Energy consumption simulation method of central air conditioning system based on multi-zone coupling model

Abstract

A method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model, the method comprising: obtaining basic data and initial parameters at the site of a cold chain logistics center, and preprocessing the basic data and initial parameters. A multi-region coupling physical mechanism model is constructed based on the heat and mass transfer equations, and the preprocessed basic data and initial parameters are used as boundary conditions, and the physical model state is output through the finite difference method. The physical model state is used as the input of a convolutional neural network to extract short-term infiltration load characteristics through time series convolution, and the short-term infiltration load prediction result is output. The short-term infiltration load prediction result is combined with the cargo thermal parameters and the partition state as the input of the Transformer model, and the multi-region energy consumption distribution prediction data is output. The multi-region energy consumption distribution prediction data is optimized by MPC to calculate the preliminary control strategy. The global optimal solution is searched in the high-dimensional nonlinear parameter space by the slime mold optimization algorithm, and the optimal control strategy is output.

Description

Technical Field

[0001] The present invention belongs to the technical field of environmental control of large cold chain logistics centers, and more specifically, relates to an energy consumption simulation method for a central air-conditioning system based on a multi-region coupling model. Background Art

[0002] In the whole process of low temperature control of large cold chain logistics centers, the energy consumption simulation method of central air-conditioning system based on multi-region coupling model is a key technical means for energy optimization and operation decision-making. In current engineering practice, technicians usually use steady-state or quasi-steady-state models to approximate the cooling demand of the warehouse area to determine the selection of refrigeration equipment, pipe network layout and control strategy. However, these traditional simulation methods are mainly suitable for scenes with relatively stable heat load, less dynamic changes in inbound and outbound shipments and space use, such as office buildings, commercial buildings or standard cold storage.

[0003] First, compared with constant environments such as data centers, large cold chain logistics centers must frequently open and close cold storage doors to meet the requirements of goods entering and exiting. This frequent opening and closing of the door will introduce a large amount of warm and humid air from the outside in a short period of time, resulting in significant transient heat and humidity load fluctuations. Traditional steady-state or quasi-steady-state models often regard door opening as a simple fixed infiltration volume, and it is difficult to consider the coupling effects of door opening frequency (5-15 times / hour), opening duration (a few minutes), and external climate conditions (30°C / relative humidity above 60% in summer) on transient loads, resulting in insufficient simulation response to short-term heat and humidity fluctuations.

[0004] Secondly, the cold chain logistics center stores a wide variety of items (fresh food, meat, semi-finished food, etc.), and the initial temperature, specific heat capacity and thermal conductivity of the goods entering the warehouse area vary greatly (for example, newly stored meat may be close to 5°C, while the environment in the warehouse is -2°C~5°C). The goods will release or absorb a considerable amount of heat (sensible heat, latent heat) during the initial storage stage, resulting in obvious time-varying characteristics of the heat load in the warehouse. Traditional air conditioning simulations usually use a uniform and simplified internal load model, which is difficult to accurately simulate the transient heat flow changes when different batches of goods enter the warehouse. Since the heat capacity of the goods (such as 0.8~3.5 kJ / (kg·K)) and quantity changes (ton level in and out) have a great impact on the indoor energy balance, if it is difficult to refine the sub-model for different types of goods and storage temperature conditions, the energy consumption prediction will have a large deviation.

[0005] Finally, in the low-temperature control of the cold chain center, there may be multiple independent or graded refrigeration areas (such as quick-freezing areas, refrigerated areas, and sorting and loading areas), and the flow of goods is realized through internal automated conveyor lines and forklift flow paths. The demand for cooling capacity of this regional division and cargo flow line is not constant, and may change with peak shipments (such as morning and evening shipments). It is difficult for conventional models to accurately associate changes in logistics flow routes (changes in the amount of goods entering and leaving different cold storages at different time periods) with the transfer of cooling capacity between low-temperature areas, the frequency of access control use, and the air exchange of internal transmission channels. The main result is that the model often regards the energy consumption of each area as an independent constant value, and it is difficult to accurately simulate the dynamic redistribution of cooling and heat under the synergy of multiple areas, resulting in large errors in energy consumption statistics. Summary of the invention

[0006] In order to solve the deficiencies in the prior art, the purpose of the present invention is to solve the above-mentioned defects and further propose a.

[0007] The present invention adopts the following technical solution.

[0008] The first aspect of the present invention discloses a method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model, the method comprising:

[0009] Obtaining basic data and initial parameters at the cold chain logistics center site, and preprocessing the basic data and initial parameters;

[0010] A multi-region coupled physical mechanism model is constructed based on the heat and mass transfer equations, and the pre-processed basic data and initial parameters are used as boundary conditions to output the physical model state through the finite difference method;

[0011] The physical model state is used as an input of a convolutional neural network to extract short-period seepage load characteristics through time series convolution to output a short-term seepage load prediction result;

[0012] The short-term infiltration load prediction results are combined with cargo thermal parameters and partition status as inputs to the Transformer model, so as to combine the short-term infiltration load prediction results with cargo distribution and regional characteristics through the attention mechanism and output multi-region energy consumption distribution prediction data;

[0013] Performing strategy optimization processing on the multi-region energy consumption distribution prediction data through MPC to calculate a preliminary control strategy within a rolling time window;

[0014] The preliminary control strategy is used as the initial solution of the slime mold optimization algorithm to search for a global optimal solution in a high-dimensional nonlinear parameter space, output an optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center site;

[0015] Among them, the basic data and initial parameters include access control opening and closing records, outdoor meteorological conditions, cargo types and storage temperature, refrigeration equipment parameters and temperature and humidity of each partition; the preprocessing includes data cleaning, data synchronization and time alignment, outlier correction or elimination, data interpolation and missing value filling; the physical model state includes the temperature and humidity field of each partition on site of the cold chain logistics center at the current moment, the cargo heat exchange status and short-term historical penetration load records; the short-term penetration load prediction result is the change of door opening and closing penetration heat and humidity load in the first time period in the future; the optimal control strategy includes refrigeration equipment power setting, fan speed and valve adjustment and access control pre-control strategy.

[0016] Furthermore, the obtaining of basic data and initial parameters at the cold chain logistics center site and preprocessing of the basic data and initial parameters include:

[0017] The access control opening and closing records are obtained through the access control sensor, and the access control opening and closing records include the opening and closing frequency and duration; the temperature and humidity of each partition and the outdoor meteorological conditions are obtained through the temperature and humidity sensor and the meteorological sensor, and the outdoor meteorological conditions include the outdoor temperature and relative humidity; and the operating parameters of the refrigeration equipment are obtained through the equipment parameter sensor;

[0018] The data collected by different sensors are aligned according to timestamps to synchronize and time-align the basic data and initial parameters, and outlier detection is performed on the sensor data to correct or eliminate the detected outliers, and missing data points are filled in using linear interpolation to obtain preprocessed basic data and initial parameters.

[0019] Furthermore, the multi-region coupling physical mechanism model is constructed based on the heat and mass transfer equations, and the pre-processed basic data and initial parameters are used as boundary conditions to output the physical model state through the finite difference method, including:

[0020] The cold chain logistics center site is divided into multiple control areas, and the physical parameters of each control area are determined. The energy conservation equation and mass conservation equation of each area are time discretized by using the explicit finite difference method to establish the discretized heat and mass transfer equation. The physical parameters include volume, air density, specific heat capacity and air flow resistance. The expression of the heat and mass transfer equation is:

[0021]

[0022]

[0023] Where n is the current time step, is the air density of the z-th control area, is the volume of the z-th control region, is the specific heat capacity of air in the zth control area, and is the input and output heat of the z-th control area at time step n, and is the input and output water vapor of the zth control area at time step n, and is the humidity of the zth region at time steps n and n+1, and is the water vapor content in the zth region at time steps n and n+1;

[0024] Selecting access control penetration load and cargo heat exchange load as boundary conditions from the preprocessed basic data and initial parameters, calculating the temperature and humidity field of each partition and the short-term historical penetration load to output the physical model state;

[0025] The physical model state is composed of the temperature vector of each partition, the relative humidity vector of each partition, the heat penetration load vector of each partition, and the water vapor penetration load vector of each partition.

[0026] Furthermore, the physical model state is used as the input of the convolutional neural network to extract the short-period seepage load characteristics through time series convolution to output the short-term seepage load prediction result, including:

[0027] Combining the temperature vector and the relative humidity vector in the physical model state into a two-dimensional matrix, and selecting the physical model states of multiple time steps to construct a time series input feature matrix, wherein the time series input feature matrix is ​​composed of multiple two-dimensional matrices;

[0028] Using the time series input feature matrix as the input of the convolutional neural network to perform a convolution operation on the time series input feature matrix to generate a feature map, and converting the feature map into a one-dimensional vector as the input of a fully connected layer to output a high-level feature vector;

[0029] Two output nodes are defined in the output layer of the convolutional neural network to respectively predict the future short-term heat permeation load and water vapor permeation load to obtain the permeation load prediction result vector, and the prediction result vector is structured and stored. The prediction result vector is composed of the heat permeation load and the water vapor permeation load.

[0030] Furthermore, the short-term penetration load prediction result is combined with the cargo thermal parameters and the partition status as the input of the Transformer model, so as to combine the short-term penetration load prediction result with the cargo distribution and regional characteristics through the attention mechanism, and output multi-region energy consumption distribution prediction data, including:

[0031] Combining the heat permeation load and the water vapor permeation load with corresponding cargo distribution data to generate a first characteristic matrix, and combining the regional characteristic data with the first characteristic matrix to generate a second characteristic matrix;

[0032] Standardize each column feature in the second feature matrix, normalize the commodity type number in the second feature matrix to obtain a third feature matrix, select a time window of fixed length, and generate an input sequence set through a sliding window;

[0033] Using the input sequence set as the input of the trained Transformer model, outputting the multi-region energy consumption distribution prediction data through an encoder and a decoder, and performing structured storage on the multi-region energy consumption distribution prediction data;

[0034] Among them, the trained Transformer model is obtained by training the Transformer model based on the input sequence set and defining multi-region energy consumption distribution prediction data as the target output, using mean square error as the loss function and adopting Adam optimizer for parameter update.

[0035] Furthermore, the multi-region energy consumption distribution prediction data is subjected to strategy optimization processing by MPC to calculate a preliminary control strategy within a rolling time window, including:

[0036] Acquire control variables, and construct a control strategy parameter set based on the control variables, wherein the control variables include refrigeration equipment power setting, fan speed, valve opening, and access control pre-control strategy;

[0037] The control strategy parameter set and multi-region energy consumption distribution prediction data are used as inputs of the MPC, and the minimization of comprehensive energy consumption, temperature control constraints, refrigeration equipment power limits, fan speed limits, valve opening limits, and access control strategy constraints are defined to construct an MPC optimization model;

[0038] The MPC optimization model is solved by an optimization solver to obtain an initial control strategy parameter set, and the initial control strategy parameter set is sent to the refrigeration equipment, fans and valves in each partition to perform real-time control on the refrigeration equipment, fans and valves in each partition.

[0039] Furthermore, the preliminary control strategy is used as the initial solution of the slime mold optimization algorithm to search for the global optimal solution in the high-dimensional nonlinear parameter space, output the optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center, including:

[0040] Initializing the slime mold optimization algorithm through the control variables, the set SAM parameters and the initialization population, wherein the SAM parameters include the population size, the maximum number of iterations, the learning factor and the random factor, and the initialization population is generated based on the initial control strategy parameter set;

[0041] Based on the initialized slime mold optimization algorithm, the objective function is to minimize the comprehensive energy consumption, and the temperature control constraint, refrigeration equipment power limit, fan speed limit, valve opening limit and access control strategy constraint are defined as constraint conditions. The position of the slime mold is updated and the fitness of each individual is calculated. By comparing the fitness of all individuals, the individual with the smallest fitness is updated as the optimal solution, and the optimal control strategy parameter set is obtained.

[0042] The second aspect of the present invention discloses a central air conditioning system energy consumption simulation device, the device comprising:

[0043] A data preprocessing module is used to obtain basic data and initial parameters at the cold chain logistics center site and preprocess the basic data and initial parameters;

[0044] A physical model building module is used to build a multi-region coupled physical mechanism model based on the heat and mass transfer equations, and output the physical model state through the finite difference method using the preprocessed basic data and initial parameters as boundary conditions;

[0045] A short-term seepage load prediction module, used to use the physical model state as an input of a convolutional neural network to extract short-period seepage load characteristics through time series convolution to output a short-term seepage load prediction result;

[0046] A multi-region energy consumption distribution prediction module is used to combine the short-term penetration load prediction results with the cargo thermal parameters and the partition status as the input of the Transformer model, so as to combine the short-term penetration load prediction results with the cargo distribution and regional characteristics through the attention mechanism, and output the multi-region energy consumption distribution prediction data;

[0047] A control strategy optimization module, used for performing strategy optimization processing on the multi-region energy consumption distribution prediction data through MPC to calculate a preliminary control strategy within a rolling time window;

[0048] An optimal control strategy generation module is used to use the preliminary control strategy as the initial solution of the slime mold optimization algorithm to search for a global optimal solution in a high-dimensional nonlinear parameter space, output an optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center site;

[0049] Among them, the basic data and initial parameters include access control opening and closing records, outdoor meteorological conditions, cargo types and storage temperature, refrigeration equipment parameters and temperature and humidity of each partition; the preprocessing includes data cleaning, data synchronization and time alignment, outlier correction or elimination, data interpolation and missing value filling; the physical model state includes the temperature and humidity field of each partition on site of the cold chain logistics center at the current moment, the cargo heat exchange status and short-term historical penetration load records; the short-term penetration load prediction result is the change of door opening and closing penetration heat and humidity load in the first time period in the future; the optimal control strategy includes refrigeration equipment power setting, fan speed and valve adjustment and access control pre-control strategy.

[0050] The third aspect of the present invention discloses a terminal, including a processor and a storage medium; the characteristics are:

[0051] The storage medium is used to store instructions;

[0052] The processor is used to operate according to the instructions to execute the steps of the method described in the first aspect.

[0053] The fourth aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0054] The beneficial effects of the present invention are that, compared with the prior art, the present invention has the following advantages:

[0055] Through the convolutional neural network, the short-term seepage load characteristics are extracted according to the state of the physical model, and the future short-term seepage load results are predicted. The Transformer model is used to divide the area of ​​the cold chain logistics center site, and the predicted data of multi-area energy consumption distribution is output. Then, the preliminary control strategy is generated through the MPC optimization model, and finally the optimal control strategy is output in combination with the slime mold optimization algorithm. In the face of the high-frequency opening and closing of access control and the rapid replacement of cargo types and batches unique to large cold chain logistics centers, it can more accurately simulate the short-term heat and humidity fluctuation load when the access control is frequently opened and closed, which improves the accuracy of energy consumption prediction to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of the energy consumption simulation method of a central air-conditioning system based on a multi-region coupling model provided by the present invention. DETAILED DESCRIPTION

[0057] The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present application.

[0058] like Figure 1As shown, in one embodiment, a method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model comprises the following steps:

[0059] Step S110, obtaining basic data and initial parameters of the cold chain logistics center site, and preprocessing the basic data and initial parameters.

[0060] Among them, the basic data and initial parameters include access control opening and closing records, outdoor meteorological conditions, cargo types and storage temperature, refrigeration equipment parameters, and temperature and humidity in each partition. Preprocessing includes data cleaning, data synchronization and time alignment, outlier correction or elimination, data interpolation and missing value filling.

[0061] In some embodiments, the central air-conditioning system energy consumption simulation method based on the multi-region coupling model provided by the present invention obtains basic data and initial parameters of the cold chain logistics center site, and pre-processes the basic data and initial parameters, specifically including the following steps:

[0062] Step S111, obtain the access control opening and closing records through the access control sensor, the access control opening and closing records include the opening and closing frequency and duration, obtain the temperature and humidity of each partition and the outdoor meteorological conditions through the temperature and humidity sensor and the meteorological sensor, the outdoor meteorological conditions include outdoor temperature and relative humidity, and obtain the operating parameters of the refrigeration equipment through the equipment parameter sensor.

[0063] Step S112, align the data collected by different sensors according to the timestamps to synchronize and time-align the basic data and initial parameters, perform outlier detection on the sensor data to correct or eliminate the detected outliers, and fill in the missing data points using linear interpolation to obtain the preprocessed basic data and initial parameters.

[0064] Step S120, constructing a multi-region coupling physical mechanism model based on the heat and mass transfer equations, and using the preprocessed basic data and initial parameters as boundary conditions, outputting the physical model state through the finite difference method.

[0065] Among them, the physical model status includes the temperature and humidity fields of each partition of the cold chain logistics center at the current moment, the heat exchange status of the goods, and the short-term historical penetration load records.

[0066] In some embodiments, the energy consumption simulation method of a central air-conditioning system based on a multi-region coupling model provided by the present invention constructs a multi-region coupling physical mechanism model based on the heat and mass transfer equations, and uses the preprocessed basic data and initial parameters as boundary conditions, and outputs the physical model state through the finite difference method, specifically including the following steps:

[0067] Step S121, divide the cold chain logistics center site into multiple control areas, and determine the physical parameters of each control area, so as to use the explicit finite difference method to time discretize the energy conservation equation and mass conservation equation of each area, and establish a discretized heat and mass transfer equation. The physical parameters include volume, air density, specific heat capacity and air flow resistance. The expression of the heat and mass transfer equation is:

[0068]

[0069]

[0070] Where n is the current time step, is the air density of the z-th control area, is the volume of the z-th control region, is the specific heat capacity of air in the zth control area, and is the input and output heat of the z-th control area at time step n, and is the input and output water vapor of the zth control area at time step n, and is the humidity of the zth region at time steps n and n+1, and is the water vapor content in the zth region at time steps n and n+1.

[0071] Step S122, select the access control penetration load and the cargo heat exchange load as boundary conditions from the preprocessed basic data and initial parameters, calculate the temperature and humidity field of each partition and the short-term historical penetration load, and output the physical model state.

[0072] The physical model state is composed of the temperature vector of each partition, the relative humidity vector of each partition, the heat penetration load vector of each partition, and the water vapor penetration load vector of each partition.

[0073] Step S130, taking the physical model state as the input of the convolutional neural network to extract the short-period seepage load characteristics through time series convolution to output the short-term seepage load prediction result.

[0074] Among them, the short-term infiltration load prediction result is the change in the door opening and closing infiltration heat and moisture load in the first time period in the future. The first time period is a period of time in the future whose duration is lower than the set threshold, which is defined as short-term.

[0075] In some embodiments, the energy consumption simulation method of a central air-conditioning system based on a multi-region coupling model provided by the present invention uses the physical model state as the input of a convolutional neural network to extract short-period infiltration load characteristics through time series convolution to output a short-term infiltration load prediction result, specifically comprising the following steps:

[0076] Step S131, combining the temperature vector and the relative humidity vector in the physical model state into a two-dimensional matrix, and selecting the physical model states of multiple time steps to construct a time series input feature matrix, the time series input feature matrix is ​​composed of multiple two-dimensional matrices.

[0077] Step S132, using the time series input feature matrix as the input of the convolutional neural network to perform a convolution operation on the time series input feature matrix to generate a feature map, and converting the feature map into a one-dimensional vector as the input of the fully connected layer to output a high-level feature vector.

[0078] Step S133, defining two output nodes in the output layer of the convolutional neural network, respectively predicting the future short-term heat permeation load and water vapor permeation load to obtain a permeation load prediction result vector, and performing structured storage on the prediction result vector, which is composed of the heat permeation load and the water vapor permeation load.

[0079] Step S140, taking the short-term penetration load prediction result in combination with the cargo thermal parameters and the partition status as the input of the Transformer model, so as to combine the short-term penetration load prediction result with the cargo distribution and regional characteristics through the attention mechanism, and output the multi-region energy consumption distribution prediction data.

[0080] In some embodiments, the energy consumption simulation method of a central air-conditioning system based on a multi-region coupling model provided by the present invention combines the short-term penetration load prediction result with the cargo thermal parameters and the partition state as the input of the Transformer model, so as to combine the short-term penetration load prediction result with the cargo distribution and regional characteristics through the attention mechanism, and output the multi-region energy consumption distribution prediction data, which specifically includes the following steps:

[0081] Step S141, combining the heat permeation load and the water vapor permeation load with the corresponding cargo distribution data to generate a first characteristic matrix, and combining the regional characteristic data with the first characteristic matrix to generate a second characteristic matrix.

[0082] Step S142, standardize each column feature in the second feature matrix, normalize the commodity category numbers in the second feature matrix to obtain a third feature matrix, select a time window of fixed length, and generate an input sequence set through a sliding window.

[0083] Step S143, using the input sequence set as the input of the trained Transformer model, outputting the multi-region energy consumption distribution prediction data through the encoder and decoder, and storing the multi-region energy consumption distribution prediction data in a structured manner.

[0084] Among them, the trained Transformer model is obtained by training the Transformer model based on the input sequence set and defining the multi-region energy consumption distribution prediction data as the target output, using the mean square error as the loss function and the Adam optimizer for parameter update.

[0085] Step S150, performing strategy optimization processing on the multi-region energy consumption distribution prediction data through MPC to calculate a preliminary control strategy within a rolling time window.

[0086] In some embodiments, the central air-conditioning system energy consumption simulation method based on the multi-region coupling model provided by the present invention performs strategy optimization processing on the multi-region energy consumption distribution prediction data through MPC to calculate the preliminary control strategy within the rolling time window, specifically including the following steps:

[0087] Step S151, obtaining control variables, and constructing a control strategy parameter set based on the control variables, the control variables including refrigeration equipment power setting, fan speed, valve opening, and access control pre-control strategy.

[0088] Step S152, using the control strategy parameter set and multi-region energy consumption distribution prediction data as the input of MPC, and defining the minimization of comprehensive energy consumption, temperature control constraints, refrigeration equipment power limitations, fan speed limitations, valve opening limitations, and access control strategy constraints to build an MPC optimization model.

[0089] Step S153, solving the MPC optimization model through the optimization solver to obtain an initial control strategy parameter set, and sending the initial control strategy parameter set to the refrigeration equipment, fans and valves in each partition to perform real-time control on the refrigeration equipment, fans and valves in each partition.

[0090] In step S160, the preliminary control strategy is used as the initial solution of the slime mold optimization algorithm to search for the global optimal solution in the high-dimensional nonlinear parameter space, output the optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center.

[0091] Among them, the optimal control strategy includes refrigeration equipment power setting, fan speed and valve adjustment, and access control pre-control strategy.

[0092] In some embodiments, the energy consumption simulation method of the central air-conditioning system based on the multi-region coupling model provided by the present invention uses the preliminary control strategy as the initial solution of the slime mold optimization algorithm to search for the global optimal solution in the high-dimensional nonlinear parameter space, output the optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center site, specifically including the following steps:

[0093] Step S161, initializing the slime mold optimization algorithm by controlling variables, setting SAM parameters and initializing the population. The SAM parameters include population size, maximum number of iterations, learning factor and random factor. The initialization population is generated based on the initial control strategy parameter set.

[0094] Step S162, based on the initialized slime mold optimization algorithm, define minimizing the comprehensive energy consumption as the objective function, and define temperature control constraints, refrigeration equipment power limitations, fan speed limitations, valve opening limitations, and access control strategy constraints as constraints, update the slime mold position and calculate the fitness of each individual, so as to compare the fitness of all individuals, update the individual with the smallest fitness as the optimal solution, and obtain the optimal control strategy parameter set.

[0095] The above-mentioned central air-conditioning system energy consumption simulation method based on the multi-region coupling model extracts the short-term penetration load characteristics according to the physical model state through the convolutional neural network, and predicts the future short-term penetration load results. It combines the Transformer model to divide the area of ​​the cold chain logistics center site, outputs the predicted data of multi-region energy consumption distribution, and then generates a preliminary control strategy through the MPC optimization model. Finally, it combines the slime mold optimization algorithm to output the optimal control strategy. In the face of the high-frequency opening and closing of access control and the rapid replacement of cargo types and batches unique to large cold chain logistics centers, it can more accurately simulate the short-term heat and humidity fluctuation load when the access control is frequently opened and closed, which improves the accuracy of energy consumption prediction to a certain extent.

[0096] In a specific embodiment, the energy consumption simulation method of a central air-conditioning system based on a multi-region coupling model provided by the present invention includes steps 1 to 6:

[0097] Step 1: Obtain basic data and initial parameters.

[0098] Specifically, on-site data of the cold chain logistics center is collected, including access control opening and closing records (frequency 5 to 15 times / hour, opening time of several minutes), outdoor weather conditions (30°C, 60%RH typical), types of goods and storage temperature (-2°C to 5°C range), refrigeration equipment parameters and current temperature and humidity in each partition.

[0099] In this embodiment, the sensor types include access control sensors, temperature and humidity sensors, meteorological sensors, and equipment parameter sensors. Among them, the access control sensor is installed at the main entrance and exit to record the frequency and duration of opening and closing. The temperature and humidity sensors are distributed at key locations of each partition on the cold chain logistics center site to measure the temperature and relative humidity of each indoor partition. The meteorological sensor is installed at a designated location outdoors and avoids direct sunlight and rain to measure the outdoor temperature and relative humidity. The equipment parameter sensor is installed at key parts of the refrigeration equipment, such as chillers and fan coils, to monitor the operating status of the refrigeration equipment. In addition, the access control sensor records the opening and closing status once a second, the temperature and humidity sensors and meteorological sensors record data once a minute, and the sensor data is transmitted to the central data server using a wired or wireless network.

[0100] In this embodiment, in the raw data collected by the sensor, the door opening and closing records include the timestamp, opening and closing times, and opening / closing duration of each door opening and closing. The outdoor meteorological condition data includes the outdoor temperature and relative humidity every minute, the type of goods and the storage temperature collection records the type and storage temperature of each batch of goods, and the refrigeration equipment parameter collection records the operating parameters of the refrigeration equipment, including the cooling capacity and fan speed of the chiller. The current temperature and relative humidity of each partition are collected to record the current temperature and relative humidity of each partition.

[0101] After data collection is completed, the above raw data is preprocessed to align the data from different sensors according to the timestamps to ensure that the data corresponds to each other within the same time step. It is also necessary to detect abnormal values ​​in the sensor data, such as sudden temperature changes and humidity jumps, and modify or eliminate them. Missing data points are filled with linear interpolation or other interpolation methods to ensure data continuity.

[0102] Step 2: Build a multi-region coupled physical mechanism model and output the current physical model status.

[0103] Specifically, a multi-region coupling performance simulation model is established based on the heat and mass transfer equations. The data output in step 1 is used as boundary conditions, and the temperature and humidity fields, cargo heat exchange conditions and short-term historical penetration load records of each partition at the current moment are obtained through the finite difference method.

[0104] First, define the multi-region model structure and divide the cold chain logistics center into multiple control areas. Each area corresponds to a unique area number. Then determine the physical parameters of each area, namely, volume, air density, specific heat capacity, and air flow resistance. Secondly, determine the energy conservation equation and mass conservation equation for each area. Use the explicit finite difference method to time discretize the determined energy conservation equation and mass conservation equation. The discretized equation is the discretized heat and mass transfer equation, which is expressed as:

[0105]

[0106]

[0107] Where n is the current time step, is the air density of the z-th control area, is the volume of the z-th control region, is the specific heat capacity of air in the zth control area, and is the input and output heat of the z-th control area at time step n, and is the input and output water vapor of the zth control area at time step n, and is the humidity of the zth region at time steps n and n+1, and is the water vapor content in the zth region at time steps n and n+1.

[0108] In this embodiment, the access control penetration load and the cargo heat exchange load are used as boundary conditions to calculate the temperature and humidity fields of each partition, and to couple heat transfer and mass transfer, that is, heat transfer between partitions and water vapor transfer between partitions. The temperature update formula is:

[0109]

[0110] In the formula, is the heat introduced into the zth region at time step n due to the door opening and closing, is the heat introduced into the zth region at time step n due to heat exchange of cargo. The other variables have the same meanings as the above variables and will not be repeated here.

[0111] The humidity update formula is:

[0112]

[0113] In the formula, is the amount of water vapor introduced into the zth region at time step n due to the opening and closing of the door, is the amount of water vapor introduced into the zth region at time step n due to heat exchange of cargo. Other variables have the same meanings as the above variables and will not be repeated here.

[0114] The heat transfer formula between the partitions is:

[0115]

[0116] In the formula, is the heat transfer between the z1th region and the z2th region at time step n, is the thermal resistance between the z1th region and the z2th region.

[0117] Similarly, the water vapor transfer formula between each partition is:

[0118]

[0119] In the formula, is the water vapor transfer between the z1th region and the z2th region at time step n, is the water vapor transfer resistance between the z1th region and the z2th region.

[0120] Subsequently, the short-term historical infiltration load is calculated based on the above temperature and humidity updates and transfer amounts. The short-term historical infiltration load is divided into the total infiltration heat load and the total infiltration water vapor load. The current physical model state vector is defined based on the short-term historical infiltration load. The current physical model state vector is composed of the temperature and humidity vectors of each partition, the relative humidity vectors of each partition, the heat infiltration load vectors of each partition, and the water vapor infiltration load vectors of each partition. Among them, a temperature vector is composed of temperature data of different regions at the same time step, a humidity vector is composed of humidity data of different regions at the same time step, and a heat infiltration load vector and a water vapor infiltration load vector are respectively composed of heat infiltration loads and water vapor infiltration loads of different regions at the same time step.

[0121] Step 3: Use CNN to predict short-term infiltration load changes and output short-term infiltration load prediction results

[0122] Specifically, the recent infiltration load history sequence output from step 2 and the current regional status are input into the convolutional neural network (CNN) model, and the short-term features are extracted through time series convolution to predict the changes in the door opening and closing infiltration heat and moisture load in the future short period of time (such as 5 to 10 minutes later).

[0123] First, it is necessary to define the feature matrix, combining the temperature vector and humidity vector in the current physical model state into a two-dimensional matrix. One column consists of temperature data of multiple regions at the same time step, and the other column consists of humidity data of multiple regions at the same time step. The total number of regions is the total number of partitions in the cold chain logistics center. In order to capture the time series characteristics, the physical model state of the most recent multiple time steps is selected as input. For example, 10 time steps represent ten minutes, and the time series input feature matrix is ​​obtained.

[0124] Secondly, define the layer parameters of the CNN model, including the convolution kernel size K×K, the number of convolution kernels, the stride, and padding. Apply the convolution kernel to the time series input feature matrix to generate the corresponding feature map. The feature map output after convolution can also be pooled to reduce the dimension of the feature map and reduce the amount of calculation. Subsequently, the alignment feature map is converted into a one-dimensional vector to obtain a flattened feature vector, which is then input into the fully connected layer to further extract high-level features. Two output nodes are defined in the output layer to predict the future short-term heat infiltration load and water vapor infiltration load respectively. Linear or nonlinear activation functions can also be applied to the output layer to limit the output range as needed. Finally, define the prediction result vector, which is the short-term heat infiltration load and water vapor infiltration load prediction output, and store the output prediction results in a structured data format, such as CSV, JSON, etc., for easy calling in subsequent steps.

[0125] Step 4: Use Transformer to predict the multi-region energy consumption distribution and output the multi-region energy consumption distribution prediction data.

[0126] Specifically, the short-term infiltration load prediction result output from step 3 and the cargo thermal parameters and partition status retained at the same time are input into the Transformer model. The Transformer combines the short-term infiltration prediction with the cargo distribution and regional characteristics through the attention mechanism to predict the dynamic energy consumption distribution and linkage relationship of each partition in the future period.

[0127] First, the heat permeation load and water vapor permeation load are combined with the cargo distribution data to form an input feature matrix, and then the regional characteristic data is combined with the input feature matrix to form a complete input feature matrix. Subsequently, each column feature in the complete input feature matrix is ​​standardized to have a mean of 0 and a standard deviation of 1. The cargo category number is then normalized to the range of [0,1].

[0128] Select a time window of fixed length to capture time series features, apply a sliding window to the entire data set corresponding to the standardized and normalized input feature matrix, and generate an input sequence set consisting of multiple input sequences. Define the model architecture of the Transformer model, including the encoder layer, decoder layer, and output layer. The parameter settings include the number of attention heads, embedding dimension, feedforward network dimension, and number of layers. The encoder layer includes a multi-head self-attention mechanism, a feedforward neural network, residual connections, and layer normalization. The decoder layer includes a multi-head self-attention mechanism, an encoder-decoder attention mechanism, a feedforward neural network, residual connections, and layer normalization. The output layer includes a fully connected layer to generate multi-region energy consumption distribution prediction data.

[0129] During the training of the Transformer model, the target output of the Transformer model is defined as the multi-region energy consumption distribution prediction data, the mean square error MSE is used as the loss function, the Adam optimizer is used to update the parameters, the prediction value is calculated through forward propagation, the model parameters are updated through back propagation, and the iteration is repeated until the loss function converges or the preset number of training rounds is reached. After that, the trained Transformer model is used to predict the new input sequence, output the predicted multi-region energy consumption distribution data, and store the prediction results in a structured data format for easy calling in subsequent steps.

[0130] Step 5: Model predictive control (MPC) strategy optimization, outputting the control strategy parameter set.

[0131] Specifically, the multi-region energy consumption distribution prediction data output from step 4 is used as the input of MPC. MPC calculates the optimal control strategy within the rolling time window, including refrigeration equipment power setting, fan speed and valve adjustment, and access control pre-control strategy to meet the low temperature requirements and reduce the overall energy consumption.

[0132] First, determine the control variables, including refrigeration equipment power setting, fan speed, valve opening, and access control pre-control strategy, and build a control strategy parameter set based on these control variables. In the process of establishing the MPC optimization model, define the minimization of comprehensive energy consumption as the objective function, define constraints, including temperature control constraints, refrigeration equipment power limits, fan speed limits, valve opening limits, and access control strategy constraints, and then complete the construction of the MPC optimization model. Subsequently, select a linear programming or nonlinear programming solver, and use the selected optimization solver to solve the defined MPC optimization model to obtain a control strategy parameter set. Then extract each control variable from the control strategy parameter set, and send the parameter set to the refrigeration equipment, fans, and valves in each partition for real-time control. The system status after implementation can also be monitored to ensure the effective execution of the control strategy, and the mobile phone feedback data is used for subsequent iterative optimization.

[0133] Step 6: Use the slime mold optimization algorithm (SMA) to optimize global parameters and output the final strategy solution.

[0134] Specifically, the preliminary control strategy parameters of step 5 are used as the initial solution of SMA. SMA searches for the global optimal solution in the high-dimensional nonlinear parameter space and further optimizes the cooling scheme, access control and load distribution parameters to obtain lower energy consumption and higher system stability.

[0135] First, define the optimization variables, i.e., refrigeration equipment power setting, fan speed, valve opening, and access control pre-control strategy, set SMA parameters, including population size, maximum number of iterations, learning factor, and random factor, and generate the initial population according to the control strategy parameter set output in step 5. Secondly, define the minimization of comprehensive energy consumption as the objective function, define temperature control constraints, refrigeration equipment power limits, fan speed limits, valve opening limits, and access control strategy constraints as constraints, iteratively update the slime mold position and calculate the fitness of each individual, and update the individual with the smallest fitness as the optimal solution by comparing the fitness of all individuals. If the maximum number of iterations or the fitness change is less than the preset threshold, terminate the algorithm, obtain the optimal control strategy parameter set, and extract each control variable from the optimal control strategy parameter set to generate the final strategy plan. Finally, send control instructions based on the final strategy plan, and send the optimal control strategy set to each partition quota refrigeration equipment, fan, valve, and access control system for real-time control.

[0136] The central air-conditioning system energy consumption simulation device provided by the present invention is described below. The central air-conditioning system energy consumption simulation device described below and the central air-conditioning system energy consumption simulation method based on the multi-region coupling model described above can be referenced to each other.

[0137] In one embodiment, a central air conditioning system energy consumption simulation device includes a data preprocessing module, a physical model building module, a short-term penetration load prediction module, a multi-region energy consumption distribution prediction module, a control strategy optimization module, and an optimal control strategy generation module.

[0138] The data preprocessing module is used to obtain the basic data and initial parameters of the cold chain logistics center site, and preprocess the basic data and initial parameters.

[0139] The physical model building module is used to build a multi-region coupled physical mechanism model based on the heat and mass transfer equations, and uses the preprocessed basic data and initial parameters as boundary conditions to output the physical model state through the finite difference method.

[0140] The short-term seepage load prediction module is used to take the physical model state as the input of the convolutional neural network to extract the short-period seepage load characteristics through time series convolution to output the short-term seepage load prediction results.

[0141] The multi-region energy consumption distribution prediction module is used to combine the short-term penetration load prediction results with the cargo thermal parameters and partition status as the input of the Transformer model, so as to combine the short-term penetration load prediction results with the cargo distribution and regional characteristics through the attention mechanism, and output the multi-region energy consumption distribution prediction data.

[0142] The control strategy optimization module is used to perform strategy optimization processing on the multi-region energy consumption distribution prediction data through MPC, so as to calculate the preliminary control strategy within the rolling time window.

[0143] The optimal control strategy generation module is used to use the preliminary control strategy as the initial solution of the slime mold optimization algorithm to search for the global optimal solution in the high-dimensional nonlinear parameter space, output the optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center.

[0144] Among them, the basic data and initial parameters include access control opening and closing records, outdoor meteorological conditions, cargo types and storage temperature, refrigeration equipment parameters, and temperature and humidity of each partition. Preprocessing includes data cleaning, data synchronization and time alignment, outlier correction or elimination, data interpolation and missing value filling. The physical model state includes the temperature and humidity field of each partition of the cold chain logistics center at the current moment, the heat exchange status of the cargo, and the short-term historical penetration load records. The short-term penetration load prediction result is the change of door opening and closing penetration heat and humidity load in the first time period in the future. The optimal control strategy includes refrigeration equipment power setting, fan speed and valve adjustment, and access control pre-control strategy.

[0145] In this embodiment, the central air conditioning system energy consumption simulation device provided by the present invention, the data preprocessing module is specifically used for:

[0146] The access control opening and closing records are obtained through the access control sensor, including the opening and closing frequency and duration. The temperature and humidity of each partition and the outdoor meteorological conditions are obtained through the temperature and humidity sensor and the meteorological sensor. The outdoor meteorological conditions include outdoor temperature and relative humidity. The operating parameters of the refrigeration equipment are obtained through the equipment parameter sensor.

[0147] The data collected by different sensors are aligned according to timestamps to synchronize and time-align the basic data and initial parameters, and outlier detection is performed on the sensor data to correct or eliminate the detected outliers, and the missing data points are filled by linear interpolation to obtain the preprocessed basic data and initial parameters.

[0148] In this embodiment, the central air conditioning system energy consumption simulation device provided by the present invention, the physical model building module is specifically used for:

[0149] The cold chain logistics center is divided into multiple control areas, and the physical parameters of each control area are determined. The energy conservation equation and mass conservation equation of each area are time discretized by explicit finite difference method, and the discretized heat and mass transfer equation is established. The physical parameters include volume, air density, specific heat capacity and air flow resistance. The expression of heat and mass transfer equation is:

[0150]

[0151]

[0152] Where n is the current time step, is the air density of the z-th control area, is the volume of the z-th control region, is the specific heat capacity of air in the zth control area, and is the input and output heat of the z-th control area at time step n, and is the input and output water vapor of the zth control area at time step n, and is the humidity of the zth region at time steps n and n+1, and is the water vapor content in the zth region at time steps n and n+1.

[0153] The access control penetration load and cargo heat exchange load are selected as boundary conditions from the preprocessed basic data and initial parameters, and the temperature and humidity fields of each partition and the short-term historical penetration load are calculated to output the physical model state.

[0154] The physical model state is composed of the temperature vector of each partition, the relative humidity vector of each partition, the heat penetration load vector of each partition, and the water vapor penetration load vector of each partition.

[0155] In this embodiment, the central air conditioning system energy consumption simulation device provided by the present invention, the short-term penetration load prediction module is specifically used for:

[0156] The temperature vector and the relative humidity vector in the physical model state are combined into a two-dimensional matrix, and the physical model states of multiple time steps are selected to construct a time series input feature matrix, which is composed of multiple two-dimensional matrices.

[0157] The time series input feature matrix is ​​used as the input of the convolutional neural network to perform a convolution operation on the time series input feature matrix to generate a feature map, and the feature map is converted into a one-dimensional vector as the input of the fully connected layer to output a high-level feature vector.

[0158] Two output nodes are defined in the output layer of the convolutional neural network to predict the future short-term heat infiltration load and water vapor infiltration load respectively to obtain the infiltration load prediction result vector, and the prediction result vector is structured and stored. The prediction result vector consists of the heat infiltration load and the water vapor infiltration load.

[0159] In this embodiment, the central air conditioning system energy consumption simulation device provided by the present invention, the multi-region energy consumption distribution prediction module is specifically used for:

[0160] The heat permeation load and the water vapor permeation load are combined with the corresponding cargo distribution data to generate a first characteristic matrix, and the regional characteristic data are combined with the first characteristic matrix to generate a second characteristic matrix.

[0161] Each column feature in the second feature matrix is ​​standardized, and the commodity type number in the second feature matrix is ​​normalized to obtain the third feature matrix. A time window of fixed length is selected, and an input sequence set is generated through a sliding window.

[0162] The input sequence set is used as the input of the trained Transformer model, and the multi-region energy consumption distribution prediction data is output through the encoder and decoder, and the multi-region energy consumption distribution prediction data is structured and stored.

[0163] Among them, the trained Transformer model is obtained by training the Transformer model based on the input sequence set and defining the multi-region energy consumption distribution prediction data as the target output, using the mean square error as the loss function and the Adam optimizer for parameter update.

[0164] In this embodiment, the central air conditioning system energy consumption simulation device provided by the present invention, the control strategy optimization module is specifically used for:

[0165] The control variables are obtained and a control strategy parameter set is constructed based on the control variables. The control variables include the power setting of the refrigeration equipment, the fan speed, the valve opening, and the access control pre-control strategy.

[0166] The control strategy parameter set and multi-region energy consumption distribution prediction data are used as the input of MPC, and the minimization of comprehensive energy consumption, temperature control constraints, refrigeration equipment power limitations, fan speed limitations, valve opening limitations, and access control strategy constraints are defined to build an MPC optimization model.

[0167] The MPC optimization model is solved by the optimization solver to obtain the initial control strategy parameter set, and the initial control strategy parameter set is sent to the refrigeration equipment, fans and valves in each partition to perform real-time control of the refrigeration equipment, fans and valves in each partition.

[0168] In this embodiment, the central air conditioning system energy consumption simulation device provided by the present invention, the optimal control strategy generation module is specifically used for:

[0169] The slime mold optimization algorithm is initialized by controlling variables, setting SAM parameters and initializing the population. SAM parameters include population size, maximum number of iterations, learning factor and random factor. The initialization population is generated based on the initial control strategy parameter set.

[0170] Based on the initialized slime mold optimization algorithm, the objective function is to minimize the comprehensive energy consumption, and the temperature control constraint, refrigeration equipment power limit, fan speed limit, valve opening limit and access control strategy constraint are defined as constraint conditions. The position of the slime mold is updated and the fitness of each individual is calculated. By comparing the fitness of all individuals, the individual with the smallest fitness is updated as the optimal solution, and the optimal control strategy parameter set is obtained.

[0171] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0172] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the above. The computer-readable storage medium used herein is not to be interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0173] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0174] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0175] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0176] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0177] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0178] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model, characterized in that: The method comprises: Obtaining basic data and initial parameters at the cold chain logistics center site, and preprocessing the basic data and initial parameters; A multi-region coupled physical mechanism model is constructed based on the heat and mass transfer equations, and the pre-processed basic data and initial parameters are used as boundary conditions to output the physical model state through the finite difference method; The physical model state is used as an input of a convolutional neural network to extract short-period seepage load characteristics through time series convolution to output a short-term seepage load prediction result; The short-term infiltration load prediction results are combined with cargo thermal parameters and partition status as inputs to the Transformer model, so as to combine the short-term infiltration load prediction results with cargo distribution and regional characteristics through the attention mechanism and output multi-region energy consumption distribution prediction data; Performing strategy optimization processing on the multi-region energy consumption distribution prediction data through MPC to calculate a preliminary control strategy within a rolling time window; The preliminary control strategy is used as the initial solution of the slime mold optimization algorithm to search for a global optimal solution in a high-dimensional nonlinear parameter space, output an optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center site; Among them, the basic data and initial parameters include access control opening and closing records, outdoor meteorological conditions, cargo types and storage temperatures, refrigeration equipment parameters, and temperature and humidity of each partition; the preprocessing includes data cleaning, data synchronization and time alignment, outlier correction or elimination, data interpolation, and missing value filling; the physical model state includes the temperature and humidity field of each partition of the cold chain logistics center at the current moment, the cargo heat exchange status, and the short-term historical penetration load record; the short-term penetration load prediction result is the change of the door opening and closing penetration heat and humidity load in the first time period in the future; the optimal control strategy includes refrigeration equipment power setting, fan speed and valve adjustment, and access control pre-control strategy; The obtaining of basic data and initial parameters at the cold chain logistics center site and preprocessing of the basic data and initial parameters include: The access control opening and closing records are obtained through the access control sensor, and the access control opening and closing records include the opening and closing frequency and duration; the temperature and humidity of each partition and the outdoor meteorological conditions are obtained through the temperature and humidity sensor and the meteorological sensor, and the outdoor meteorological conditions include the outdoor temperature and relative humidity; and the operating parameters of the refrigeration equipment are obtained through the equipment parameter sensor; Aligning the data collected by different sensors according to timestamps to synchronize and time-align the basic data and initial parameters, performing outlier detection on the sensor data to correct or remove the detected outliers, and filling in missing data points using linear interpolation to obtain preprocessed basic data and initial parameters; The multi-region coupling physical mechanism model is constructed based on the heat and mass transfer equations, and the pre-processed basic data and initial parameters are used as boundary conditions to output the physical model state through the finite difference method, including: The cold chain logistics center site is divided into multiple control areas, and the physical parameters of each control area are determined. The energy conservation equation and mass conservation equation of each area are time discretized by using the explicit finite difference method to establish the discretized heat and mass transfer equation. The physical parameters include volume, air density, specific heat capacity and air flow resistance. The expression of the heat and mass transfer equation is: Where n is the current time step, is the air density of the z-th control area, is the volume of the z-th control region, is the specific heat capacity of air in the zth control area, and is the input and output heat of the z-th control area at time step n, and is the input and output water vapor of the zth control area at time step n, and is the humidity of the zth region at time steps n and n+1, and is the water vapor content in the zth region at time steps n and n+1; Selecting access control penetration load and cargo heat exchange load as boundary conditions from the preprocessed basic data and initial parameters, calculating the temperature and humidity field of each partition and the short-term historical penetration load to output the physical model state; The physical model state is composed of the temperature vector of each partition, the relative humidity vector of each partition, the heat penetration load vector of each partition, and the water vapor penetration load vector of each partition.

2. The method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model according to claim 1 is characterized in that: The physical model state is used as the input of the convolutional neural network to extract the short-period seepage load characteristics through time series convolution to output the short-term seepage load prediction result, including: Combining the temperature vector and the relative humidity vector in the physical model state into a two-dimensional matrix, and selecting the physical model states of multiple time steps to construct a time series input feature matrix, wherein the time series input feature matrix is ​​composed of multiple two-dimensional matrices; Using the time series input feature matrix as the input of the convolutional neural network to perform a convolution operation on the time series input feature matrix to generate a feature map, and converting the feature map into a one-dimensional vector as the input of a fully connected layer to output a high-level feature vector; Two output nodes are defined in the output layer of the convolutional neural network to respectively predict the future short-term heat permeation load and water vapor permeation load to obtain the permeation load prediction result vector, and the prediction result vector is structured and stored. The prediction result vector is composed of the heat permeation load and the water vapor permeation load.

3. The method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model according to claim 2 is characterized in that: The short-term penetration load prediction result is combined with the cargo thermal parameters and the partition status as the input of the Transformer model, so as to combine the short-term penetration load prediction result with the cargo distribution and regional characteristics through the attention mechanism, and output multi-region energy consumption distribution prediction data, including: Combining the heat permeation load and the water vapor permeation load with corresponding cargo distribution data to generate a first characteristic matrix, and combining the regional characteristic data with the first characteristic matrix to generate a second characteristic matrix; Standardize each column feature in the second feature matrix, normalize the commodity type number in the second feature matrix to obtain a third feature matrix, select a time window of fixed length, and generate an input sequence set through a sliding window; Using the input sequence set as the input of the trained Transformer model, outputting the multi-region energy consumption distribution prediction data through an encoder and a decoder, and performing structured storage on the multi-region energy consumption distribution prediction data; Among them, the trained Transformer model is obtained by training the Transformer model based on the input sequence set and defining multi-region energy consumption distribution prediction data as the target output, using mean square error as the loss function and adopting Adam optimizer for parameter update.

4. The method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model according to claim 3 is characterized in that: The performing strategy optimization processing on the multi-region energy consumption distribution prediction data by MPC to calculate a preliminary control strategy within a rolling time window includes: Acquire control variables and construct a control strategy parameter set based on the control variables, wherein the control variables include refrigeration equipment power setting, fan speed, valve opening, and access control pre-control strategy; The control strategy parameter set and multi-region energy consumption distribution prediction data are used as inputs of the MPC, and the minimization of comprehensive energy consumption, temperature control constraints, refrigeration equipment power limits, fan speed limits, valve opening limits, and access control strategy constraints are defined to construct an MPC optimization model; The MPC optimization model is solved by an optimization solver to obtain an initial control strategy parameter set, and the initial control strategy parameter set is sent to the refrigeration equipment, fans and valves in each partition to perform real-time control on the refrigeration equipment, fans and valves in each partition.

5. The method for simulating energy consumption of a central air-conditioning system based on a multi-region coupling model according to claim 4 is characterized in that: The method uses the preliminary control strategy as the initial solution of the slime mold optimization algorithm to search for the global optimal solution in the high-dimensional nonlinear parameter space, outputs the optimal control strategy, and optimizes the control of the central air-conditioning system at the cold chain logistics center, including: Initializing the slime mold optimization algorithm through the control variables, the set SAM parameters and the initialization population, wherein the SAM parameters include the population size, the maximum number of iterations, the learning factor and the random factor, and the initialization population is generated based on the initial control strategy parameter set; Based on the initialized slime mold optimization algorithm, the objective function is to minimize the comprehensive energy consumption, and the temperature control constraint, refrigeration equipment power limit, fan speed limit, valve opening limit and access control strategy constraint are defined as constraint conditions. The position of the slime mold is updated and the fitness of each individual is calculated. By comparing the fitness of all individuals, the individual with the smallest fitness is updated as the optimal solution, and the optimal control strategy parameter set is obtained.

6. A central air conditioning system energy consumption simulation device, used to execute the method according to any one of claims 1 to 5, characterized in that: The device comprises: A data preprocessing module is used to obtain basic data and initial parameters at the cold chain logistics center site and preprocess the basic data and initial parameters; A physical model building module is used to build a multi-region coupled physical mechanism model based on the heat and mass transfer equations, and output the physical model state through the finite difference method using the preprocessed basic data and initial parameters as boundary conditions; A short-term seepage load prediction module, used to use the physical model state as an input of a convolutional neural network to extract short-period seepage load characteristics through time series convolution to output a short-term seepage load prediction result; A multi-region energy consumption distribution prediction module is used to combine the short-term penetration load prediction results with the cargo thermal parameters and the partition status as the input of the Transformer model, so as to combine the short-term penetration load prediction results with the cargo distribution and regional characteristics through the attention mechanism, and output the multi-region energy consumption distribution prediction data; A control strategy optimization module, used for performing strategy optimization processing on the multi-region energy consumption distribution prediction data through MPC to calculate a preliminary control strategy within a rolling time window; An optimal control strategy generation module is used to use the preliminary control strategy as the initial solution of the slime mold optimization algorithm to search for a global optimal solution in a high-dimensional nonlinear parameter space, output an optimal control strategy, and optimize the control of the central air-conditioning system at the cold chain logistics center site; Among them, the basic data and initial parameters include access control opening and closing records, outdoor meteorological conditions, cargo types and storage temperature, refrigeration equipment parameters and temperature and humidity of each partition; the preprocessing includes data cleaning, data synchronization and time alignment, outlier correction or elimination, data interpolation and missing value filling; the physical model state includes the temperature and humidity field of each partition on site of the cold chain logistics center at the current moment, the cargo heat exchange status and short-term historical penetration load records; the short-term penetration load prediction result is the change of door opening and closing penetration heat and humidity load in the first time period in the future; the optimal control strategy includes refrigeration equipment power setting, fan speed and valve adjustment and access control pre-control strategy.

7. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images