Terminal model of chilled water supply temperature optimization method and system

Abstract

The application provides a chilled water supply temperature optimization method and system of a terminal model, comprising the following steps: S1, a mechanism model of terminal equipment of an air conditioning system is built; S2, historical data is processed by clustering, and then parameters of the mechanism model are identified; and S3, the mechanism model is controlled by PID, and an upper limit of a chilled water demand temperature and a supply temperature of the terminal equipment of the air conditioning system is obtained as a boundary condition of an optimization task. The application improves the modeling accuracy, robustness and safety of the terminal model, and provides a new solution for chilled water supply temperature optimization in chilled station energy efficiency optimization.

Description

Technical Field

[0001] This invention belongs to the field of energy and energy-saving technology for HVAC systems, specifically relating to a method and system for optimizing chilled water supply temperature using a terminal model. More specifically, it is a method for optimizing chilled water supply temperature using a terminal model based on mechanism and data fusion, applicable to the optimization and energy-saving control of chiller supply temperature in large industrial and commercial buildings and factories. Background Technology

[0002] The upper limit of chilled water supply temperature is one of the core boundary conditions for energy efficiency optimization of chiller plants. Its value will directly affect the heat and humidity treatment capacity of terminal equipment, the operating efficiency of chiller units, and the energy consumption of water pumps.

[0003] Traditional chiller plant optimization typically relies on manually setting the upper limit of chilled water supply temperature for chillers based on engineering experience, neglecting changing climatic conditions and the actual operating status of terminal equipment. In some methods that have incorporated terminal equipment models into the optimization considerations, issues exist such as oversimplification of terminal modeling methods, insufficient multiphysics coupling, inadequate model robustness, and limited parameter identification methods. These problems may prevent the terminal model from accurately simulating various operating conditions and lack the establishment of safety boundaries based on equipment parameters with practical physical meaning.

[0004] Therefore, how to combine equipment mechanisms and historical data to improve the accuracy, robustness, and safety of the model while ensuring the model simulation efficiency and engineering response requirements is a challenge currently faced in optimizing chilled water supply temperature for energy efficiency optimization of chilled plants.

[0005] Patent document CN115688479A discloses a data-driven method for optimizing energy saving in chilled water circulation. The method includes: preprocessing historical production data, using a support vector regression algorithm to calculate a chilled water circulation energy consumption model; using a genetic algorithm to encode the start-up and shutdown of the chiller unit, the start-up and shutdown frequency of the chilled water pump, and the chilled water outlet temperature of the chiller unit; using the predicted value function of the chilled water circulation energy consumption model as the fitness function; using the chilled water main supply temperature as the limiting condition; and performing iterative cycles of genetic operations to obtain the minimum energy consumption, the corresponding equipment start-up and shutdown combinations, and the set value of the chilled water outlet temperature of the chiller unit.

[0006] However, this scheme cannot improve the global search capability of parameter identification and cannot avoid the local optimum problem of traditional genetic algorithms. This problem urgently needs to be solved. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for optimizing chilled water supply temperature in a terminal model.

[0008] A method for optimizing chilled water supply temperature in a terminal model according to the present invention includes:

[0009] Step S1: Build a mechanistic model of the terminal equipment of the air conditioning system;

[0010] Step S2: Cluster the historical data to identify the parameters of the mechanism model;

[0011] Step S3: By using the PID control mechanism model, the upper limits of the chilled water demand temperature and supply temperature of the air conditioning system terminal equipment are obtained as boundary conditions for the optimization task.

[0012] Preferably, in step S1, the mechanism model can simulate the operation of the terminal equipment of the air conditioning system under different operating conditions;

[0013] The input variables of the mechanism model include: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, fan operating frequency, humidifier on / off status, and fan operating status.

[0014] Preferably, step S2 includes:

[0015] Step S2.1: According to the preset task objective, remove abnormal and missing data from the historical data, and then construct a feature dataset based on the historical data;

[0016] Step S2.2: Use the K-means++ algorithm to cluster the feature dataset and generate typical working conditions;

[0017] Step S2.3: Based on typical working conditions, obtain the parameter identification results of the mechanism model through a genetic algorithm.

[0018] Preferably, in step S2.1, the feature dataset includes: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, hot water supply temperature, hot water supply pressure, air conditioning water valve opening percentage, air supply temperature and air supply dew point temperature.

[0019] In step S2.2, the K-means++ algorithm is used to perform a traversal search within a preset range of cluster numbers, and then the optimal number of clusters is determined by the silhouette coefficient method. Then, the P samples closest to the cluster center in each cluster are selected as typical working cases, where the value of P ranges from 10 to 100. The optimal number of clusters is the number of clusters corresponding to the maximum silhouette coefficient.

[0020] The mathematical expression for the contour coefficient method is:

[0021]

[0022] in, For the first Profile coefficients for each data point For the first The average distance between a data point and other data points in its cluster Indicates the first The average distance between each data point and all data points in its nearest neighbor cluster;

[0023] In step S2.3, the parameter identification results of the mechanistic model are obtained through a genetic algorithm, including:

[0024] Step S2.3.1: Generate 50×3 individuals and divide them into 3 initial subpopulations, each containing 50 individuals;

[0025] Step S2.3.2: Use tournament selection. Randomly select 10 individuals from the subpopulation each time and let them compete until the number of individuals preserved to the next generation reaches 25.

[0026] Step S2.3.3: Based on the preset probability values ​​and proportion values, crossover and mutate individuals within each subpopulation;

[0027] Step S2.3.4: Every 10 generations, the best individual in the current subpopulation is migrated to an adjacent subpopulation using a circular migration strategy, replacing the worst individual in the adjacent subpopulation. After each migration, the top 20% of individuals with the best fitness are selected from each subpopulation and stored in the elite population. It is determined whether the globally optimal individual in the elite population has been maintained for more than 10 generations. If yes, the globally optimal individual is output as a parameter of the mechanism model. If no, step S2.3.2 is executed again.

[0028] The mathematical expression for fitness is:

[0029]

[0030] in, For the sample size, For prediction data, i.e. simulation results, This refers to real data, i.e., observational data.

[0031] Preferably, in step S3, the required temperature of the chilled water is set to T. sup_demand The mathematical expressions for the upper limit of the temperature under summer and transitional season conditions, from top to bottom, are as follows:

[0032] T sup_uplim =T sup_demand +0.5℃

[0033] T sup_uplim =T sup_demand +0.3℃

[0034] Among them, T sup_uplim Indicates the upper limit of temperature;

[0035] The mathematical expression for the mechanism model based on PID control is:

[0036]

[0037] in, , , In order, they are proportional coefficient, integral coefficient, and differential coefficient; for The time-based setpoint, i.e., the target supply air temperature and humidity in this scenario, and the process value, i.e., the difference between the supply air temperature and humidity calculated by the model in this scenario, are used. The control objective of PID is to gradually reduce the temperature and humidity through feedback control. Adjust to 0; for The control variable at any given time is the water supply temperature in this scenario. Represents auxiliary variables within the integral; To represent the differential; This represents the integral.

[0038] A chilled water supply temperature optimization system for a terminal model according to the present invention includes:

[0039] Module M1: Constructing a mechanistic model of the terminal equipment in the air conditioning system;

[0040] Module M2: Clusters historical data to identify the parameters of the mechanistic model;

[0041] Module M3: By using the PID control mechanism model, the upper limits of the chilled water demand temperature and supply temperature of the terminal equipment of the air conditioning system are obtained as boundary conditions for the optimization task.

[0042] Preferably, in module M1, the mechanism model can simulate the operation of the terminal equipment of the air conditioning system under different operating conditions;

[0043] The input variables of the mechanism model include: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, fan operating frequency, humidifier on / off status, and fan operating status.

[0044] Preferably, module M2 includes:

[0045] Module M2.1: Based on the preset task objectives, it removes abnormal and missing data from historical data, and then constructs a feature dataset based on the historical data;

[0046] Module M2.2: The feature dataset is clustered using the K-means++ algorithm to generate typical working conditions;

[0047] Module M2.3: Based on typical working conditions, the parameter identification results of the mechanism model are obtained through genetic algorithm.

[0048] Preferably, in module M2.1, the feature dataset includes: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, hot water supply temperature, hot water supply pressure, air conditioning water valve opening percentage, supply air temperature and supply air dew point temperature;

[0049] In module M2.2, the K-means++ algorithm is used to perform a traversal search within a preset range of cluster numbers, and then the optimal number of clusters is determined by the silhouette coefficient method. Then, the P samples closest to the cluster center in each cluster are selected as typical working conditions, where the value of P ranges from 10 to 100; wherein, the optimal number of clusters is the number of clusters corresponding to the maximum silhouette coefficient.

[0050] The mathematical expression for the contour coefficient method is:

[0051]

[0052] in, For the first Profile coefficients for each data point For the first The average distance between a data point and other data points in its cluster Indicates the first The average distance between each data point and all data points in its nearest neighbor cluster;

[0053] In module M2.3, the parameter identification results of the mechanistic model are obtained through a genetic algorithm, including:

[0054] Module M2.3.1: Generates 50×3 individuals, divided into 3 initial subpopulations, each containing 50 individuals;

[0055] Module M2.3.2: Uses tournament selection, randomly selecting 10 individuals from the subpopulation each time to compete, until the number of individuals preserved to the next generation reaches 25;

[0056] Module M2.3.3: Based on preset probability values ​​and proportion values, crossover and mutation are performed on individuals within each subpopulation;

[0057] Module M2.3.4: Every 10 generations, the best individual in the current subpopulation is migrated to an adjacent subpopulation through a circular migration strategy, replacing the worst individual in the adjacent subpopulation. After each migration, the top 20% of individuals with the best fitness are selected from each subpopulation and stored in the elite population. It is determined whether the globally best individual in the elite population has been maintained for more than 10 generations. If the result is yes, the globally best individual is output as a parameter of the mechanism model. If the result is no, module M2.3.2 is triggered again.

[0058] The mathematical expression for fitness is:

[0059]

[0060] in, For the sample size, For prediction data, i.e. simulation results, This refers to real data, i.e., observational data.

[0061] Preferably, in module M3, the required temperature of the chilled water is set to T. sup_demand The mathematical expressions for the upper limit of the temperature under summer and transitional season conditions, from top to bottom, are as follows:

[0062] T sup_uplim =T sup_demand +0.5℃

[0063] T sup_uplim =T sup_demand +0.3℃

[0064] Among them, T sup_uplim Indicates the upper limit of temperature;

[0065] The mathematical expression for the mechanism model based on PID control is:

[0066]

[0067] in, , , In order, they are proportional coefficient, integral coefficient, and differential coefficient; for The time-based setpoint, i.e., the target supply air temperature and humidity in this scenario, and the process value, i.e., the difference between the supply air temperature and humidity calculated by the model in this scenario, are used. The control objective of PID is to gradually reduce the temperature and humidity through feedback control. Adjust to 0; for The control variable at any given time is the water supply temperature in this scenario. Represents auxiliary variables within the integral; To represent the differential; This represents the integral.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] 1. This invention achieves data collaboration between the Modelica mechanism model and the Python genetic algorithm through cross-platform model interaction technology, and establishes a terminal device model that integrates mechanism and data. While ensuring computational efficiency, it improves the robustness and security of the terminal device model.

[0070] 2. This invention incorporates end-point demand and equipment models into the boundary of chiller plant optimization, and proposes an end-point modeling method that integrates mechanisms and data, which improves the modeling accuracy, robustness and safety of the end-point model, and provides a new solution for optimizing chilled water supply temperature in chiller plant energy efficiency optimization.

[0071] 3. Based on the terminal equipment model, this invention uses an ideal PID control model and valve opening threshold to reverse simulate and calculate the chilled water supply temperature requirements of the terminal equipment, thereby deriving the theoretical upper limit of the chilled water supply temperature and optimizing the traditional method of setting the upper limit of the chilled water supply temperature in chilled plants. Attached Figure Description

[0072] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0073] Figure 1 This is a flowchart provided for the present invention;

[0074] Figure 2 This is a schematic diagram illustrating the process of collaborating the Modelica mechanism model with historical data provided by the present invention.

[0075] Figure 3 This is a schematic diagram of the multi-population genetic algorithm provided by the present invention;

[0076] Figure 4 The logic diagram for reverse simulation calculation of chilled water supply temperature provided by this invention. Detailed Implementation

[0077] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0078] This invention provides a method for optimizing chilled water supply temperature based on a terminal model that integrates mechanism and data fusion. First, a mechanism model of the terminal equipment in the air conditioning system is built in Modelica. Then, based on the historical operating data of the equipment after cleaning and clustering, an improved genetic algorithm is applied to identify the parameters of the equipment mechanism model. Based on the identified model parameters, a terminal equipment mechanism model with a PID controller is built, and the terminal water supply temperature that meets the setpoints for supply air temperature and supply air dew point temperature under the current outdoor temperature and humidity conditions is calculated. Based on this temperature, the upper limit of the chilled water supply temperature is calculated seasonally as the boundary condition for chilled water station optimization.

[0079] In other words, this invention uses a multi-population genetic algorithm to identify parameters of the end-device mechanism model. Through five processes—natural selection, crossover, mutation, migration, and artificial selection—it improves the global search capability of parameter identification and avoids the local optimum defect of traditional genetic algorithms.

[0080] The present invention provides a method for optimizing chilled water supply temperature based on a terminal model using mechanism and data fusion, comprising:

[0081] Step 1: Build a mechanistic model of the terminal equipment of the air conditioning system;

[0082] Step 2: After cleaning and clustering the historical data, an improved genetic algorithm is applied to identify the parameters of the end-device mechanism model.

[0083] Step 3: Based on the terminal equipment mechanism model after parameter identification, add ideal PID control, calculate the required chilled water temperature of the terminal equipment to meet the supply air temperature and humidity requirements of the terminal equipment, and calculate the upper limit of the chilled water supply temperature of the chiller plant based on this temperature, as the boundary condition for chiller plant optimization.

[0084] Specifically, the mechanism model of the air conditioning system terminal equipment in step 1 is built using Open Modelica software based on the Modelica language;

[0085] Specifically, the mechanism model of the air conditioning system terminal equipment in step 1 includes cold coils, hot coils, humidifiers, fans, water valves and pipeline pressure drop components, which can simulate the operation of the air conditioning system terminal equipment under different working conditions.

[0086] Specifically, the input variables of the air conditioning system terminal equipment mechanism model in step 1 include the main chilled water supply temperature, hot water supply temperature, valve opening degree, outdoor ambient temperature, outdoor ambient humidity, fan operating frequency, humidifier on / off status, and fan operating status.

[0087] Specifically, in step 2, the historical data required for identifying the mechanism model parameters of the terminal equipment includes: outdoor ambient temperature, outdoor ambient humidity, chilled water supply temperature, chilled water supply pressure, hot water supply temperature, hot water supply pressure, air conditioning operating status, air conditioning water valve opening percentage, fan operating frequency, humidifier on / off status, fan operating status, supply air temperature, and supply air dew point temperature.

[0088] Specifically, in step 2, the K-Means++ method is used to perform clustering processing on the cleaned historical data;

[0089] Specifically, in step 2, for the mechanism model of the terminal equipment of the air conditioning system, the parameters to be identified are the rated air volume, the rated water flow of the coil, the heat exchange capacity UA of the coil under the rated flow, the rated humidification capacity, and the rated pressure drop of the pipeline.

[0090] Specifically, in step 2, a cross-platform interface between Python and Modelica is established through the OMPython library in the Python environment. This is manifested in the following ways: OMPython is used to load and compile the Modelica terminal device mechanism model to achieve dynamic instantiation of the model. The mechanism model is then driven to perform dynamic simulation calculations through the running interface. This seamlessly embeds the Modelica-based air conditioning system terminal device mechanism model into the Python-based improved genetic algorithm framework. This allows the improved genetic algorithm to obtain the device state parameters output by the mechanism model in real time during the iterative optimization process, and dynamically feeds the offspring generated by the algorithm back to the mechanism model for verification, forming a closed-loop collaborative optimization between the mechanism model and the data model.

[0091] Specifically, the improved genetic algorithm in step 2 is a multi-population genetic algorithm, which breaks through the framework of traditional genetic algorithms that rely solely on a single population for evolution, and is beneficial for escaping local optima. Its specific implementation process is as follows:

[0092] Define the fitness evaluation function: Drive the Modelica end-device mechanism model to perform dynamic simulation calculations in the Python environment, and define the mean absolute percentage error (MAPE) between the simulation results and the observed data as the fitness evaluation function; 2. Initialize multiple populations: Based on the parameters to be identified in the end-device mechanism model and the identification range of each parameter, randomly generate M×N individuals to form N initial subpopulations, N≥3, and each initial subpopulation contains M individuals;

[0093] 3. Selection operation: Tournament selection is adopted within each subpopulation. A certain number of individuals are randomly selected from the subpopulation each time to compete. The individual with the highest fitness is saved to the next generation. This process is repeated until the number of individuals saved to the next generation reaches the preset number.

[0094] In the tournament selection mechanism of the genetic algorithm, the "highest fitness" is a relative evaluation standard based on the current state of the population, rather than a preset absolute numerical threshold.

[0095] Since the individuals in the initial population are randomly generated, pre-set optimization objectives, such as MAPE < 10%, may not be achievable for the initial population or some offspring. Therefore, forcibly setting a numerical threshold may cause the algorithm to terminate due to the inability to select effective individuals.

[0096] 4. Crossover operation: Apply two-point crossover within each subpopulation. Randomly select two individuals and randomly set a crossover start point and a crossover end point in each individual. Exchange genes between the crossover points to generate new offspring individuals.

[0097] 5. Mutation operation: Uniform mutation is applied within each subpopulation. For each gene locus of each individual in the subpopulation, the mutation operation is independently determined with the same probability. If the determination is true, the value of the gene locus is completely replaced with a uniform random number within the corresponding identification range.

[0098] 6. Migration Operation: Every K generations, the best individual in the current subpopulation is migrated to the neighboring subpopulation through a circular migration strategy, replacing the worst individual in the neighboring subpopulation.

[0099] 7. Elite Population Update: After each migration, select the top X% of individuals with the best fitness from each subpopulation and store them in the elite population. When the globally optimal individual in the elite population remains for more than a preset threshold number of generations, the algorithm is considered to have converged and training is stopped.

[0100] Specifically, the chilled water demand temperature of the terminal equipment in step 3 is based on the Modelica terminal equipment mechanism model after parameter identification in step 2. With the supply air temperature and supply air dew point temperature set as the target and 90% water valve opening as the boundary condition, the chilled water supply temperature is adjusted through an ideal PID control model so that the chilled water supply temperature of the terminal equipment is the simulated output air temperature and humidity with a dynamic error of ≤2% compared with the set value under the current outdoor temperature and humidity.

[0101] Specifically, the upper limit T of the chilled water supply temperature in step 3. sup_uplim The calculation is as follows:

[0102] Summer operating conditions: T sup_uplim = T sup_demand +0.5℃;

[0103] Transitional Season Operating Conditions: T sup_uplim = T sup_demand +0.3℃.

[0104] Among them, T sup_demand The chilled water supply temperature at the terminal is required to meet the requirements of the supply air temperature and supply air dew point temperature of the terminal equipment.

[0105] In other words, the method for dynamically optimizing chilled water supply temperature based on a terminal model fusion of mechanism and data, provided by the present invention, includes:

[0106] Step 1: Build a mechanistic model of the terminal equipment of the air conditioning system;

[0107] Step 2: After cleaning and clustering the historical data, an improved genetic algorithm is applied to identify the parameters of the end-device mechanism model.

[0108] Step 3: Based on the completed mechanism model of the terminal equipment with parameter identification, add ideal PID control, calculate the required chilled water temperature of the terminal equipment to meet the supply air temperature and humidity requirements of the terminal equipment, and calculate the upper limit of the chilled water supply temperature of the chiller plant based on this temperature, as the boundary condition for the optimization of the chiller plant.

[0109] Specifically, in step 1, the mechanism model of the terminal equipment of the air conditioning system is built using Open Modelica software based on the Modelica language. The model includes cold coils, hot coils, humidifiers, fans, water valves, and pipeline pressure drop components.

[0110] Specifically, historical data is collected to identify the parameters of the terminal equipment mechanism model. The required data includes: outdoor ambient temperature, outdoor ambient humidity, chilled water supply temperature, chilled water supply pressure, hot water supply temperature, hot water supply pressure, air conditioning operating status, air conditioning water valve opening percentage, fan operating frequency, humidifier on / off status, fan operating status, supply air temperature and supply air dew point temperature. The K-Means++ method is then used to cluster the cleaned historical data.

[0111] The parameters of the terminal equipment mechanism model that need to be identified in the air conditioning system are rated air volume, rated water flow rate of the coil, heat exchange capacity of the coil UA at the rated flow rate, rated humidification capacity, and rated pressure drop of the pipeline.

[0112] Specifically, in step 2, a cross-platform interface between Python and Modelica is established through the OMPython library in the Python environment. OMPython is used to load and compile the Modelica terminal device mechanism model to achieve dynamic instantiation of the model. The mechanism model is then driven to perform dynamic simulation calculations through the running interface. This seamlessly embeds the Modelica-based air conditioning system terminal device mechanism model into the Python-based improved genetic algorithm framework. This allows the improved genetic algorithm to obtain the device state parameters output by the mechanism model in real time during the iterative optimization process, and dynamically feeds the offspring generated by the algorithm back to the mechanism model for verification, forming a closed-loop collaborative optimization between the mechanism model and the data model.

[0113] Genetic algorithms are essentially iterative processes of random search, which cannot be fully described by a single closed mathematical definition. Improved genetic algorithms typically include hierarchical genetic algorithms, adaptive genetic algorithms, and multi-population genetic algorithms. The improved genetic algorithm used in this embodiment is a multi-population genetic algorithm. Based on traditional genetic algorithms, it maintains population diversity and avoids getting trapped in local optima through multi-population parallel evolution and migration strategies.

[0114] Specifically, the improved genetic algorithm used in this scheme is a multi-population genetic algorithm. Based on the traditional genetic algorithm, it maintains the diversity of the population and avoids getting trapped in local optima through multi-population parallel evolution and migration strategies.

[0115] The method for determining the chilled water supply temperature requirement of the terminal equipment is as follows: using the Modelica terminal equipment mechanism model after parameter identification, with the supply air temperature and supply air dew point temperature setpoints as targets and a water valve opening of 90% as boundary conditions, the chilled water supply temperature is adjusted through an ideal PID control model. When the dynamic relative error between the simulated air temperature and humidity output of the model under the current outdoor temperature and humidity and the setpoint is ≤2%, the corresponding chilled water temperature is the target supply water temperature.

[0116] like Figure 1 As shown in the embodiment, the chilled water supply temperature optimization method based on a terminal model fusion of mechanism and data provided by the present invention includes the following steps:

[0117] Step 1: Build a mechanistic model of the air conditioning system's terminal equipment:

[0118] The mechanism model of the terminal device was built using Open Modelica software based on the Modelica language. The model components mainly come from the Buildings library in Modelica, including cold coils, hot coils, humidifiers, fans, water valves, and pipe pressure drop components. The model components were connected in the same order as the actual device components to construct a multiphysics coupled terminal device mechanism model.

[0119] In some embodiments, the terminal equipment of the air conditioning system adopts a six-pipe fresh air handling unit, namely a Make-up Air Unit (MAU). Its summer air handling process is as follows: fresh air flows sequentially through a pre-cooling coil (design supply / return water temperature 12℃ / 18℃), a cooling coil (design supply / return water temperature 6℃ / 12℃), a humidifier, and a reheat coil (design supply / return water temperature 37℃ / 29℃) for processing. The pre-cooling coil bears the sensible heat load and part of the latent heat load, the cooling coil bears the main latent heat load, and the reheat coil regulates the air temperature according to the set supply air temperature. Finally, a variable frequency fan delivers the processed air to the terminal rooms at the set temperature and humidity.

[0120] In the specific implementation process, the terminal equipment mechanism model is built according to the air handling process of the aforementioned six-control MAU.

[0121] The cold coil uses the discretized numerical wet coil model WetCoilCounterFlow from the Buildings library to simulate the water-air heat transfer process with condensate. This method can calculate the heat and mass transfer of each discrete coil unit separately. For sensible heat transfer, the heat transfer calculation is as follows:

[0122]

[0123] in, For heat transfer rate, The thermal capacity flow rate of water, The inlet temperature of the water. This refers to the outlet temperature of the water; the same applies to the air side. Indicates the inlet temperature of the air; Indicates the outlet temperature of the air; This indicates the heat capacity flow rate of air;

[0124] The thermal flow rate of the water is expressed mathematically as follows:

[0125] =

[0126] in, The mass flow rate of water; This is the specific heat capacity of water at constant pressure.

[0127] Heat exchanger efficiency Defined as:

[0128]

[0129] in, To determine the theoretical maximum heat exchange of the coil, use calculate, For the fluid on both sides The smaller one. Refers to the aforementioned heat capacity flow rate.

[0130] and, It can also be expressed as the number of heat transfer units, i.e. Dimensionless heat capacity flow rate ratio and coil heat exchange method, i.e. The function, that is:

[0131]

[0132] in, This equation represents the number of heat transfer units; it also represents the function. It can be expressed as , and The function.

[0133] The number of heat transfer units, i.e. For dimensionless numbers, the calculation is as follows:

[0134]

[0135] in, The average heat transfer coefficient of the heat exchanger. This refers to the heat exchange area.

[0136] Generally speaking, It can be calculated from the operating parameters under the design conditions, but due to the actual operating conditions... The actual operating conditions may differ significantly from the design conditions; therefore, during parameter identification, the parameters of each coil will be considered. Values ​​are identified synchronously.

[0137] The input variables of the model include the main chilled water supply temperature, hot water supply temperature, valve opening degree, outdoor ambient temperature, outdoor ambient humidity, fan operating frequency, humidifier on / off status, and fan operating status.

[0138] During the model encapsulation stage, an explicit declaration method is used to define the mechanistic parameters of the terminal equipment, namely rated air volume, rated water flow of the coil, coil heat exchange capacity UA at rated flow, rated humidification capacity, and rated pressure drop of the pipeline, as adjustable parameters to support external interface access and realize parameter identification based on the improved genetic algorithm.

[0139] Step 2: After clustering the historical data, an improved genetic algorithm is applied to identify the parameters of the end-device mechanism model.

[0140] After the model parameters are set, a cross-platform interface between Python and Modelica is established using the OMPython library in the Python environment based on the FMI2.0 protocol, so as to realize the data interaction between external data and the mechanistic model.

[0141] Specifically, by instantiating the Modelica System class in OMPython, importing the pre-packaged Modelica end-device model, and performing model compilation, simulation solution settings, parameter passing, and simulation calculations, the calculation results will be stored in a .mat file via OMPython for subsequent reading and processing.

[0142] The historical data required for model parameter identification includes: outdoor ambient temperature, outdoor ambient humidity, chilled water supply temperature, chilled water supply pressure, hot water supply temperature, hot water supply pressure, air conditioning operating status, air conditioning water valve opening percentage, fan operating frequency, humidifier on / off status, fan operating status, supply air temperature, and supply air dew point temperature.

[0143] The steps for cleaning historical data include: identifying and correcting outliers that are slightly off-limits, deleting severely outliers, and deleting missing data.

[0144] Based on the cleaned historical dataset, a feature dataset was constructed, including outdoor ambient temperature, outdoor ambient humidity, chilled water supply temperature, chilled water supply pressure, hot water supply temperature, hot water supply pressure, air conditioning water valve opening percentage, humidifier on / off status, supply air temperature, and supply air dew point temperature. The K-means++ algorithm was then used to cluster the feature dataset. The specific steps are as follows:

[0145] To eliminate dimensional differences, the max-min normalization method is used to standardize the feature data. The specific formula is as follows:

[0146] In the formula, This is the original data. The minimum value of the sample data. The maximum value of the sample data. This is the normalized data.

[0147] The K-means++ initialization algorithm was used to perform a traversal search within the preset cluster number range k∈[8,20]. The optimal number of clusters was determined to be 12 using the silhouette coefficient method. The formula for calculating the silhouette coefficient is as follows:

[0148] In the formula, For the first Profile coefficients for each data point For the first The average distance between a data point and other data points in its cluster Indicates the first The average distance between each data point and all data points in its nearest neighbor cluster.

[0149] Specifically, this includes: Step A1: Pre-setting the range of the number of clusters k to be traversed, which is [8,20] in this embodiment;

[0150] Step A2: Perform a traversal within the preset range and calculate the silhouette coefficient corresponding to the number of clusters. In each traversal, perform the following steps:

[0151] Step B1: Select k data points as initial centroids. The selection of centroids adopts the K-means++ algorithm, as detailed below;

[0152] Step B2: Calculate the distance between each sample point and each centroid, and assign it to the nearest cluster;

[0153] Step B3: For each cluster, take the average of all data points within the cluster as the new centroid;

[0154] Step B4: Repeat step B3. When the cluster centroid no longer changes, stop the loop and output the clustering results. Calculate the silhouette coefficient corresponding to the cluster number based on the clustering results.

[0155] The optimal number of clusters is the number of clusters that corresponds to the maximum silhouette coefficient.

[0156] The initialization of the centroids uses the K-means++ optimization algorithm, including:

[0157] Step C1: Randomly select k objects from N sample data as the initial cluster centers;

[0158] Step C2: Calculate the shortest distance between each sample and the current cluster center, denoted by D(x);

[0159] Step C3: Calculate the probability that each sample point is selected as the next cluster center. This probability is proportional to D(x)^2, and select the sample point corresponding to the maximum probability value as the new cluster center.

[0160] Step C4: Repeat steps C2 and C3 until k new cluster centers are selected as initial centroids, and then execute the standard K-means algorithm.

[0161] For each cluster, Euclidean distance is used as the metric, and the 10 samples closest to the cluster center within each cluster are selected as typical operating conditions, serving as input variables for subsequent model parameter identification. For the mechanism model of the air conditioning system's terminal equipment, the parameters to be identified are rated air volume, rated coil water flow rate, coil heat exchange capacity UA at rated flow rate, rated humidification capacity, and rated pipeline pressure drop. Based on the actual equipment selection for the project, the identification range for each parameter is set as follows:

[0162] Rated air volume: 60000 m³ 3 / h ~ 100000 m 3 / h;

[0163] Rated water flow rate of precooling coil: 15 kg / s ~ 45 kg / s;

[0164] Rated water flow rate for cold coils: 8 kg / s ~ 30 kg / s;

[0165] Rated water flow rate for reheat coil: 1 kg / s ~ 15 kg / s;

[0166] Heat exchange capacity of precooling coil at rated flow rate UA: 30000 W / K ~ 90000 W / K;

[0167] Heat exchange capacity of the cold coil at rated flow rate UA: 30000 W / K ~ 90000 W / K;

[0168] Heat exchange capacity of reheat coil at rated flow rate UA: 5000 W / K ~ 40000 W / K;

[0169] Rated humidification capacity: 0.01 kg / s ~ 0.1 kg / s;

[0170] Rated pressure drop of pipeline: 20 kPa ~ 80 kPa.

[0171] An improved genetic algorithm framework was built in the Python environment, and the mechanism model of the Modelica air conditioning system terminal equipment was embedded into the improved genetic algorithm using OMPython to identify model parameters. The specific steps are as follows:

[0172] Define the fitness evaluation function: Using the clustered typical operating conditions as input variables for model simulation, dynamic simulation calculations are performed by calling the Modelica terminal device mechanism model through OMPython. The mean absolute percentage error (MAPE) between the simulation results and the observed data is defined as the fitness evaluation function. The formula for calculating the mean absolute percentage error is as follows:

[0173]

[0174] in, For the sample size, For the first The predicted data, i.e., the simulation results, For the first Real data, i.e., observational data.

[0175] Initialize multiple populations: Based on the parameters to be identified in the end device mechanism model and the set range of each parameter, 50×3 individuals are randomly generated and divided into 3 initial subpopulations. Each initial subpopulation contains 50 individuals, and each individual is composed of its genes, i.e., the parameters to be identified and fitness.

[0176] Selection operation: Tournament selection is used within each subpopulation. Ten individuals are randomly selected from the subpopulation each time to compete. The individual with the highest fitness is saved to the next generation. This process is repeated until 25 individuals are saved to the next generation.

[0177] Crossover operation: Two-point crossover is applied within each subpopulation. Two individuals are randomly selected with a 50% probability, and a crossover start point and a crossover end point are randomly set in each individual. Genes in the middle of the crossover point are exchanged to generate new offspring individuals.

[0178] Mutation operation: Within each subpopulation, individuals are randomly selected at a rate of 20% to perform mutation operations. For each gene locus of the selected individual, a 10% probability is used to independently determine whether to perform the mutation operation. If the determination is true, the value of the gene locus is replaced with a uniformly random number within the corresponding parameter range.

[0179] Migration operation: Every 10 generations, the best individual in the current subpopulation is migrated to the neighboring subpopulation through a circular migration strategy, replacing the worst individual in the neighboring subpopulation.

[0180] Elite population update: After each migration, select the top 20% of individuals with the best fitness from each subpopulation and store them in the elite population. When the globally optimal individual in the elite population remains for more than 10 generations, the algorithm is considered to have converged and training is stopped. Otherwise, repeat steps (3) to (6).

[0181] After the genetic algorithm is completed, the individual with the best fitness in the elite population is output as the parameter identification result of the end device mechanism model.

[0182] Step 3: Based on the completed mechanism model of the terminal equipment with parameter identification, add ideal PID control to calculate the chilled water demand temperature of the terminal equipment that meets the supply air temperature and humidity requirements of the terminal equipment. Based on this temperature, calculate the upper limit of the chilled water supply temperature of the chiller plant as the boundary condition for chiller plant optimization.

[0183] Based on the identified parameters, a mechanism model of the Modelica terminal equipment with a PID controller was built. The chilled water supply temperature was adjusted using an ideal PID controller. The required chilled water supply temperature, T, was calculated when the water valve opening was 90%, and under the current outdoor temperature and humidity, the supply air temperature reached the set value of 16℃, and the supply air dew point temperature reached the set value of 12℃. sup_demand .

[0184] After parameter identification is completed, the following steps are followed:

[0185] Build a parametric model: Substitute the identified parameters, including air volume and UA value, into the Modelica model component;

[0186] By adding a PID controller to the existing terminal equipment mechanism model, the function of reverse calculation of the water supply temperature is realized. Simply put, the model's simulation calculation is based on the identified model parameters, using the inlet air temperature and humidity, water supply temperature, fan frequency, and water valve opening (inputs) to calculate the supply air temperature and humidity (output). With the addition of the PID controller, the model's inputs and outputs can be adjusted, enabling the calculation of the water supply temperature (output) based on the supply air temperature and humidity requirements (target input), as well as the inlet air temperature and humidity, fan frequency, and water valve opening (inputs).

[0187] PID control is one of the most widely used control methods in industrial scenarios. Its principle is based on feedback control theory. It measures the error between the actual output of the system (the supply air temperature and humidity calculated by the model in this scenario) and the desired output (the target supply air temperature and humidity in this scenario), and adjusts the controller's output according to the magnitude of the error. In real-world scenarios, the output of a PID controller is typically 0~100%, while in an ideal PID model, it can directly output the water supply temperature, with a range set from 3~12℃.

[0188] The calculation of the terminal water supply temperature is achieved through a forward model, namely, coil heat exchange mechanism + PID controller.

[0189] The heat transfer mechanism of the coil has been explained in the discretized numerical wet coil model WetCoilCounterFlow, and the mathematical expression of the PID controller is as follows:

[0190]

[0191] in, , , In order, they are proportional coefficient, integral coefficient, and differential coefficient; for The time-based setpoint, i.e., the target supply air temperature and humidity in this scenario, and the process value, i.e., the difference between the supply air temperature and humidity calculated by the model in this scenario, are used. The control objective of PID is to gradually reduce the temperature and humidity through feedback control. Adjust to 0; for The control variable at any given time is the water supply temperature in this scenario. Represents auxiliary variables within the integral; To represent the differential; This represents the integral.

[0192] Calculate the upper limit of the chilled water supply temperature (T) for the chilled water plant based on the chilled water supply temperature requirements of the terminal equipment. sup_uplim The calculation method is as follows: Summer working condition: T sup_uplim = T sup_demand +0.5℃; Transitional season operating conditions: T sup_uplim = T sup_demand +0.3℃. Wherein, T sup_demand The chilled water supply temperature of the terminal equipment is required to meet the requirements of the supply air temperature and supply air dew point temperature.

[0193] In other words, based on the required chilled water supply temperature T of the terminal equipment sup_demand Calculate the upper limit of the cooling plant water supply temperature T sup_uplim as follows:

[0194] Summer operating conditions: T sup_uplim = T sup_demand +0.5℃;

[0195] Transitional Season Operating Conditions: T sup_uplim = T sup_demand +0.3℃.

[0196] The present invention also provides a chilled water supply temperature optimization system for a terminal model. The chilled water supply temperature optimization system for a terminal model can be implemented by executing the process steps of the chilled water supply temperature optimization method for a terminal model. That is, those skilled in the art can understand the chilled water supply temperature optimization method for a terminal model as a preferred embodiment of the chilled water supply temperature optimization system for a terminal model.

[0197] A chilled water supply temperature optimization system for a terminal model according to the present invention includes:

[0198] Module M1: Constructing a mechanistic model of the terminal equipment in the air conditioning system;

[0199] Module M2: Clusters historical data to identify the parameters of the mechanistic model;

[0200] Module M3: By using the PID control mechanism model, the upper limits of the chilled water demand temperature and supply temperature of the terminal equipment of the air conditioning system are obtained as boundary conditions for the optimization task.

[0201] Specifically, in module M1, the mechanism model can simulate the operation of the terminal equipment of the air conditioning system under different operating conditions;

[0202] The input variables of the mechanism model include: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, fan operating frequency, humidifier on / off status, and fan operating status.

[0203] Specifically, module M2 includes:

[0204] Module M2.1: Based on the preset task objectives, it removes abnormal and missing data from historical data, and then constructs a feature dataset based on the historical data;

[0205] Module M2.2: The feature dataset is clustered using the K-means++ algorithm to generate typical working conditions;

[0206] Module M2.3: Based on typical working conditions, the parameter identification results of the mechanism model are obtained through genetic algorithm.

[0207] Specifically, in module M2.1, the feature dataset includes: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, hot water supply temperature, hot water supply pressure, air conditioning water valve opening percentage, air supply temperature and air supply dew point temperature.

[0208] In module M2.2, the K-means++ algorithm is used to perform a traversal search within a preset range of cluster numbers, and then the optimal number of clusters is determined by the silhouette coefficient method. Then, the P samples closest to the cluster center in each cluster are selected as typical working conditions, where the value of P ranges from 10 to 100; wherein, the optimal number of clusters is the number of clusters corresponding to the maximum silhouette coefficient.

[0209] The mathematical expression for the contour coefficient method is:

[0210]

[0211] in, The contour coefficient for the i-th data point. Let be the average distance between the i-th data point and all other data points in its cluster. This represents the average distance between the i-th data point and all data points in the nearest neighbor cluster;

[0212] In module M2.3, the parameter identification results of the mechanistic model are obtained through a genetic algorithm, including:

[0213] Module M2.3.1: Generates 50×3 individuals, divided into 3 initial subpopulations, each containing 50 individuals;

[0214] Module M2.3.2: Uses tournament selection, randomly selecting 10 individuals from the subpopulation each time to compete, until the number of individuals preserved to the next generation reaches 25;

[0215] Module M2.3.3: Based on preset probability values ​​and proportion values, crossover and mutation are performed on individuals within each subpopulation;

[0216] Module M2.3.4: Every 10 generations, the best individual in the current subpopulation is migrated to an adjacent subpopulation through a circular migration strategy, replacing the worst individual in the adjacent subpopulation. After each migration, the top 20% of individuals with the best fitness are selected from each subpopulation and stored in the elite population. It is determined whether the globally best individual in the elite population has been maintained for more than 10 generations. If the result is yes, the globally best individual is output as a parameter of the mechanism model. If the result is no, module M2.3.2 is triggered again.

[0217] The mathematical expression for fitness is:

[0218]

[0219] in, For the sample size, For the first The predicted data, i.e., the simulation results, For the first Real data, i.e., observational data.

[0220] Specifically, in module M3, the required temperature of the chilled water is set to T. sup_demand The mathematical expressions for the upper limit of the temperature under summer and transitional season conditions, from top to bottom, are as follows:

[0221] T sup_uplim =T sup_demand +0.5℃

[0222] T sup_uplim =T sup_demand +0.3℃

[0223] Among them, T sup_uplim This indicates the upper limit of the temperature.

[0224] The mathematical expression for the mechanism model based on PID control is:

[0225]

[0226] in, , , In order, they are proportional coefficient, integral coefficient, and differential coefficient; for The time-based setpoint, i.e., the target supply air temperature and humidity in this scenario, and the process value, i.e., the difference between the supply air temperature and humidity calculated by the model in this scenario, are used. The control objective of PID is to gradually reduce the temperature and humidity through feedback control. Adjust to 0; for The control variable at any given time is the water supply temperature in this scenario. Represents auxiliary variables within the integral; To represent the differential; This represents the integral.

[0227] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0228] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for optimizing chilled water supply temperature in a terminal model, characterized in that, include: Step S1: Build a mechanistic model of the terminal equipment of the air conditioning system; Step S2: Cluster the historical data to identify the parameters of the mechanism model; Step S3: Using the PID control mechanism model, the upper limits of the chilled water demand temperature and supply temperature of the air conditioning system terminal equipment are obtained as boundary conditions for the optimization task. In step S1, the mechanism model can simulate the operation of the terminal equipment of the air conditioning system under different operating conditions. The input variables of the mechanism model include: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, fan operating frequency, humidifier on / off status, and fan operating status. Step S2 includes: Step S2.1: According to the preset task objective, remove abnormal and missing data from the historical data, and then construct a feature dataset based on the historical data; Step S2.2: Use the K-means++ algorithm to cluster the feature dataset and generate typical working conditions; Step S2.3: Based on typical working conditions, obtain the parameter identification results of the mechanism model through a genetic algorithm; In step S3, the required temperature of the chilled water is set to T. sup_demand The mathematical expressions for the upper limit of the temperature under summer and transitional season conditions, from top to bottom, are as follows: T sup_uplim =T sup_demand +0.5℃ T sup_uplim =T sup_demand +0.3℃ Among them, T sup_uplim Indicates the upper limit of temperature; The mathematical expression for the mechanism model based on PID control is: in, , , In order, they are proportional coefficient, integral coefficient, and differential coefficient; for Time setting value; for Control variables at any given time; Represents auxiliary variables within the integral; To represent the differential; This represents the integral.

2. The method for optimizing the chilled water supply temperature of the terminal model according to claim 1, characterized in that, In step S2.1, the feature dataset includes: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, hot water supply temperature, hot water supply pressure, air conditioning water valve opening percentage, air supply temperature and air supply dew point temperature. In step S2.2, the K-means++ algorithm is used to perform a traversal search within a preset range of cluster numbers, and then the optimal number of clusters is determined by the silhouette coefficient method. Then, the P samples closest to the cluster center in each cluster are selected as typical working cases, where the value of P ranges from 10 to 100. The optimal number of clusters is the number of clusters corresponding to the maximum silhouette coefficient. The mathematical expression for the contour coefficient method is: in, Let i be the contour coefficient of the i-th data point. Let be the average distance between the i-th data point and all other data points in its cluster. This represents the average distance between the i-th data point and all data points in the nearest neighbor cluster; In step S2.3, the parameter identification results of the mechanistic model are obtained through a genetic algorithm, including: Step S2.3.1: Generate M×N individuals, divide them into N initial subpopulations, N≥3, and each initial subpopulation contains M individuals; Step S2.3.2: Use tournament selection. Each time, randomly select Y individuals from the subpopulation and let them compete until the number of individuals preserved to the next generation reaches M / 2. Step S2.3.3: Based on the preset probability values ​​and proportion values, crossover and mutate individuals within each subpopulation; Step S2.3.4: Every 10 generations, the best individual in the current subpopulation is migrated to an adjacent subpopulation using a circular migration strategy, replacing the worst individual in the adjacent subpopulation. After each migration, the top 20% of individuals with the best fitness are selected from each subpopulation and stored in the elite population. It is determined whether the globally optimal individual in the elite population has been maintained for more than 10 generations. If yes, the globally optimal individual is output as a parameter of the mechanism model. If no, step S2.3.2 is executed again. The mathematical expression for fitness is: in, For the sample size, For the first The predicted data, i.e., the simulation results, For the first Real data, i.e., observational data.

3. A chilled water supply temperature optimization system for a terminal model, characterized in that, include: Module M1: Constructing a mechanistic model of the terminal equipment in the air conditioning system; Module M2: Clusters historical data to identify the parameters of the mechanistic model; Module M3: By using the PID control mechanism model, the upper limits of the chilled water demand temperature and supply temperature of the air conditioning system terminal equipment are obtained as boundary conditions for the optimization task; In module M1, the mechanism model can simulate the operation of the terminal equipment of the air conditioning system under different operating conditions; The input variables of the mechanism model include: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, fan operating frequency, humidifier on / off status, and fan operating status. The module M2 includes: Module M2.1: Based on the preset task objectives, it removes abnormal and missing data from historical data, and then constructs a feature dataset based on the historical data; Module M2.2: The feature dataset is clustered using the K-means++ algorithm to generate typical working conditions; Module M2.3: Based on typical working conditions, the parameter identification results of the mechanism model are obtained through genetic algorithm; In module M3, the required temperature of the chilled water is set to T. sup_demand The mathematical expressions for the upper limit of the temperature under summer and transitional season conditions, from top to bottom, are as follows: T sup_uplim =T sup_demand +0.5℃ T sup_uplim =T sup_demand +0.3℃ Among them, T sup_uplim Indicates the upper limit of temperature; The mathematical expression for the mechanism model based on PID control is: in, , , In order, they are proportional coefficient, integral coefficient, and differential coefficient; for Time setting value; for Control variables at any given time; Represents auxiliary variables within the integral; To represent the differential; This represents the integral.

4. The chilled water supply temperature optimization system for the terminal model according to claim 3, characterized in that, In module M2.1, the feature dataset includes: chilled water supply temperature, outdoor ambient temperature, outdoor ambient humidity, hot water supply temperature, hot water supply pressure, air conditioning water valve opening percentage, air supply temperature and air supply dew point temperature. In module M2.2, the K-means++ algorithm is used to perform a traversal search within a preset range of cluster numbers, and then the optimal number of clusters is determined by the silhouette coefficient method. Then, the P samples closest to the cluster center in each cluster are selected as typical working conditions, where the value of P ranges from 10 to 100; wherein, the optimal number of clusters is the number of clusters corresponding to the maximum silhouette coefficient. The mathematical expression for the contour coefficient method is: in, Let i be the contour coefficient of the i-th data point. Let be the average distance between the i-th data point and all other data points in its cluster. This represents the average distance between the i-th data point and all data points in the nearest neighbor cluster; In module M2.3, the parameter identification results of the mechanistic model are obtained through a genetic algorithm, including: Module M2.3.1: Generates M×N individuals, divided into N initial subpopulations, where N≥3, and each initial subpopulation contains M individuals; Module M2.3.2: A tournament selection method is adopted, in which Y individuals are randomly selected from the subpopulation each time to compete with each other until the number of individuals preserved to the next generation reaches M / 2. Module M2.3.3: Based on preset probability values ​​and proportion values, crossover and mutation are performed on individuals within each subpopulation; Module M2.3.4: Every 10 generations, the best individual in the current subpopulation is migrated to an adjacent subpopulation through a circular migration strategy, replacing the worst individual in the adjacent subpopulation. After each migration, the top 20% of individuals with the best fitness are selected from each subpopulation and stored in the elite population. It is determined whether the globally best individual in the elite population has been maintained for more than 10 generations. If the result is yes, the globally best individual is output as a parameter of the mechanism model. If the result is no, module M2.3.2 is triggered again. The mathematical expression for fitness is: in, For the sample size, For the first The predicted data, i.e., the simulation results, For the first Real data, i.e., observational data.

Images