Method, system, medium and computer program product for predicting cooling load of subway station air-conditioning system based on improved BPNN model

Abstract

The present invention provides a method, system, medium, and computer program product for predicting the cooling load of subway station air-conditioning systems based on an improved BPNN model. The method comprises: establishing a big data processing method for repairing missing values ​​and outliers based on subway station operating data, generating a database in hourly units; performing a correlation analysis on factors affecting the cooling load of subway stations to determine the main influencing factors; determining BPNN structural parameters based on the main influencing factors and the cooling load, and establishing a BPNN model; determining the optimal weights, thresholds, and number of hidden layer nodes of the BPNN model based on the Dayan optimization algorithm, obtaining an optimized BPNN model, and predicting the cooling load of air conditioners based on the optimized BPNN model. The method of the present invention can perform effective data processing and input screening based on a data-driven method for predicting the cooling load of subway stations, and achieve prediction in combination with historical cooling loads, thereby having the advantages of high accuracy, fast speed, strong adaptability, and good robustness.

Description

Technical Field

[0001] The present invention relates to the technical field of cooling load prediction of subway station air-conditioning systems, and in particular to a cooling load prediction method and system for subway station air-conditioning systems based on an improved BPNN model. Background Art

[0002] Cooling load forecasting provides clear direction and effective support for energy-efficient air conditioning system regulation. However, different buildings have varying cooling load characteristics and factors influencing them. Therefore, it is necessary to clearly define the cooling load influencing patterns and establish a prediction mechanism for each building type. Subway stations, as special types of buildings, require a highly targeted, widely applicable, and integrated load forecasting method and device to guide subsequent regulation.

[0003] Neural network models are widely used in cooling load forecasting, but their accuracy is dependent on input and internal hyperparameter settings. Existing methods for predicting subway station air conditioning system load, such as those based on BP neural networks and cluster optimization integrated neural networks, lack input screening, including variables with low correlation with cooling load, which reduces the neural network's prediction accuracy. Regarding hyperparameter settings, the random generation and step-by-step iteration logic inherent in neural networks can also lead to unstable prediction results and slow convergence when solving certain specific scenarios. Summary of the Invention

[0004] In response to the problems existing in the prior art, the purpose of the present invention is to provide a method and system for predicting the cooling load of subway station air-conditioning systems based on an improved BPNN model, so as to solve the problems of unstable model prediction results and slow convergence caused by low input correlation and improper setting of neural network hyperparameters in the existing subway station air-conditioning system load prediction methods.

[0005] According to a first aspect of the present invention, a method for predicting cooling load of air conditioners in subway stations based on an improved BPNN is proposed, comprising the following steps:

[0006] Based on subway station operation data, a big data processing method for missing values ​​and outlier repair is established to generate an hourly database.

[0007] Conduct a correlation analysis on factors affecting the cooling load of subway stations to determine the main influencing factors, including outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system in the 24 hours before the forecast time;

[0008] Determine the structural parameters of the BPNN network according to the main influencing factors and cooling load, and establish a BPNN model;

[0009] The optimal weights, thresholds and number of hidden layer nodes of the BPNN model are determined based on the Dayan optimization algorithm to obtain the optimized BPNN model. The air conditioning cooling load of the subway station at a certain moment is predicted based on the optimized BPNN model.

[0010] According to a second aspect of the present invention, a subway station air conditioning cooling load prediction system based on an improved BPNN is proposed, comprising:

[0011] A database generation module is configured to establish a big data processing method for repairing missing values ​​and outliers based on subway station operation data, and generate a database in hourly units;

[0012] a correlation analysis module configured to perform a correlation analysis on relevant factors affecting the cooling load of the subway station to determine the main influencing factors, wherein the relevant factors include: outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system in the 24 hours before the forecast time;

[0013] The BPNN model building module is configured to determine the structural parameters of the BPNN network according to the main influencing factors and the cooling load, and to establish the BPNN model;

[0014] The BPNN model optimization module is configured to determine the optimal weights, thresholds, and number of hidden layer nodes of the BPNN model based on the Dayan optimization algorithm, obtain the optimized BPNN model, and predict the air conditioning cooling load of the subway station at a certain moment based on the optimized BPNN model.

[0015] According to a third aspect of the present invention, a computer system is provided, comprising:

[0016] one or more processors;

[0017] A memory stores operable instructions, which, when executed by the one or more processors, cause the one or more processors to perform operations, including the process of executing the aforementioned subway station air conditioning cooling load prediction method based on the improved BPNN.

[0018] According to a fourth aspect of the present invention, a computer-readable medium storing software is proposed, wherein the software includes instructions that can be executed by one or more computers, and when the instructions are executed by the one or more computers, the process of the aforementioned subway station air conditioning cooling load prediction method based on the improved BPNN is performed.

[0019] According to a fifth aspect of the present invention, a computer program product is proposed, comprising a computer program, which, when executed by a processor, implements the steps of the aforementioned subway station air conditioning cooling load prediction method based on the improved BPNN.

[0020] Compared to existing technologies, Sunrise's improved BPNN-based subway station air conditioning cooling load forecasting method uses historical cooling load data as one of its inputs and provides a comprehensive load forecasting method for subway stations, including data processing, input analysis, model building, and model optimization. This provides an important reference for load forecasting of subway station buildings. Furthermore, the Dayan optimization algorithm is used to find the optimal hyperparameter settings for the BPNN, resulting in higher prediction accuracy, more stable prediction results, and faster convergence.

[0021] The method of the present invention can perform effective data processing and input screening based on a data-driven method for predicting subway station cooling load, and realize prediction in combination with historical cooling load, which has the advantages of high accuracy, fast speed, strong adaptability and good robustness in practical applications.

[0022] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below, as long as such concepts are not mutually inconsistent, can be considered part of the inventive subject matter of this disclosure. In addition, all combinations of the claimed subject matter are considered part of the inventive subject matter of this disclosure.

[0023] The foregoing and other aspects, embodiments, and features of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as features and / or beneficial effects of the exemplary embodiments, will become apparent from the following description or through practice of specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings are not intended to be drawn to scale. In the accompanying drawings, each identical or nearly identical component shown in various figures may be represented by the same reference numeral. For clarity, not every component is labeled in every figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings.

[0025] Figure 1 4 is a flowchart of a subway station air conditioning cooling load prediction method based on an improved BPNN according to an embodiment of the present invention.

[0026] Figure 2 2 is a schematic diagram of correlation coefficients of candidate factors according to an embodiment of the present invention.

[0027] Figure 3 1 is a block diagram of a BPNN structure according to an embodiment of the present invention.

[0028] Figure 4 2 is a flow chart of the Dayan optimization algorithm according to an embodiment of the present invention.

[0029] Figure 5 3 is a schematic diagram of the cooling load prediction effect for a certain day in summer according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to better understand the technical content of the present invention, specific embodiments are given and described below with reference to the accompanying drawings.

[0031] Various aspects of the present invention are described in this disclosure with reference to the accompanying drawings, in which a number of illustrative embodiments are shown. The embodiments of the present disclosure are not necessarily intended to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of many ways, because the concepts and embodiments disclosed herein are not limited to any embodiment. In addition, some aspects of the present disclosure may be used alone or in any appropriate combination with other aspects disclosed herein.

[0032] {Example 1}

[0033] Combine Figure 1-4 As shown, the subway station air conditioning cooling load prediction method based on the improved BPNN according to this embodiment includes the following steps:

[0034] Based on subway station operation data, a big data processing method for missing values ​​and outlier repair is established to generate an hourly database.

[0035] Conduct a correlation analysis on factors affecting the cooling load of subway stations to determine the main influencing factors, including outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system in the 24 hours before the forecast time;

[0036] Determine the structural parameters of the BPNN network according to the main influencing factors and cooling load, and establish a BPNN model;

[0037] The optimal weights, thresholds and number of hidden layer nodes of the BPNN model are determined based on the Dayan optimization algorithm to obtain the optimized BPNN model. The air conditioning cooling load of the subway station at a certain moment is predicted based on the optimized BPNN model.

[0038] As an optional implementation, the method for processing missing values ​​and outliers based on subway station operation data is established to generate a database in hourly units, including:

[0039] A data traversal program is established, which is configured to automatically divide the original subway station operation data into hourly units;

[0040] An outlier diagnostic program is established, which is configured to calculate the theoretical upper and lower limits of the running data based on the quartile method, and to find outliers that exceed the upper and lower limits according to the theoretical upper and lower limits;

[0041] Establish missing value and outlier repair procedures, which are set to repair missing values ​​and outliers based on weighted average;

[0042] According to the above three procedures, the repaired database in hours is obtained, and the average value of each factor is calculated and the processed database is output.

[0043] As an optional implementation, the method of repairing missing values ​​and outliers based on weighted average includes:

[0044] By calculating the first five data in the normal data whose other dimensions are similar to the missing data, and calculating based on the weighted average, the weighted average is used as the corresponding missing value day or abnormal value;

[0045] Among them, the search for similar data is based on Euclidean distance, and the calculation formula of Euclidean distance is:

[0046]

[0047] Where: d(p,q) is the Euclidean distance between vectors p and q, p i and q i are the i-th components of vectors p and q respectively;

[0048] For the identified outliers and missing values, repair is performed based on the weighted average of the first five data with the closest Euclidean distance to the data to be repaired in the remaining dimensions found in the database.

[0049] As an optional implementation, the correlation analysis of the factors affecting the cooling load of the subway station to determine the main influencing factors includes:

[0050] Calculate the Pearson correlation coefficient between outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system 24 hours before the forecast time and air conditioning cooling load;

[0051] Factors with a Pearson correlation coefficient greater than the preset threshold were selected as the main influencing factors.

[0052] As an optional implementation, determining the structural parameters of the BPNN network based on the main influencing factors and the cooling load and establishing the BPNN model includes:

[0053] The BPNN model is configured as a three-layer neural network, including an input layer, a hidden layer, and an output layer;

[0054] Among them, the input layer corresponds to the main influencing factors one by one, and the output layer is the cooling load prediction value;

[0055] The value of each node in the hidden layer is calculated based on the internal weight and threshold of the grid, and the number of nodes in the hidden layer is determined based on the number of nodes in the input layer and the output layer.

[0056] As an optional implementation, determining the optimal weights, thresholds, and number of hidden layer nodes of the BPNN model based on the Dayan optimization algorithm to obtain the optimized BPNN model, and predicting the air conditioning cooling load of the subway station at a certain time based on the optimized BPNN model includes:

[0057] Set the parameters of the Dayan optimization algorithm, including the maximum number of iterations M gen , population size N, inertia weight w, acceleration factors c1 and c2;

[0058] Initialize the first generation of geese, calculate the fitness of each individual in the first generation of geese, and find the best individual as the leader goose P best ;

[0059] Generate random numbers and update the speed and position of each goose in the first generation population;

[0060] Update individual fitness and leader goose, when the number of iterations reaches the maximum number of iterations M gen When , the position and fitness of the final leader goose of the goose optimization algorithm are output;

[0061] Repeat the above steps to find the position and fitness of the leader goose under each possible number of hidden layer nodes of BPNN;

[0062] Based on the fitness, the optimal leader goose position and the corresponding number of hidden layer nodes are found, and the BPNN model is set and assigned to obtain the optimized BPNN model.

[0063] {Example 2}

[0064] In conjunction with the implementation process of the improved BPNN-based subway station air conditioning cooling load prediction method in the above embodiment, we use a subway station in a certain city as an example in this example to more specifically illustrate the implementation of the method of the present invention, which includes the following process:

[0065] Step 101: Based on the subway station operation data, a big data processing method for repairing missing values ​​and outliers is established to generate a database in hourly units;

[0066] Step 102: performing a correlation analysis on the factors affecting the cooling load of the subway station to determine the main factors, wherein the aforementioned related factors are selected as: outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system in the 24 hours before the forecast;

[0067] Step 103: Determine the structural parameters of the BPNN based on the main factors and the cooling load, and establish a BPNN model; the model is a three-layer neural network, including an input layer, a hidden layer, and an output layer;

[0068] Step 104: Determine the optimal weight, threshold, and number of hidden layer nodes of the BPNN using the Dayan optimization algorithm, and predict the cooling load based on the optimized BPNN.

[0069] Furthermore, in step 101, based on the historical operation data of a certain subway station, a big data processing method for repairing missing values ​​and outliers is established to generate a database in hourly units. The specific operation process includes:

[0070] (1) Establish a data traversal program to automatically divide the data of the original data set into hourly units;

[0071] (2) Establish an outlier diagnosis procedure, calculate the theoretical upper and lower limits of the data set based on the quartile method, and find outliers based on the theoretical upper and lower limits;

[0072] (3) Establish a missing value and outlier repair procedure by calculating the top five data in the normal data whose remaining dimensions are similar to the missing data, and repair the missing and outlier values ​​based on the weighted average;

[0073] (4) Calculate the average value of the processed data within each hour and output the data for each hour to the database.

[0074] As an optional implementation method, the theoretical upper limit and the theoretical lower limit are calculated as follows:

[0075] U1=X 0.75 +1.5*(X 0.75 -X 0.25 )

[0076] L1=X 0.25 -1.5*(X 0.75 -X 0.25 )

[0077] Where: U1 and L1 are the reasonable upper and lower limits of normal data respectively; X 0.75 and X 0.25 are the upper and lower quartiles in the processed data, respectively.

[0078] As an optional implementation, similar data is found based on Euclidean distance, and the calculation formula of Euclidean distance is:

[0079]

[0080] Where: d(p,q) is the Euclidean distance between vectors p and q, p i and q i are the i-th components of vectors p and q respectively.

[0081] For the identified outliers and missing values, they will be repaired based on the weighted average of the top five data with the closest Euclidean distance to the data to be repaired in the remaining dimensions found in the database.

[0082] Furthermore, the specific operation process of step 102 includes:

[0083] (1) Calculate the Pearson correlation coefficient between each factor and cooling load;

[0084] The calculation formula of Pearson correlation coefficient is:

[0085]

[0086] Where: ρ xy is the Pearson correlation coefficient between variables x and y; i and y i are the i-th sample points of variables x and y, respectively. ave and y ave are the mean values ​​of variables x and y, respectively, and n is the number of data logarithms of variables x and y;

[0087] (2) Taking July 25 to October 24, 2021 as an example, the correlation analysis of the hourly cooling load of the case subway station and the outdoor temperature, outdoor humidity, forecast time and the hourly cooling load of the system 24 hours before the forecast is carried out. The results are as follows: Figure 2 As shown in this example, the relevant factors with a Pearson correlation coefficient greater than the threshold of 0.4 are selected as the main influencing factors, combined with Figure 3 In the example shown, a total of nine factors, including outdoor temperature, forecast time, cooling load 1-3 hours before forecast, and cooling load 21-24 hours before forecast, are finally selected as the main factors.

[0088] Furthermore, the specific operation process of step 103 includes:

[0089] like Figure 3 As shown in the figure, the structural parameters of BPNN are determined according to the above 9 main factors and cooling load, and a BPNN model is established. The BPNN model is a three-layer neural network, including an input layer, a hidden layer, and an output layer.

[0090] Among them, the input layer corresponds to the nine main factors one by one, the output layer is the cooling load prediction value, the value of each node in the hidden layer is calculated based on the internal weight and threshold of the grid, and the number of nodes in the hidden layer is determined based on the number of nodes in the input and output layers.

[0091] Before training the neural network, the database data can be normalized. The normalization formula is:

[0092]

[0093] Where: x is the variable to be normalized, x max and x min are the maximum and minimum values ​​of the variable x, respectively.

[0094] Among them, after obtaining the prediction results based on BPNN, the predicted cooling load needs to be denormalized. The denormalization formula is:

[0095] CL p =CL model ·(CL max -CL min )+CL min

[0096] Where: CL p is the cooling load prediction value after denormalization, CL model is the cooling load prediction value obtained based on BPNN, CL max and CL min are the maximum and minimum values ​​of cooling load before normalization.

[0097] Further, if Figure 4 As shown, in step 104, the optimal weight, threshold, and number of hidden layer nodes of the BPNN are determined by the Dayan optimization algorithm, and the cooling load is predicted based on the optimized BPNN model. The specific operation process includes:

[0098] (1) Set the parameters of the Dayan optimization algorithm, including the maximum number of iterations M gen , population size N, inertia weight w, acceleration factors c1 and c2;

[0099] (2) Initialize the first generation of geese, calculate the fitness of each individual in the first generation of geese, and find the best individual as the leader goose P best ;

[0100] The evaluation criteria for the best individual are based on the error when the individual parameters are used for the weights and thresholds of BPNN to predict the cooling load. The cooling load prediction error is measured by MAE, and the MAE calculation formula is:

[0101]

[0102] Where: y i and y i,pre are the actual and predicted values ​​of cooling load, and n is the number of predicted sample points;

[0103] (3) Generate random numbers and update the speed and position of each goose in the first generation population;

[0104] Among them, for each goose, its speed update formula is:

[0105] V i (t+1)=w·V i (t)+c1·r1·(X leader (t)-X i (t))+c2·r2·(X pbest (t)-X i (t))

[0106] Where: V i (t) and V i (t+1) are the values ​​before and after the speed update, w is the inertia weight, c1 and c2 are acceleration factors, r1 and r2 are random numbers, X pbest (t) is the individual's historical optimal value, X leader (t) is the historical optimal value of the group;

[0107] Among them, for each goose, its position update formula is:

[0108] X i (t+1)=X i (t)+V i (t+1)

[0109] Where: X i (t) and X i (t+1) are the positions before and after the update respectively;

[0110] (4) Update individual fitness and leader goose, when the number of iterations reaches the maximum number of iterations M gen When , the position and fitness of the final leader goose of the goose optimization algorithm are output;

[0111] (5) Repeat the above steps to find the position and fitness of the leader goose under each possible number of hidden layer nodes of BPNN;

[0112] Among them, the calculation formula for the possible range of node numbers of BPNN is:

[0113]

[0114] Where: h is the number of hidden layer nodes of BPNN, m and n are the number of input layer and output layer nodes respectively, and a is a positive integer between 1 and 10;

[0115] (6) Based on the fitness, the optimal leader goose position and the corresponding number of hidden layer nodes are found, and the BPNN model is set and assigned to obtain the optimized BPNN model.

[0116] {Example 3}

[0117] In combination with the above embodiments, the present disclosure further proposes a subway station air conditioning cooling load prediction system based on an improved BPNN, comprising:

[0118] A database generation module is configured to establish a big data processing method for repairing missing values ​​and outliers based on subway station operation data, and generate a database in hourly units;

[0119] a correlation analysis module configured to perform a correlation analysis on relevant factors affecting the cooling load of the subway station to determine the main influencing factors, wherein the relevant factors include: outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system in the 24 hours before the forecast time;

[0120] The BPNN model building module is configured to determine the structural parameters of the BPNN network according to the main influencing factors and the cooling load, and to establish the BPNN model;

[0121] The BPNN model optimization module is configured to determine the optimal weights, thresholds, and number of hidden layer nodes of the BPNN model based on the Dayan optimization algorithm, obtain the optimized BPNN model, and predict the air conditioning cooling load of the subway station at a certain moment based on the optimized BPNN model.

[0122] It should be understood that the functions, outputs and specific implementation processes of the above-mentioned database generation module, correlation analysis module, BPNN model construction module and BPNN model optimization module can be implemented in the same manner as the corresponding process of the subway station air-conditioning cooling load prediction method based on improved BPNN in the aforementioned embodiment.

[0123] As an optional implementation, the BPNN model construction module is configured to construct the BPNN model in the following manner:

[0124] The BPNN model is configured as a three-layer neural network, including an input layer, a hidden layer, and an output layer;

[0125] Among them, the input layer corresponds to the main influencing factors one by one, and the output layer is the cooling load prediction value;

[0126] The value of each node in the hidden layer is calculated based on the internal weight and threshold of the grid, and the number of nodes in the hidden layer is determined based on the number of nodes in the input layer and the output layer.

[0127] As an optional embodiment, the BPNN model optimization module is configured to obtain an optimized BPNN model based on the following method for predicting the air conditioning cooling load of a subway station at a certain moment, including:

[0128] Set the parameters of the Dayan optimization algorithm, including the maximum number of iterations M gen , population size N, inertia weight w, acceleration factors c1 and c2;

[0129] Initialize the first generation of geese, calculate the fitness of each individual in the first generation of geese, and find the best individual as the leader goose P best ;

[0130] Generate random numbers and update the speed and position of each goose in the first generation population;

[0131] Update individual fitness and leader goose, when the number of iterations reaches the maximum number of iterations M gen When , the position and fitness of the final leader goose of the goose optimization algorithm are output;

[0132] Repeat the above steps to find the position and fitness of the leader goose under each possible number of hidden layer nodes of BPNN;

[0133] Based on the fitness, the optimal leader goose position and the corresponding number of hidden layer nodes are found, and the BPNN model is set and assigned to obtain the optimized BPNN model.

[0134] {Example 4}

[0135] In combination with the above embodiments, the present disclosure further provides a computer system, including:

[0136] one or more processors;

[0137] Memory stores instructions that can be operated.

[0138] Wherein, when the instruction is executed by the one or more processors, the one or more processors perform an operation, and the operation includes the process of executing the subway station air conditioning cooling load prediction method based on improved BPNN in any of the aforementioned embodiments.

[0139] {Example 5}

[0140] A computer-readable medium storing software, characterized in that the software includes instructions that can be executed by one or more computers, and when the instructions are executed by the one or more computers, the process of the subway station air conditioning cooling load prediction method based on improved BPNN is performed in any embodiment.

[0141] {Example 6}

[0142] A computer program product includes a computer program, characterized in that when the computer program is executed by a processor, it implements the steps of the subway station air conditioning cooling load prediction method based on improved BPNN in any of the aforementioned embodiments.

[0143] While the present invention has been disclosed above with reference to preferred embodiments, this is not intended to limit the present invention. Persons skilled in the art will readily appreciate that various modifications and variations can be made without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A subway station air conditioning cooling load prediction method based on improved BPNN, characterized in that: The following processes are included: Based on subway station operation data, a big data processing method for missing values ​​and outlier repair is established to generate an hourly database. Conduct a correlation analysis on factors affecting the cooling load of subway stations to determine the main influencing factors, including outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system in the 24 hours before the forecast time; Determine the structural parameters of the BPNN network according to the main influencing factors and cooling load, and establish a BPNN model; The optimal weights, thresholds, and number of hidden layer nodes of the BPNN model are determined based on the Dayan optimization algorithm. The optimized BPNN model is then used to predict the cooling load of the subway station air conditioner at a certain time. The method of establishing a big data processing method for repairing missing values ​​and outliers based on subway station operation data and generating a database in hourly units includes: A data traversal program is established, which is configured to automatically divide the original subway station operation data into hourly units; An outlier diagnostic program is established, which is configured to calculate the theoretical upper and lower limits of the running data based on the quartile method, and to find outliers that exceed the upper and lower limits according to the theoretical upper and lower limits; Establish missing value and outlier repair procedures, which are set to repair missing values ​​and outliers based on weighted average; According to the above three procedures, the repaired database in hours is obtained, and the average value of each factor is calculated and the processed database is output; The correlation analysis of the factors affecting the cooling load of the subway station is carried out to determine the main influencing factors, including: Calculate the Pearson correlation coefficient between outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system 24 hours before the forecast time and air conditioning cooling load; Factors with a Pearson correlation coefficient greater than the preset threshold were selected as the main influencing factors; The method of determining the structural parameters of the BPNN network according to the main influencing factors and the cooling load and establishing the BPNN model includes: The BPNN model is configured as a three-layer neural network, including an input layer, a hidden layer, and an output layer; Among them, the input layer corresponds to the main influencing factors one by one, and the output layer is the cooling load prediction value; The value of each node in the hidden layer is calculated based on the internal weight and threshold of the grid, and the number of nodes in the hidden layer is determined based on the number of nodes in the input layer and the output layer.

2. The subway station air conditioning cooling load prediction method based on improved BPNN according to claim 1 is characterized in that: The method of repairing missing values ​​and outliers based on weighted average includes: By calculating the first five data in the normal data that are similar to the missing data in other dimensions, and calculating based on the weighted average, the weighted average is used as the corresponding missing value day or abnormal value; Among them, the search for similar data is based on Euclidean distance, and the calculation formula of Euclidean distance is: ； Where: d ( p , q ) is a vector p and q The Euclidean distance, p i and q i They are vectors p and q The i-th component of ; For the identified outliers and missing values, repair is performed based on the weighted average of the first five data with the closest Euclidean distance to the data to be repaired in the remaining dimensions found in the database.

3. The subway station air conditioning cooling load prediction method based on improved BPNN according to claim 1 is characterized in that: The method of determining the optimal weight, threshold, and number of hidden layer nodes of the BPNN model based on the Dayan optimization algorithm to obtain the optimized BPNN model and predicting the air conditioning cooling load of the subway station at a certain time based on the optimized BPNN model includes: Set the parameters of the Dayan optimization algorithm, including the maximum number of iterations M gen , population size N , inertia weight w , acceleration factor c 1 and c 2; Initialize the first generation of geese, calculate the fitness of each individual in the first generation of geese, and find the best individual as the leader goose P best ; Generate random numbers and update the speed and position of each goose in the first generation population; Update individual fitness and leader goose, when the number of iterations reaches the maximum number of iterations M gen When , the position and fitness of the final leader goose of the goose optimization algorithm are output; Repeat the above steps to find the position and fitness of the leader goose under each possible number of hidden layer nodes of BPNN; Based on the fitness, the optimal leader goose position and the corresponding number of hidden layer nodes are found, and the BPNN model is set and assigned to obtain the optimized BPNN model.

4. A subway station air conditioning cooling load prediction system based on an improved BPNN, applying the subway station air conditioning cooling load prediction method based on an improved BPNN according to claim 1, characterized in that: include: A database generation module is configured to establish a big data processing method for repairing missing values ​​and outliers based on subway station operation data, and generate a database in hourly units; a correlation analysis module configured to perform a correlation analysis on relevant factors affecting the cooling load of the subway station to determine the main influencing factors, wherein the relevant factors include: outdoor temperature, outdoor humidity, forecast time, and hourly cooling load of the system in the 24 hours before the forecast time; The BPNN model building module is configured to determine the structural parameters of the BPNN network according to the main influencing factors and the cooling load, and to establish the BPNN model; The BPNN model optimization module is configured to determine the optimal weights, thresholds, and number of hidden layer nodes of the BPNN model based on the Dayan optimization algorithm, obtain the optimized BPNN model, and predict the air conditioning cooling load of the subway station at a certain moment based on the optimized BPNN model.

5. The subway station air conditioning cooling load prediction system based on improved BPNN according to claim 3 is characterized in that: The BPNN model construction module is configured to construct a BPNN model in the following manner: The BPNN model is configured as a three-layer neural network, including an input layer, a hidden layer, and an output layer; Among them, the input layer corresponds to the main influencing factors one by one, and the output layer is the cooling load prediction value; The value of each node in the hidden layer is calculated based on the internal weight and threshold of the grid, and the number of nodes in the hidden layer is determined based on the number of nodes in the input layer and the output layer.

6. The subway station air conditioning cooling load prediction system based on improved BPNN according to claim 3 is characterized in that: The BPNN model optimization module is configured to obtain an optimized BPNN model based on the following method for predicting the air conditioning cooling load of a subway station at a certain time, including: Set the parameters of the Dayan optimization algorithm, including the maximum number of iterations M gen , population size N , inertia weight w , acceleration factor c 1 and c 2; Initialize the first generation of geese, calculate the fitness of each individual in the first generation of geese, and find the best individual as the leader goose P best ; Generate random numbers and update the speed and position of each goose in the first generation population; Update individual fitness and leader goose, when the number of iterations reaches the maximum number of iterations M gen When , the position and fitness of the final leader goose of the goose optimization algorithm are output; Repeat the above steps to find the position and fitness of the leader goose under each possible number of hidden layer nodes of BPNN; Based on the fitness, the optimal leader goose position and the corresponding number of hidden layer nodes are found, and the BPNN model is set and assigned to obtain the optimized BPNN model.

7. A computer system, characterized in that: include: one or more processors; A memory storing operable instructions, wherein when the instructions are executed by the one or more processors, the one or more processors are caused to perform operations, wherein the operations include the process of executing the method according to any one of claims 1 to 3.

8. A computer-readable medium storing software, characterized in that: The software includes instructions that can be executed by one or more computers, and when the instructions are executed by the one or more computers, the process of the method according to any one of claims 1 to 3 is performed.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images