Central air conditioner performance optimization method and system based on whale algorithm and BP neural network

Abstract

The invention discloses a central air-conditioning performance optimization method and system based on a whale algorithm and a BP neural network, and the method comprises the steps: collecting and preprocessing operation data of a central air-conditioning system, including equipment operation parameters, environment parameters and energy consumption data, and obtaining stable operation data of the system; a Boruta feature selection algorithm is adopted for feature selection, the energy efficiency ratio serves as a performance index, the equipment operation parameters and the environment operation parameters serve as feature variables, the feature importance degree is calculated, and a feature subset is obtained; constructing a BP neural network prediction model, selecting the feature subset as model input, taking the performance index as model output, and carrying out model training; a whale algorithm is introduced to optimize the BP neural network prediction model; s5, evaluating the optimized performance indexes, finally obtaining control parameters when the maximum energy consumption ratio is achieved, and applying the control parameters to an actual control system of the central air conditioner; the energy consumption of the central air-conditioning system can be effectively reduced, and the performance and the energy efficiency ratio of the system are improved.

Description

Technical Field

[0001] This invention relates to the field of central air conditioning system performance optimization technology, and more specifically to a central air conditioning performance optimization method and system based on whale algorithm and BP neural network. Background Technology

[0002] With the acceleration of urbanization and the improvement of people's living standards, the energy consumption of central air conditioning systems has become increasingly prominent, becoming a common problem faced by cities around the world. On the one hand, building energy consumption is increasing year by year, and on the other hand, as the main source of building energy consumption, the optimization and energy saving of central air conditioning systems are of paramount importance.

[0003] Traditional control methods are mainly based on manual control, which has serious lag and cannot achieve real-time adjustment, resulting in a lot of energy waste. Air conditioning load has nonlinearity, non-stationarity and randomness, and is easily affected by weather factors. Traditional control methods are difficult to meet the precise control requirements under complex operating conditions.

[0004] In existing technologies, the control of central air conditioning systems is mostly based on empirical models or simple historical data analysis, which is difficult to adapt to complex dynamic operating environments. For example, traditional PID control methods often cannot adjust parameters in time when faced with rapid load changes, resulting in slow system response or increased energy consumption. In addition, some optimization methods based on physical models require a lot of prior knowledge and complex calculations, which are difficult to promote in practical applications.

[0005] In recent years, machine learning techniques have been introduced into the optimization of central air conditioning systems, such as BP neural networks and support vector machines. These methods, through data-driven approaches, can better fit the nonlinear characteristics of the system and improve the accuracy of prediction and control. However, existing machine learning methods still have shortcomings in feature selection and model optimization, resulting in limited model generalization ability and unsatisfactory optimization effects.

[0006] Therefore, how to improve the prediction accuracy and generalization ability of the model to adapt to the nonlinear characteristics and dynamic changes of the central air conditioning system, so as to reduce the energy consumption of the air conditioning system and improve the operating efficiency of the air conditioning system, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a central air conditioning performance optimization method and system based on whale algorithm and BP neural network. Through optimization by BP neural network and whale algorithm, it can better adapt to the nonlinear characteristics and dynamic changes of central air conditioning system, improve the prediction accuracy and generalization ability of model, effectively reduce the energy consumption of central air conditioning system, improve system performance and energy efficiency ratio, without the need for a large amount of prior knowledge and complex calculation, and has good practicality and promotion value.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A method for optimizing the performance of central air conditioning systems based on the whale algorithm and BP neural network includes the following steps:

[0010] S1. Collect the operating data of the central air conditioning system, including equipment operating parameters, environmental parameters and energy consumption data, and perform data preprocessing to obtain stable operating data of the central air conditioning system;

[0011] S2. The Boruta feature selection algorithm is used to select features, with energy efficiency ratio as the performance index, and equipment operating parameters and environmental operating parameters as feature variables. The feature importance of multiple feature variables is calculated to obtain a feature subset.

[0012] S3. Construct a BP neural network prediction model using stable operation data of the central air conditioning system, select a subset of features as model input, use performance indicators as model output, and train the model.

[0013] S4. Introduce the whale algorithm to optimize the BP neural network prediction model;

[0014] S5. Evaluate the optimized performance indicators to obtain the control parameters that maximize the energy efficiency ratio, and apply them to the actual control system of the central air conditioning system to achieve intelligent control of the central air conditioning system.

[0015] Preferably, data preprocessing includes cleaning and normalizing the operating data of the central air conditioning system, and using a Bayesian algorithm to fill in missing data to obtain stable operating data of the central air conditioning system.

[0016] Preferably, the specific content of using the Bayesian algorithm to fill in missing data is as follows:

[0017] The collected central air conditioning system operation data is divided into two parts: complete data and missing data, and the location of the missing values ​​is recorded.

[0018] Identify variables in the dataset, determine the probabilistic dependencies between variables, and construct a Bayesian network to represent the probabilistic dependencies;

[0019] Using the complete data portion, the conditional probability distribution of each variable in the Bayesian network is learned through Bayesian estimation methods to construct a Bayesian model;

[0020] Initialize the missing data by filling it with the mean, median, or random values ​​as an initial estimate;

[0021] For each missing value, the posterior probability distribution is calculated using the known data and Bayes' theorem.

[0022] Based on the posterior probability distribution, the value with the highest probability is selected as the estimate of the missing value.

[0023] Preferably, the specific content of step S2 is as follows:

[0024] Randomly shuffle each original feature data of the operating parameters and environmental operating parameters to generate corresponding shadow features;

[0025] The original features and shadow features are combined into a new dataset to train the random forest model;

[0026] The importance of each feature is calculated using a random forest model. The importance of each original feature is statistically tested using the modified Brownian bridge statistic to determine whether the original features are significantly more important than the shadow features. Insignificant features are removed, and a subset of features that have a significant impact on the energy efficiency ratio is retained.

[0027] Preferably, the specific content of step S3 is as follows:

[0028] S31. Select a feature subset as the model input. The feature subset includes chiller load rate, chilled water supply temperature, cooling water return temperature and outdoor temperature. Use the performance index energy efficiency ratio as the model output.

[0029] S32. Initialize network parameters, set network structure, and define the ReLU activation function; the network structure includes an input layer, hidden layers, and an output layer; the number of nodes in the input layer is equal to the number of feature subsets, and the number of nodes in the output layer is 1, i.e., the energy efficiency ratio; randomly initialize the network weights and biases, setting the weight range to [-1, 1].

[0030] S33. Input the central air conditioning system's operating data into the network for forward propagation. The input layer receives the normalized cooling unit load rate, chilled water supply temperature, cooling water return temperature, and outdoor temperature, and passes them to the hidden layer. The hidden layer calculates the weighted sum of the inputs and activates them using the ReLU function. The output of the hidden layer is then passed to the output layer, where the input and output of each neuron in the output layer are calculated. The output of the output layer is the model's prediction result.

[0031] S34. Using the mean squared error (MSE) as the loss function, calculate the error between the model output and the actual value, propagate the error back from the output layer to the input layer, adjust the network parameters, calculate the gradient of each weight, update the weights using gradient descent to minimize the error, and update the biases.

[0032] S35. Repeat steps S33-S34 for forward propagation, error calculation, back propagation, and parameter update until the convergence condition is met.

[0033] Preferably, the specific content of introducing the whale algorithm to optimize the BP neural network prediction model is as follows:

[0034] S41. Randomly initialize the locations of the whale population, with each individual whale representing a set of possible system operating parameters; when initializing the population, the range of each location component is consistent with the initialization of the BP network;

[0035] S42. Using the reciprocal of the prediction error as the fitness function, the prediction error of each individual whale is calculated through a BP neural network using the training dataset and used as the fitness value. The position of the whale is iteratively updated according to the fitness value. If the current individual is the optimal solution, the position is adjusted according to the spiral update formula.

[0036] S43. Check if the convergence condition is met. If it is met, stop the optimization. If it is not met, return to the fitness calculation step and continue the iteration.

[0037] Preferably, the fitness function is as follows:

[0038]

[0039] Where E is the mean squared error, which is the error between the model output and the actual value;

[0040] The whale's location is updated using the following method:

[0041] D = |C·X * (t)-X(t)|

[0042] X(t+1)=X * (t)-A·D

[0043] Where D is the distance vector between the current whale individual and the optimal solution, representing the relative position of the individual and the optimal solution in the search space; A and C are coefficient vectors used to control the direction and step size of the individual's movement; X * X(t) represents the position of the current optimal solution, i.e., the position of the individual with the highest fitness value in the population; X(t) represents the current position of the whale individual, and X(t+1) represents the new position of the whale individual after the update;

[0044] The coefficient vector is:

[0045] A = 2·a·r1-α

[0046] C = 2·r²

[0047] Where 'a' is a parameter that decreases linearly from 2 to 0, used to control the size of the search range; initially a = 2, and gradually decreases as iterations proceed, causing the search range to gradually shift from global to local, a = 2 - 2·(t / T) max ), T max Let r1 and r2 be the maximum number of iterations, and r1 and r2 be random vectors in the interval [0, 1].

[0048] A central air conditioning performance optimization system based on whale algorithm and BP neural network is deployed in a central air conditioning control system. It includes a data acquisition and preprocessing module, and a central air conditioning performance optimization module constructed using the aforementioned central air conditioning performance optimization method based on whale algorithm and BP neural network.

[0049] The data acquisition and preprocessing module is used to collect the operating data of the central air conditioning system, including equipment operating parameters, environmental parameters and energy consumption data, and to perform data preprocessing to obtain stable operating data of the central air conditioning system.

[0050] The central air conditioning performance optimization module is used to obtain the control parameters for maximizing the energy efficiency ratio based on the collected stable operation data of the central air conditioning system using a trained and optimized BP neural network prediction model, thereby realizing intelligent control of the central air conditioning system.

[0051] The central air conditioning performance optimization module includes a database, a feature selection unit, a model building and training unit, and a whale optimization unit;

[0052] The database is used to store historical stable operation data of the central air conditioning system after the data acquisition and preprocessing module has collected and processed it.

[0053] The feature selection unit is used to perform feature selection using the Boruta feature selection algorithm, selecting energy efficiency ratio as the performance index, equipment operating parameters and environmental operating parameters as feature variables, calculating the feature importance of multiple feature variables, and obtaining a feature subset.

[0054] The model building and training unit is used to build a BP neural network prediction model using stable operation data of the central air conditioning system, select a subset of features as model input, use performance indicators as model output, and train the model.

[0055] The whale optimization unit is used to introduce the whale algorithm to optimize the BP neural network prediction model, resulting in an optimized prediction model.

[0056] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for optimizing the performance of a central air conditioning system based on the whale algorithm and a BP neural network.

[0057] A processing terminal includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the aforementioned method for optimizing the performance of a central air conditioning system based on the whale algorithm and a BP neural network.

[0058] As can be seen from the above technical solution, compared with the prior art, this invention discloses a method and system for optimizing the performance of central air conditioning based on the whale algorithm and BP neural network. The method optimizes the BP neural network based on the whale algorithm and combines historical operating data and environmental parameters to predict the energy efficiency ratio in advance, dynamically adjust system operating parameters, and achieve intelligent optimization control of the central air conditioning system, reducing energy consumption and improving the energy efficiency ratio. Specifically: by simulating the hunting behavior of whales and dynamically adjusting network weights and biases, the prediction accuracy and generalization ability of the model can be effectively improved, adapting to the nonlinear characteristics and dynamic changes of the central air conditioning system; the Boruta feature selection algorithm is introduced to effectively screen the key feature subsets that affect the performance of the central air conditioning system, reducing data dimensionality and improving model training efficiency and generalization performance; the ReLU activation function is used to enhance the nonlinear fitting ability of the model, and gradient descent is combined for parameter updates to accelerate model convergence and improve prediction accuracy. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0060] Figure 1 A schematic diagram illustrating a central air conditioning performance optimization method based on the whale algorithm and BP neural network provided by the present invention;

[0061] Figure 2 A schematic diagram of central air conditioning system operation data provided in an embodiment of the present invention.

[0062] Figure 3 This is a schematic diagram of the operation data preprocessing provided by the present invention;

[0063] Figure 4 This is a schematic diagram illustrating the optimization of the whale algorithm provided by the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] This invention discloses a method for optimizing the performance of central air conditioning systems based on the whale algorithm and a backpropagation neural network. Figure 1 This includes the following steps:

[0066] S1. Collect the operating data of the central air conditioning system, including equipment operating parameters, environmental parameters and energy consumption data, and perform data preprocessing to obtain stable operating data of the central air conditioning system;

[0067] S2. The Boruta feature selection algorithm is used to select features, with energy efficiency ratio as the performance index, and equipment operating parameters and environmental operating parameters as feature variables. The feature importance of multiple feature variables is calculated to obtain a feature subset.

[0068] S3. Construct a BP neural network prediction model using stable operation data of the central air conditioning system, select a subset of features as model input, use performance indicators as model output, and train the model.

[0069] S4. Introduce the whale algorithm to optimize the BP neural network prediction model;

[0070] S5. Evaluate the optimized performance indicators to obtain the control parameters that maximize the energy efficiency ratio, and apply them to the actual control system of the central air conditioning system to achieve intelligent control of the central air conditioning system.

[0071] In this embodiment, the equipment operating parameters include chiller unit load rate, chilled water supply temperature, cooling water return temperature, water pump frequency, and cooling tower fan frequency; environmental parameters include outdoor temperature and humidity; and energy consumption data includes chiller unit energy consumption, chilled water pump energy consumption, cooling water pump energy consumption, and cooling tower energy consumption. For example, taking the collected operating data of the central air conditioning system of an office building during the cooling season, collected every hour, for example, from 9:00 to 18:00 on a certain day, such as... Figure 2 .

[0072] To further implement the above technical solutions, such as Figure 3 Data preprocessing includes cleaning and normalizing the operating data of the central air conditioning system, and using Bayesian algorithms to fill in missing data to obtain stable operating data of the central air conditioning system.

[0073] In this embodiment, the cleaning of the central air conditioning system's operating data involves: using boxplot analysis to remove outlier data; based on the interquartile range of the data distribution, data exceeding 1.5 times the interquartile range are considered outliers; comparing the time and value of the data records, if multiple records are highly similar in both time and value, only one is retained to ensure the conciseness of the dataset.

[0074] To ensure that data of different dimensions and magnitudes participate in subsequent modeling on a uniform scale and to improve model stability and accuracy, the normalization process includes:

[0075]

[0076] Where x is the original data, x min and x max , respectively, are the minimum and maximum values ​​of the feature, and x′ is the normalized data;

[0077] Taking the load rate of a chiller unit as an example: the minimum value is 55% and the maximum value is 85%, then the normalization formula is: The normalized chiller load rate data can then be obtained as: [0.33, 0.5, 0.77, 0.9, 1, 0.83, 0.67, 0.43, 0.23, 0]; similarly, the normalized results of other feature values ​​can be obtained.

[0078] To further implement the above technical solution, the specific details of using the Bayesian algorithm to fill in missing data are as follows:

[0079] The collected central air conditioning system operation data is divided into two parts: complete data and missing data, and the location of the missing values ​​is recorded.

[0080] Identify the variables X1, X2, ..., X in the dataset. n And determine the probabilistic dependencies between variables, and construct a Bayesian network to represent the probabilistic dependencies;

[0081] Using the complete data portion, the conditional probability distribution P(X) of each variable in the Bayesian network is learned through Bayesian estimation. i Parents(X) i Parents(X) i ) is X i The parent node in the Bayesian network is used to construct the Bayesian model;

[0082] Initialize the missing data by filling it with the mean, median, or random values ​​as an initial estimate;

[0083] For each missing value, the posterior probability distribution is calculated using the known data and Bayes' theorem.

[0084]

[0085] Wherein, P(X) miss |E) is the posterior probability, representing the missing value X given the data E. miss The probability distribution of P(E|X); miss Let be the likelihood function, representing the likelihood of missing values ​​X. miss Given the probability of E occurring, P(X) miss ) represents the missing value X miss The prior probability, P(E), is the marginal probability used for normalization;

[0086] According to the posterior probability distribution P(X) miss |E), select the value with the highest probability as the estimate of the missing value.

[0087] To further implement the above technical solution, the specific content of step S2 is as follows:

[0088] Randomly shuffle each original feature data of the operating parameters and environmental operating parameters to generate corresponding shadow features;

[0089] For example, the normalized chiller load rate data [0.33, 0.5, 0.77, 0.9, 1, 0.83, 0.67, 0.43, 0.23, 0], after being randomly shuffled, may be: [0.50, 0.83, 0.33, 0.23, 0.67, 0.9, 0.43, 0.77, 0, 1];

[0090] The original features and shadow features are combined into a new dataset to train the random forest model;

[0091] The importance of each feature is calculated using a random forest model:

[0092] In this embodiment, the feature importance is:

[0093]

[0094] Where, N tree It is the total number of decision trees in the random forest, errOObt j It is variable x j The prediction error of a decision tree, errOObt, is the variable x. j The prediction error of the decision tree corresponding to the features after random perturbation;

[0095] The original feature importance calculated by the random forest model is as follows: chiller load rate: 0.5; chilled water supply temperature: 0.67; cooling water return temperature: 0.67; pump frequency: 0.43; cooling tower fan frequency: 0.4; outdoor temperature: 0.53; humidity: 0.3.

[0096] The importance of each original feature is statistically tested using the modified Brownian bridge statistic to determine whether the original features are significantly more important than the shadow features. Insignificant features are removed, and the subset of features that have a significant impact on the energy efficiency ratio is retained.

[0097]

[0098] Among them, Importance j The importance of the j-th original feature, μ shadow and σ shadowThese are the mean and standard deviation of the importance of shadow features, respectively;

[0099] As a result, the chiller unit load rate, chilled water supply temperature, cooling water return temperature, and outdoor temperature were selected as feature subsets.

[0100] To further implement the above technical solution, the specific content of step S3 is as follows:

[0101] S31. Select a feature subset as the model input. The feature subset includes chiller load rate, chilled water supply temperature, cooling water return temperature and outdoor temperature. Use the performance index energy efficiency ratio as the model output.

[0102] S32. Initialize network parameters, set network structure, and define ReLU activation function; the network structure includes input layer, hidden layer and output layer; the number of nodes in the input layer is equal to the number of feature subsets, and the number of nodes in the output layer is 1, i.e., energy efficiency ratio; randomly initialize the weights and biases of the network, and set the weight range to [-1,1];

[0103] The number of hidden layer nodes is:

[0104]

[0105] Where, n input n is the number of nodes in the input layer. output The number of nodes in the output layer is the number of nodes in the hidden layer, which is rounded down.

[0106] In this implementation, the feature subset includes chiller load rate, chilled water supply temperature, cooling water return temperature, and outdoor temperature; therefore, the number of input layer nodes is n. input =4, number of output layer nodes n output =1, which leads to The number of hidden layer nodes is set to 3;

[0107] The activation function chosen is the ReLU function, and the formula is as follows:

[0108] f(x) = max(0,x);

[0109] S33. Input the central air conditioning system's operating data into the network for forward propagation. The input layer receives the normalized cooling unit load rate, chilled water supply temperature, cooling water return temperature, and outdoor temperature, and passes them to the hidden layer. The hidden layer calculates the weighted sum of the inputs and activates them using the ReLU function. The output of the hidden layer is then passed to the output layer, where the input and output of each neuron in the output layer are calculated. The output of the output layer is the model's prediction result.

[0110] For example, the weight matrix from the input layer to the hidden layer is:

[0111]

[0112] The bias vector is:

[0113] b hidden = [0.1, -0.2, 0.3]

[0114] The weight vector from the hidden layer to the input layer is:

[0115] w hidden-output = [0.5, 0.3, -0.4]

[0116] The bias is:

[0117] b output =0.2

[0118] During forward propagation, taking the i-th neuron in the hidden layer as an example, its input is:

[0119]

[0120] Among them, w ji x represents the weights from the j-th neuron in the input layer to the i-th neuron in the hidden layer. j b represents the output of the j-th neuron in the input layer. i This represents the bias of the i-th neuron in the hidden layer;

[0121] Taking the data at the first time point (9:00) as an example, after normalization, the chiller unit load rate is 0.33, the chilled water supply temperature is 0, the cooling water return temperature is 0.0465, and the outdoor temperature is 0.078.

[0122] z1 = 0.19, z2 = -0.21, z3 = 0.40

[0123] Calculate the output of the i-th neuron in the hidden layer using the activation function:

[0124] f(x) = max(0,x)

[0125] y i =f(net) i )

[0126] This function outputs the input itself when the input is greater than 0, and outputs 0 otherwise, which can effectively introduce nonlinearity and avoid the gradient vanishing problem.

[0127] y1=0.19, y2=0, y3=0.4;

[0128] For the k-th neuron in the output layer, its input is:

[0129]

[0130] Among them, w iky represents the weights from the i-th neuron in the hidden layer to the k-th neuron in the output layer. i b represents the output of the i-th neuron in the hidden layer. k This represents the bias of the k-th neuron in the output layer;

[0131] Then, the output z of the output layer out =0.13 is the model's prediction result, i.e., EERpred = 0.13;

[0132] S34. Using the mean squared error (MSE) as the loss function, calculate the error between the model output and the actual value, propagate the error back from the output layer to the input layer, adjust the network parameters, calculate the gradient of each weight, update the weights using gradient descent to minimize the error, and update the biases.

[0133] The loss function is:

[0134]

[0135] Among them, y i Output values ​​for the model. Here, m represents the actual value, and m represents the sample size.

[0136] Assuming the actual energy efficiency ratio is 1.5, then MSE = (0.13 - 1.5) 2 =1.88;

[0137] For the weights w of the output layer ij The gradient is:

[0138]

[0139] Where, f′(net) i Let be the derivative of the activation function. For the ReLU function, its derivative is:

[0140]

[0141] The weight update formula is:

[0142]

[0143] Where η is the learning step size; This is the partial derivative of the error with respect to the weights;

[0144] Bias update formula:

[0145]

[0146] In this embodiment, the updated weight matrix from the input layer to the hidden layer is:

[0147]

[0148] The updated weight vector from the hidden layer to the output layer is:

[0149] w hidden-output =[0.5,0.3,-0.4]

[0150] S35. Repeat steps S33-S34 for forward propagation, error calculation, back propagation, and parameter update until the convergence condition is met; in this embodiment, the convergence condition is set to stop when the maximum number of iterations reaches 1000 or the mean square error is less than 0.01.

[0151] To further implement the above technical solutions, such as Figure 4 The specific content of introducing the whale algorithm to optimize the BP neural network prediction model is as follows:

[0152] S41. Based on the weights and biases of the BP neural network, randomly initialize the positions of the whale population, with each individual whale representing a set of possible system operating parameters; when initializing the population, the range of each position component is consistent with the BP network initialization.

[0153] For example, for individual whale 1, the weight matrix from the input layer to the hidden layer is: The weight vector from the hidden layer to the output layer is [0.6, 0.4, -0.5], the bias vector of the hidden layer is [0.2, -0.3, 0.4], and the bias of the output layer is 0.3;

[0154] S42. Using the reciprocal of the prediction error as the fitness function, the prediction error of each individual whale is calculated through a BP neural network using the training dataset and used as the fitness value. The position of the whale is iteratively updated according to the fitness value. If the current individual is the optimal solution, the position is adjusted according to the spiral update formula.

[0155] S43. Check if the convergence condition is met. If it is, stop the optimization. If not, return to the fitness calculation step and continue the iteration.

[0156] To further implement the above technical solution, the fitness function is specifically as follows:

[0157]

[0158] Where E is the mean squared error, which is the error between the model output and the actual value;

[0159] The whale's location is updated using the following method:

[0160] D = |C·X * (t)-X(t)|

[0161] X(t+1)=X * (t)-A·D

[0162] Where D is the distance vector between the current whale individual and the optimal solution, representing the relative position of the individual and the optimal solution in the search space; A and C are coefficient vectors used to control the direction and step size of the individual's movement; X * X(t) represents the position of the current optimal solution, i.e., the position of the individual with the highest fitness value in the population; X(t) represents the current position of the whale individual, and X(t+1) represents the new position of the whale individual after the update;

[0163] The coefficient vector is:

[0164] A = 2·a·r1-a

[0165] C = 2·r²

[0166] Where 'a' is a parameter that decreases linearly from 2 to 0, used to control the size of the search range; initially a = 2, and gradually decreases as iterations proceed, causing the search range to gradually shift from global to local, a = 2 - 2·(t / T) max ), T max The maximum number of iterations is given by r1 and r2, where r1 and r2 are random vectors in the interval [0, 1].

[0167] In this embodiment, the position of the current optimal solution is X. * = [0.5, 0.3, -0.4], the current position of the individual whale is X = [0.2, -0.3, 0.5], we can calculate: D = |0.5.[0.5, 0.3, -0.4] - [0.2, -0.3, 0.5]|, X(t+1) = [0.5, 0.3, -0.4] - 0.5·D;

[0168] The control parameters that maximize the energy efficiency ratio were applied to the actual control system to achieve intelligent control of the central air conditioning system. The energy efficiency ratio before optimization was 1.5, and after optimization it was 1.75, an improvement of 16.67%.

[0169] A central air conditioning performance optimization system based on whale algorithm and BP neural network is deployed in the central air conditioning control system. It includes a data acquisition and preprocessing module, and a central air conditioning performance optimization module constructed using a central air conditioning performance optimization method based on whale algorithm and BP neural network.

[0170] The data acquisition and preprocessing module is used to collect the operating data of the central air conditioning system, including equipment operating parameters, environmental parameters and energy consumption data, and to perform data preprocessing to obtain stable operating data of the central air conditioning system.

[0171] The central air conditioning performance optimization module is used to obtain the control parameters for maximizing the energy efficiency ratio based on the collected stable operation data of the central air conditioning system using a trained and optimized BP neural network prediction model, thereby realizing intelligent control of the central air conditioning system.

[0172] The central air conditioning performance optimization module includes a database, a feature selection unit, a model building and training unit, and a whale optimization unit;

[0173] The database is used to store historical stable operation data of the central air conditioning system after the data acquisition and preprocessing module has collected and processed it.

[0174] The feature selection unit is used to perform feature selection using the Boruta feature selection algorithm, selecting energy efficiency ratio as the performance index, equipment operating parameters and environmental operating parameters as feature variables, calculating the feature importance of multiple feature variables, and obtaining a feature subset.

[0175] The model building and training unit is used to build a BP neural network prediction model using stable operation data of the central air conditioning system, select a subset of features as model input, use performance indicators as model output, and train the model.

[0176] The whale optimization unit is used to introduce the whale algorithm to optimize the BP neural network prediction model, resulting in an optimized prediction model.

[0177] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a central air conditioning performance optimization method based on the whale algorithm and a BP neural network.

[0178] A processing terminal includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements a central air conditioning performance optimization method based on the whale algorithm and BP neural network.

[0179] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0180] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for optimizing the performance of a central air conditioning system based on the whale algorithm and a backpropagation neural network, characterized in that, Includes the following steps: S1. Collect the operating data of the central air conditioning system, including equipment operating parameters, environmental parameters and energy consumption data, and perform data preprocessing to obtain stable operating data of the central air conditioning system; S2. The Boruta feature selection algorithm is used to select features, with energy efficiency ratio as the performance index, and equipment operating parameters and environmental operating parameters as feature variables. The feature importance of multiple feature variables is calculated to obtain a feature subset. S3. Construct a BP neural network prediction model using stable operation data of the central air conditioning system, select a subset of features as model input, use performance indicators as model output, and train the model. S4. Introduce the whale algorithm to optimize the BP neural network prediction model; S5. Evaluate the optimized performance indicators to obtain the control parameters that maximize the energy efficiency ratio, and apply them to the actual control system of the central air conditioning system to achieve intelligent control of the central air conditioning system.

2. The central air conditioning performance optimization method based on whale algorithm and BP neural network according to claim 1, characterized in that, Data preprocessing includes cleaning and normalizing the operating data of the central air conditioning system, and using Bayesian algorithms to fill in missing data to obtain stable operating data of the central air conditioning system.

3. The central air conditioning performance optimization method based on whale algorithm and BP neural network according to claim 2, characterized in that, The specific steps for using the Bayesian algorithm to complete missing data are as follows: The collected central air conditioning system operation data is divided into two parts: complete data and missing data, and the location of missing values ​​is recorded. Identify variables in the dataset, determine the probabilistic dependencies between variables, and construct a Bayesian network to represent the probabilistic dependencies; Using the complete data portion, the conditional probability distribution of each variable in the Bayesian network is learned through Bayesian estimation methods to construct a Bayesian model; Initialize the missing data by filling it with the mean, median, or random values ​​as an initial estimate; For each missing value, the posterior probability distribution is calculated using the known data and Bayes' theorem. Based on the posterior probability distribution, the value with the highest probability is selected as the estimate of the missing value.

4. The central air conditioning performance optimization method based on whale algorithm and BP neural network according to claim 1, characterized in that, The specific content of step S2 is as follows: Randomly shuffle each original feature data of the operating parameters and environmental operating parameters to generate corresponding shadow features; The original features and shadow features are combined into a new dataset to train the random forest model; The importance of each feature is calculated using a random forest model. The importance of each original feature is statistically tested using the modified Brownian bridge statistic to determine whether the original features are significantly more important than the shadow features. Insignificant features are removed, and a subset of features that have a significant impact on the energy efficiency ratio is retained.

5. The central air conditioning performance optimization method based on whale algorithm and BP neural network according to claim 1, characterized in that, The specific content of step S3 is as follows: S31. Select a feature subset as the model input. The feature subset includes chiller load rate, chilled water supply temperature, cooling water return temperature and outdoor temperature. Use the performance index energy efficiency ratio as the model output. S32. Initialize network parameters, set network structure, and define the ReLU activation function; the network structure includes an input layer, hidden layers, and an output layer; the number of nodes in the input layer is equal to the number of feature subsets, and the number of nodes in the output layer is 1, i.e., the energy efficiency ratio; randomly initialize the network weights and biases, setting the weight range to [-1, 1]. S33. Input the central air conditioning system's operating data into the network for forward propagation. The input layer receives the normalized cooling unit load rate, chilled water supply temperature, cooling water return temperature, and outdoor temperature, and passes them to the hidden layer. The hidden layer calculates the weighted sum of the inputs and activates them using the ReLU function. The output of the hidden layer is then passed to the output layer, where the input and output of each neuron in the output layer are calculated. The output of the output layer is the model's prediction result. S34. Using the mean squared error (MSE) as the loss function, calculate the error between the model output and the actual value, propagate the error back from the output layer to the input layer, adjust the network parameters, calculate the gradient of each weight, update the weights using gradient descent to minimize the error, and update the biases. S35. Repeat steps S33-S34 for forward propagation, error calculation, back propagation, and parameter update until the convergence condition is met.

6. The central air conditioning performance optimization method based on whale algorithm and BP neural network according to claim 5, characterized in that, The specific details of using the whale algorithm to optimize the BP neural network prediction model are as follows: S41. Randomly initialize the locations of the whale population, with each individual whale representing a set of possible system operating parameters; when initializing the population, the range of each location component is consistent with the initialization of the BP network; S42. Using the reciprocal of the prediction error as the fitness function, the prediction error of each individual whale is calculated through a BP neural network using the training dataset and used as the fitness value. The position of the whale is iteratively updated according to the fitness value. If the current individual is the optimal solution, the position is adjusted according to the spiral update formula. S43. Check if the convergence condition is met. If it is met, stop the optimization. If it is not met, return to the fitness calculation step and continue the iteration.

7. The method for optimizing the performance of a central air conditioning system based on the whale algorithm and BP neural network according to claim 6, characterized in that, The fitness function is as follows: Where E is the mean squared error, which is the error between the model output and the actual value; The whale's location is updated using the following method: D=|C·X * (t)-X(t)| X(t+1)=X * (t)-A.D Where D is the distance vector between the current whale individual and the optimal solution, representing the relative position of the individual and the optimal solution in the search space; A and C are coefficient vectors used to control the direction and step size of the individual's movement; X * X(t) represents the position of the current optimal solution, i.e., the position of the individual with the highest fitness value in the population; X(t) represents the current position of the whale individual, and X(t+1) represents the new position of the whale individual after the update; The coefficient vector is: A = 2·a·r1-a C＝2·r2 Where 'a' is a parameter that decreases linearly from 2 to 0, used to control the size of the search range; initially a = 2, and gradually decreases as iterations proceed, causing the search range to gradually shift from global to local, a = 2 - 2·(t / T) max ), T max Let r1 and r2 be the maximum number of iterations, and r1 and r2 be random vectors in the interval [0, 1].

8. A central air conditioning performance optimization system based on whale algorithm and BP neural network, characterized in that, Deployed in a central air conditioning control system, it includes a data acquisition and preprocessing module, and a central air conditioning performance optimization module constructed using a central air conditioning performance optimization method based on whale algorithm and BP neural network as described in any one of claims 1-7. The data acquisition and preprocessing module is used to collect the operating data of the central air conditioning system, including equipment operating parameters, environmental parameters and energy consumption data, and to perform data preprocessing to obtain stable operating data of the central air conditioning system. The central air conditioning performance optimization module is used to obtain the control parameters for maximizing the energy efficiency ratio based on the collected stable operation data of the central air conditioning system using a trained and optimized BP neural network prediction model, thereby realizing intelligent control of the central air conditioning system. The central air conditioning performance optimization module includes a database, a feature selection unit, a model building and training unit, and a whale optimization unit; The database is used to store historical stable operation data of the central air conditioning system after the data acquisition and preprocessing module has collected and processed it. The feature selection unit is used to perform feature selection using the Boruta feature selection algorithm, selecting energy efficiency ratio as the performance index, equipment operating parameters and environmental operating parameters as feature variables, calculating the feature importance of multiple feature variables, and obtaining a feature subset. The model building and training unit is used to build a BP neural network prediction model using stable operation data of the central air conditioning system, select a subset of features as model input, use performance indicators as model output, and train the model. The whale optimization unit is used to introduce the whale algorithm to optimize the BP neural network prediction model, resulting in an optimized prediction model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the central air conditioning performance optimization method based on the whale algorithm and BP neural network as described in any one of claims 1-7.

10. A processing terminal, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the central air conditioning performance optimization method based on the whale algorithm and BP neural network as described in any one of claims 1-7.

Images