parameters to be optimized, including the rated speed ratio of the chilled water pump or load water pump, the rated speed ratio of the cooling water pump or ground source circulation pump, the rated speed ratio of the cooling tower fan, and the setpoint of the outlet water temperature of the refrigeration or heating equipment. The outer particle position vector is { , , , }

[0097] Step S3c includes the following steps: S3c1, Input system condition variables and the position vector of the outer particle swarm, initialize the position vector and velocity vector of the inner particle swarm, the position vector of the inner particle swarm is the coupling parameter, including the condenser inlet water temperature and condenser outlet water temperature, the ground source outlet water temperature and the ground source inlet water temperature; S3c2, Compare the input position vector of the outer particle swarm with the optimal solution position of the outer particle in the previous optimization step, obtain the comparison result, query the preset search space boundary adjustment rules based on the comparison result, determine the search direction adjustment strategy of the inner particle swarm coupling parameters, and adjust the inner particle swarm according to the adjustment strategy. S3c3: Perform inner-layer particle swarm search, updating the position and velocity vectors of the inner-layer particle swarm; S3c4: Input the updated inner-layer particle position vectors into the coupled device model, perform cross-prediction, and obtain cross-prediction values; S3c5: Establish a decoupled solution evaluation objective function, and use the current position of the inner-layer particles and cross-prediction values ​​to solve the objective function of each particle in the inner-layer particle swarm; S3c6: Repeat steps S3c3 to S3c5 until the convergence condition is met or the maximum number of iterations is reached, and output the solved system operating parameters. The calculation process of the inner-layer particle swarm objective function is as follows: Figure 5 As shown.

[0098] The search space boundary adjustment rules include: for chiller systems, adjusting the search direction of condenser inlet water temperature and condenser outlet water temperature according to the changing trend of cooling water pump speed ratio and cooling tower fan speed ratio; for ground source heat pump systems, adjusting the search direction of ground source outlet water temperature according to the changing trend of ground source circulation pump speed ratio.

[0099] In some embodiments, the boundary rules of the inner particle partial vector search space are adjusted as follows:

[0100] ① Increase and Decrease Search in the direction of increasing;

[0101] ② Increase and Increase Search in the direction of decreasing;

[0102] ③ Decrease and Decrease Search in the direction of increasing;

[0103] ④ Decrease and Increase Search in the direction of decreasing;

[0104] ④ Increase, under heat pump refrigeration conditions Search in the direction of decreasing, and search in the direction of increasing under heating conditions.

[0105] The total energy consumption of the system is the sum of the power of the main equipment in the cold and heat source system, including the main unit's cooling power, the main unit's heating power, the power of the chilled water pump or load water pump, the power of the cooling water pump or ground source circulation pump, and the power of the cooling tower. The calculation formula is as follows:

[0106] ;

[0107] in, This represents the total power of the system. Cooling power of the main unit; This refers to the power of the heat pump; Power of the chilled water pump (load pump); Power for cooling water pumps and ground source circulation pumps; This refers to the power of the cooling tower.

[0108] The decoupled solution calculation process for the objective function is as follows: Obtain the current particle position value; input the position value, along with condition variables, outer particle positions, and control parameters, into the coupled device model to calculate the predicted value; compare the assumed particle position value with the predicted value obtained through cross-prediction, and calculate the squared deviations of the chiller system and the ground source heat pump system respectively; add the squared deviations of the chiller system and the ground source heat pump system, and then take the square root to obtain the particle's objective function value. The calculation formula is as follows:

[0109] ;

[0110] in, The location value is the ratio of the condensate inlet water temperature to the ground source outlet water temperature for the i-th particle. The location value is the condenser outlet water temperature / ground source inlet water temperature of the i-th particle. For The condensate inlet temperature and ground source outlet temperature are predicted using the cooling tower / buried pipe model with conditional variables and external particle positions as inputs. For The condensate inlet temperature and ground source outlet temperature are predicted using a cooling tower / buried pipe model, with conditional variables and external particle positions as inputs.

[0111] Considering that the aforementioned rules for adjusting the inner particle partial vector search space boundary are empirical summaries based on the system's physical characteristics, and are essentially a set of static, heuristic rules based on prior knowledge, this invention upgrades the static rules into a dynamic, adaptive intelligent search strategy. This is done to further improve the rule adaptability in the event of system configuration changes, equipment aging, or operating environment deviations from design conditions, and to utilize real-time feedback information during the search process (such as convergence speed and the changing trend of prediction deviation) for self-optimization to improve search efficiency.

[0112] In some embodiments, the method for adjusting the search space boundary of the corresponding position vector of the inner particle swarm further includes the following steps: calling a pre-trained reinforcement learning agent, inputting a state feature vector composed of the comparison result between the position vector of the outer particle swarm and the position of the optimal solution in the previous optimization step, the current state of the inner particle swarm, and the system condition variables; having the reinforcement learning agent output the corresponding boundary adjustment action based on the state feature vector; and dynamically adjusting the search space boundary of the position vector of the inner particle swarm based on the boundary adjustment action.

[0113] The reinforcement learning agent outputs a corresponding boundary adjustment action based on the state feature vector, including the following steps: The policy network of the reinforcement learning agent performs multi-layer nonlinear transformations on the input state feature vector to output an action probability distribution. Each action in the action probability distribution corresponds to a predefined atomic operation that translates or scales the boundary of the search space. Based on the action probability distribution, a final boundary adjustment action is selected by sampling or selecting the maximum probability value. The boundary adjustment action is executed to generate a specific boundary adjustment instruction, which includes the direction and magnitude of the adjustment.

[0114] Pre-trained reinforcement learning agents can provide optimal instructions for adjusting the search space boundary based on real-time states during inner-layer particle swarm optimization. The training method involves repeatedly testing the agent in a simulated environment and learning from the results. The specific training process is as follows:

[0115] A high-fidelity simulation environment is constructed based on the entire cold and heat source system group model and the inner particle swarm optimization process. This environment provides an interactive digital twin platform for the agent. At the start of training, the core elements of the agent's interaction with the environment need to be precisely defined: the state space consists of multi-dimensional feature vectors such as the outer particle positions, inner particle swarm states, and system variables; the action space is designed as a set of instructions for fine-tuning the boundaries of the search space, such as discrete or continuous operations like increasing or scaling specific parameter boundaries; and the reward function is carefully designed as a guide for the agent's learning direction. It is a composite function that integrates the descent of the decoupled evaluation objective function, the convergence flag, and the search space size penalty term, aiming to simultaneously incentivize convergence speed, accuracy, and search efficiency.

[0116] The training employs advanced policy gradient algorithms, such as PPO or SAC, as the core learning engine. The mechanism involves the agent generating a large number of experience tuples—state-action-reward-new state—through trial and error in a simulated environment, storing these tuples in an experience replay buffer. The algorithm periodically samples small batches of historical experience data, continuously evaluating the value of actions and updating the policy network parameters. Its ultimate goal is to enable the agent to learn to select the action that maximizes cumulative reward in a specific state, i.e., the optimal boundary adjustment policy. The entire offline training process requires millions of iterative runs on system data covering massive amounts of different operating conditions. Through this large-scale trial and error and parameter updates, the agent ultimately internalizes an efficient navigation strategy capable of handling complex and ever-changing operating conditions.

[0117] By employing the aforementioned pre-trained reinforcement learning agent to dynamically adjust the search space boundary of the inner particle swarm position vector, the decoupling accuracy and convergence speed are significantly improved. On one hand, the reinforcement learning agent can dynamically and accurately shrink the search space, avoiding particles searching in invalid regions, thereby greatly accelerating convergence and reducing the number of inner iterations compared to the original static rule method. Simultaneously, it can obtain a better objective function solution, improving the accuracy and reliability of the overall energy efficiency optimization effect. On the other hand, it enhances the system's adaptability and robustness because this method no longer relies on fixed prior rules but can autonomously learn the optimal search strategy under different systems, operating conditions, and equipment performance degradation. Therefore, this energy-saving management method can be adaptively applied to various building heating and cooling source systems, such as those in commercial buildings and data centers, and can even maintain high efficiency through fine-tuning after system component replacement or aging. Finally, this method forms a continuously evolving intelligent control system with sustainable learning capabilities. As operating time accumulates, the reinforcement learning agent's strategy becomes increasingly efficient, forming an intelligent energy-saving ecosystem with self-optimization capabilities.

[0118] Example 2:

[0119] This embodiment provides an energy-saving management and control system for a cold and heat source system based on particle swarm optimization and nested decoupling. This system is used to implement the energy-saving management and control method for a cold and heat source system based on particle swarm optimization and nested decoupling as described in Embodiment 1. It includes an equipment group module, a system group module, and a parameter analysis and decision module. Specifically: the equipment group module is used to establish an equipment group model of the building's cold and heat source system. The equipment group model includes chillers and ground source heat pumps, ground source heat pumps, chilled water pumps, cooling water pumps and ground source circulating pumps, a cooling tower model, and buried pipes. The system group module is used to map and match the associated coupling variables with the same physical meaning among the equipment based on the equipment group model, establishing a cold and heat source system group model. The system group model includes a chiller system and a ground source heat pump system. The parameter analysis and decision module is used to optimize and solve the system's global control parameters using a particle swarm optimization algorithm based on the cold and heat source system group model, outputting the optimal values ​​of the system's global control parameters.

[0120] Working Principle: This invention provides an energy-saving control method and system for cold and heat source systems based on particle swarm optimization (PSO) nested decoupling. Based on a single-device data-driven model, it maps and matches the associated coupling variables with the same physical meaning among the devices. The system swarm model variables are defined as system input variables, internal associated coupling variables, and system output variables. A PSO nested control parameter optimization algorithm is proposed. The outer PSO optimization process is used to solve for the optimal solution of the system's global control parameters, while the inner PSO optimization process is used to decouple and solve the internal coupling variables. After updating the velocity and position, the outer PSO process calls the inner PSO decoupling solution process. The objective function (total system energy consumption) is calculated using the system state parameters output by the inner particle swarm optimization, and the optimal values ​​of the global control parameters are finally output. The inner particle swarm optimization process of the internal coupling variables of the cold and heat source system is solved, with the particle position vector as the coupling parameter. The coupling parameter can be accurately and quickly solved by adjusting the search space boundary and solving the evaluation objective function. Furthermore, by using a pre-trained reinforcement learning agent to dynamically adjust the search space boundary of the inner particle swarm position vector, the decoupling accuracy and convergence speed are significantly improved, the system's adaptability and robustness are enhanced, and a continuously evolving intelligent control system is formed.

[0121] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0122] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0123] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0124] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0125] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An energy-saving control method for cold and heat source systems based on particle swarm optimization and nested decoupling, characterized in that, Includes the following steps: S1. Establish a model of the equipment group of the building's cold and heat source system. The equipment group model includes a chiller unit, a ground source heat pump, a chilled water pump, a ground source circulating pump, a cooling tower model, and buried pipes. S2. Based on the equipment group model, map and match the associated coupling variables with the same physical meaning among the equipment to establish a cold and heat source system group model, which includes a chiller system and a ground source heat pump system. S3. Based on the cold and heat source system group model, the particle swarm optimization algorithm is used to optimize and solve the global control parameters of the system, and the optimal value of the global control parameters of the system is output. In step S3, the particle swarm optimization algorithm includes an outer particle swarm optimization process and an inner particle swarm optimization process; The outer particle swarm optimization process is used to find the optimal values ​​of the global system control parameters that minimize the total energy consumption of the system. The inner-layer particle swarm optimization process is used to decouple and solve the internal coupling variables of the system for the control parameters given by the outer-layer particle swarm. The outer particle swarm optimization process includes the following steps: S3a. Input system condition variables, initialize the position vector and velocity vector of the outer particle swarm, and determine the search range and boundary conditions of the position vector; S3b, Update the velocity vector and position vector of the outer particle swarm; S3c. For each particle at the updated position, call the inner particle swarm decoupling solution process to solve the system state parameters of each particle after the outer particle swarm updates its position. S3d: Calculate the objective function for the updated positions of outer particles using the system swarm model and the system state parameters output by the inner particle swarm, and update the individual extreme values ​​and the global optimal solution; S3e, Repeat steps S3b to S3d until the convergence condition is met or the maximum number of iterations is reached, and output the optimal value of the global control parameter; The inner-layer particle swarm optimization process includes the following steps: S3c1. Input the system condition variables and the position vector of the outer particle swarm, initialize the position vector and velocity vector of the inner particle swarm, the position vector of the inner particle swarm is a coupling parameter, including the condenser inlet water temperature and condenser outlet water temperature, the ground source outlet water temperature and the ground source inlet water temperature; S3c2. Compare the input position vector of the outer particle swarm with the position of the optimal solution of the outer particle in the previous optimization step to obtain the comparison result. Based on the comparison result, query the preset search space boundary adjustment rules to determine the search direction adjustment strategy of the inner particle swarm coupling parameters. Adjust the search space boundary of the corresponding position vector of the inner particle swarm according to the adjustment strategy. S3c3, Perform inner-layer particle swarm search and update the position vector and velocity vector of the inner-layer particle swarm; S3c4. Input the updated inner particle position vector into the coupled device model, perform cross-prediction, and obtain the cross-prediction value. S3c5. Establish a decoupled solution evaluation objective function, and use the current position of the inner particle and the cross prediction value to solve the objective function of each particle in the inner particle swarm. S3c6. Repeat steps S3c3 to S3c5 until the convergence condition is met or the maximum number of iterations is reached, and output the solved system state parameters. The preset search space boundary adjustment rules include: For the chiller system, the search direction of the condenser inlet water temperature and condenser outlet water temperature is adjusted according to the changing trend of the cooling water pump speed ratio and the cooling tower fan speed ratio. For ground source heat pump systems, the search direction for ground source outlet water temperature is adjusted according to the changing trend of the ground source circulating pump speed ratio. The outer particle position vector includes the rated speed ratio of chilled water pumps or load water pumps, the rated speed ratio of cooling water pumps or ground source circulation pumps, the rated speed ratio of cooling tower fans, and the set value of the outlet water temperature of refrigeration or heating equipment; the system condition variables include the unit load rate, the system load rate, the outdoor wet-bulb temperature / ground source well soil temperature, the number of chilled water pumps or load water pumps, and the number of cooling water pumps or ground source circulation pumps.

2. The energy-saving control method for cold and heat source systems based on particle swarm optimization and nested decoupling according to claim 1, characterized in that, The total energy consumption of the system is the sum of the power of the main equipment in the cold and heat source system, including the main unit's cooling power, the main unit's heating power, the power of the chilled water pump or load water pump, the power of the cooling water pump or ground source circulation pump, and the power of the cooling tower.

3. The energy-saving control method for cold and heat source systems based on particle swarm optimization and nested decoupling according to claim 1, characterized in that, The calculation process of the decoupled solution to the evaluation objective function is as follows: Get the current particle position value; The position value, along with the condition variables, the outer particle position, and the control parameters, is input into another device model coupled with it to calculate the predicted value. The position values ​​of the particle hypothesis are compared with the predicted values ​​obtained through cross-prediction, and the squared deviations of the chiller system and the ground source heat pump system are calculated respectively. The squared deviations of the chiller system and the ground source heat pump system are added together, and then the square root is taken to obtain the objective function value of the particle.

4. The energy-saving control method for cold and heat source systems based on particle swarm optimization and nested decoupling according to claim 1, characterized in that, The method for adjusting the search space boundary of the corresponding position vector of the inner particle swarm further includes the following steps: The pre-trained reinforcement learning agent is invoked, and the state feature vector is composed of the comparison result between the position vector of the outer particle swarm and the position of the optimal solution in the previous optimization step, the current state of the inner particle swarm, and the system condition variables. This vector is then input into the reinforcement learning agent. The reinforcement learning agent outputs the corresponding boundary adjustment action based on the state feature vector; Based on the boundary adjustment action, the search space boundary of the inner particle swarm position vector is dynamically adjusted.

5. The energy-saving control method for cold and heat source systems based on particle swarm optimization and nested decoupling according to claim 4, characterized in that, The step of the reinforcement learning agent outputting the corresponding boundary adjustment action based on the state feature vector includes the following steps: The policy network of the reinforcement learning agent performs multi-layer nonlinear transformation on the input state feature vector to output an action probability distribution. Each action in the action probability distribution corresponds to a predefined atomic operation that translates or scales the search space boundary. Based on the action probability distribution, a final boundary is selected to adjust the action by sampling or selecting the maximum probability value; The boundary adjustment action is executed to generate specific boundary adjustment instructions, which include the direction and magnitude of the adjustment.

6. An energy-saving control system for cold and heat source systems based on particle swarm optimization and nested decoupling, characterized in that, This system is used to implement the energy-saving control method for cold and heat source systems based on particle swarm optimization and nested decoupling as described in any one of claims 1-5, comprising: The equipment group module is used to establish an equipment group model of the building's cold and heat source system. The equipment group model includes chiller units and ground source heat pumps, ground source heat pumps, chilled water pumps, cooling water pumps and ground source circulating pumps, cooling tower models, and underground pipes. The system group module is used to map and match the associated coupling variables with the same physical meaning among the devices based on the device group model, and to establish a cold and heat source system group model, which includes a chiller system and a ground source heat pump system. The parameter analysis and decision module is used to optimize and solve the global control parameters of the system based on the cold and heat source system group model using a particle swarm optimization algorithm, and output the optimal values ​​of the global control parameters of the system.

Images