Air conditioner energy consumption prediction method, system and equipment and storage medium

Abstract

The invention belongs to the field of air conditioner energy consumption prediction, and particularly relates to an air conditioner energy consumption prediction method, system and device and a storage medium. After training air conditioner energy consumption data is processed, a covariance matrix is obtained, characteristic decomposition is carried out, characteristic values and characteristic vectors are obtained, the characteristic vectors are sorted according to the characteristic values, a characteristic vector matrix is obtained, screening is carried out according to the contribution rate, a screening matrix is obtained, and according to the screening matrix, the initial training data weight and the number of predictors, a prediction result is obtained; and training each sub-predictor to obtain a training weight of each sub-predictor, updating the training weight by combining a training error to obtain an updated weight, performing weighted summation on the updated weight of each sub-predictor to obtain a strong predictor, and performing prediction through the strong predictor to obtain predicted air conditioner energy consumption. Compared with a traditional neural network weight training mode, whether the training data is correctly classified or not is considered, the misclassification degree is also considered, and the accuracy of air conditioner energy consumption prediction is improved.

Description

Technical Field

[0001] The present invention belongs to the field of air conditioning energy consumption prediction, and in particular relates to an air conditioning energy consumption prediction method, system, device and storage medium. Background Art

[0002] Factory air conditioning energy consumption forecasting involves analyzing historical data, environmental parameters, equipment status, and other information to develop mathematical models to estimate future energy consumption for a factory's air conditioning system. The core goal is to optimize air conditioning operation strategies and reduce energy waste through scientific forecasting.

[0003] In related technologies, the energy consumption prediction of factory air conditioners is mainly carried out through data processing and analysis methods to explore its energy consumption patterns. The prediction results can also be obtained through neural network model training. When training the neural network model, the air conditioner energy consumption data is mainly input into the neural network model for training and the model weights are adjusted.

[0004] Regarding the above-mentioned related technologies, when updating weights in traditional neural network models, it only determines whether the current training data is correctly classified, without defining the degree of misclassification. This may lead to serious misclassification that is different from the actual situation during training, resulting in inaccurate prediction results. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an air conditioning energy consumption prediction method, system, equipment and storage medium, which can improve the accuracy of air conditioning energy consumption prediction, reasonably arrange production plans based on the predicted air conditioning energy consumption, and thus reduce the overall energy consumption of the factory.

[0006] A method for predicting air conditioning energy consumption, comprising:

[0007] Acquire training air conditioning energy consumption data, wherein the energy consumption data includes date, time, meteorological parameters, control information, and energy consumption value;

[0008] Normalizing the training air conditioning energy consumption data to obtain a standard matrix;

[0009] Converting the standard matrix into a covariance matrix;

[0010] Performing eigendecomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0011] Sort the corresponding eigenvectors according to the eigenvalues ​​to obtain an eigenvector matrix;

[0012] Calculating the contribution rate of each eigenvector according to the eigenvalue, and screening the eigenvector matrix according to the contribution rate to obtain a screening matrix;

[0013] Set the initial training data weight and the number of sub-predictors;

[0014] Training each sub-predictor according to the screening matrix, the initial training data weight, and the number of predictors to obtain a training weight for each sub-predictor;

[0015] Get the training error of each sub-predictor;

[0016] updating the training weight of each sub-predictor according to the training error to obtain an updated weight of each sub-predictor;

[0017] Perform weighted summation on the updated weights of each sub-predictor to obtain a strong predictor;

[0018] The actual air-conditioning energy consumption data is input into the strong predictor to obtain the predicted air-conditioning energy consumption.

[0019] Optionally, converting the standard matrix into a covariance matrix includes:

[0020] Transposing the standard matrix to obtain a transposed matrix;

[0021] Get the conversion formula;

[0022] Input the transposed matrix and the standard matrix into the conversion formula to obtain a covariance matrix;

[0023] The conversion formula is:

[0024]

[0025] Among them, X is the standard matrix, X T is the transposed matrix, and n is the number of training data.

[0026] Optionally, calculating the contribution rate of each eigenvector according to the eigenvalue, and screening the eigenvector matrix according to the contribution rate to obtain the screening matrix includes:

[0027] According to the eigenvalue corresponding to each eigenvector, the total eigenvalue is calculated;

[0028] According to the eigenvalue corresponding to each eigenvector and the total eigenvalue, the contribution rate corresponding to each eigenvector is obtained;

[0029] The contribution rates are summed from large to small until the sum of the contribution rates is greater than a preset value, and the eigenvectors whose sum of contribution rates is greater than the preset value are screened from the eigenvector matrix to form a screening matrix.

[0030] Optionally, each sub-predictor is trained according to the screening matrix, the initial training data weight, and the number of predictors, and the training weight of each sub-predictor is obtained, comprising:

[0031] Randomly extracting m groups of data from the screening matrix as training data;

[0032] Inputting the training data into each of the predictors to obtain a predicted value, and obtaining a prediction error based on the predicted value and the actual value in the training data;

[0033] Adjusting the initial training data weights according to the prediction error and the adjustment formula to obtain adjusted weights;

[0034] The adjustment formula is:

[0035]

[0036] Where i = 1, 2…, m, e c is the absolute value of the prediction error, e e is the predetermined error limit;

[0037] Calculating a prediction error rate based on the adjustment weights;

[0038] According to the prediction error rate and the weight formula, the training weight of each sub-predictor is obtained.

[0039] Optionally, the weight formula is:

[0040]

[0041] Among them, α t is the training weight, e t is the prediction error rate, and N is the number of sub-predictors.

[0042] Optionally, obtaining the training error of each sub-predictor includes:

[0043] Obtaining a training error based on the training weights, the predicted values ​​of the training data, the true values ​​of the training data, and an error formula;

[0044] The error formula is:

[0045]

[0046] Where N is the number of training data, is the predicted value of the i-th training data, y i is the true value of the i-th training data, Generate True or False according to the judgment result, that is, the value is 0 or 1, ω miis the weight corresponding to the i-th training data on the m-th classifier.

[0047] Optionally, updating the training weight of each sub-predictor according to the training error to obtain the updated weight of each sub-predictor includes:

[0048] Get the weight update formula;

[0049] Inputting the training error into the weight update formula to obtain the updated weight of each sub-predictor;

[0050] The weight update formula is:

[0051]

[0052] Among them, ω mi is the weight corresponding to the i-th training data on the m-th classifier, Z m is the weight sum, is the predicted value of the i-th training data, y i The true value of the i-th training data, α m is the weight of each sub-predictor.

[0053]

[0054] Among them, e m is the training error, and L is the number of categories.

[0055] An air conditioning energy consumption prediction system, comprising:

[0056] A first acquisition module is used to acquire training air-conditioning energy consumption data, wherein the training air-conditioning energy consumption data includes date, time, meteorological parameters, control information and energy consumption value;

[0057] A processing module, configured to normalize the training air-conditioning energy consumption data to obtain a standard matrix;

[0058] A conversion module, used for converting the standard matrix into a covariance matrix;

[0059] A decomposition module, configured to perform eigendecomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0060] A sorting module, configured to sort the corresponding eigenvectors according to the eigenvalues ​​to obtain an eigenvector matrix;

[0061] a calculation module, configured to calculate a contribution rate of each eigenvector according to the eigenvalue, and screen the eigenvector matrix according to the contribution rate to obtain a screening matrix;

[0062] The setting module is used to set the initial training data weight and the number of sub-predictors;

[0063] a training module, configured to train each sub-predictor according to the screening matrix, the initial training data weight, and the number of predictors to obtain a training weight for each sub-predictor;

[0064] The second acquisition module is used to obtain the training error of each sub-predictor;

[0065] an updating module, configured to update the training weight of each sub-predictor according to the training error to obtain an updated weight of each sub-predictor;

[0066] A summation module is used to perform weighted summation on the updated weights of each sub-predictor to obtain a strong predictor;

[0067] The prediction module is used to input the actual air-conditioning energy consumption data into the strong predictor to obtain the predicted air-conditioning energy consumption.

[0068] A terminal device includes a memory and a processor. The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, an air conditioning energy consumption prediction method is adopted.

[0069] A computer-readable storage medium stores a computer program. When the computer program is loaded and executed by a processor, an air conditioning energy consumption prediction method is adopted.

[0070] The beneficial effects of the present invention are:

[0071] After processing the training air-conditioning energy consumption data, a covariance matrix is ​​obtained, and eigendecomposition is performed to obtain eigenvalues ​​and eigenvectors. The eigenvectors are sorted according to the eigenvalues ​​to obtain an eigenvector matrix, and the eigenvector matrix is ​​screened according to the contribution rate to obtain a screening matrix. According to the screening matrix, the initial training data weights and the number of predictors, each sub-predictor is trained to obtain the training weight of each sub-predictor. The training weights are updated in combination with the training error to obtain the updated weights. The updated weights of each sub-predictor are weighted summed to obtain a strong predictor, and the predicted air-conditioning energy consumption is predicted by the strong predictor. Compared with the traditional neural network weight training method, the present application not only takes into account whether the training data is correctly classified, but also takes into account the degree of misclassification, thereby improving the accuracy of air-conditioning energy consumption prediction. According to the predicted air-conditioning energy consumption, the production plan is reasonably arranged, thereby reducing the overall energy consumption of the factory. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 The figure is a flow chart of an air conditioning energy consumption prediction method of the present invention. DETAILED DESCRIPTION

[0073] A method for predicting air conditioning energy consumption, such as Figure 1 As shown, including:

[0074] S1. Acquire training air conditioning energy consumption data, where the training air conditioning energy consumption data includes date, time, meteorological parameters, control information, and energy consumption value.

[0075] Specifically, a set of air conditioning energy consumption data includes date, time, meteorological parameters, control information, and energy consumption values. Each date, time, meteorological parameter, and control information corresponds to an energy consumption value. Multiple sets of date, time, meteorological parameters, control information, and energy consumption values ​​form a standard matrix. Control information can be set with different parameters depending on the plant. For example, for hydropower generation, control information could include high-zone secondary supply water temperature, high-zone secondary return water temperature, low-zone primary return water temperature, and low-zone secondary supply water temperature.

[0076] S2. Normalize the training air-conditioning energy consumption data to obtain a standard matrix.

[0077] Specifically, in the standard matrix, each column represents a feature variable, each row is an observation training data, and the columns are date, time, and weather, etc.

[0078] S3. Convert the standard matrix into a covariance matrix.

[0079] Converting the standard matrix into a covariance matrix includes:

[0080] The standard matrix is ​​transposed to obtain a transposed matrix.

[0081] Get the conversion formula.

[0082] The transposed matrix and the standard matrix are input into the conversion formula to obtain a covariance matrix.

[0083] The conversion formula is:

[0084]

[0085] Among them, X is the standard matrix, X T is the transposed matrix, n is the number of training data, X∈R m×n , m and n are the number of rows and columns of the matrix.

[0086] S4. Perform eigendecomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors.

[0087] Specifically, the covariance matrix R is subjected to eigenvalue decomposition, with the goal of finding a set of eigenvalues ​​(eigenvalues) and corresponding eigenvectors (eigenvectors). Eigenvalues ​​and eigenvectors are usually obtained by solving the characteristic equation.

[0088] S5. Sort the corresponding eigenvectors according to the eigenvalues ​​to obtain an eigenvector matrix.

[0089] Specifically, each eigenvector is called a pivot. The eigenvalues ​​are arranged in descending order, and the corresponding eigenvectors are also arranged in order to form an eigenvector matrix.

[0090] S6. Calculate the contribution rate of each eigenvector according to the eigenvalue, and screen the eigenvector matrix according to the contribution rate to obtain a screening matrix.

[0091] The step of calculating the contribution rate of each eigenvector according to the eigenvalue, and screening the eigenvector matrix according to the contribution rate to obtain a screening matrix includes:

[0092] According to the eigenvalue corresponding to each eigenvector, the total eigenvalue is calculated.

[0093] According to the eigenvalue corresponding to each eigenvector and the total eigenvalue, the contribution rate corresponding to each eigenvector is obtained.

[0094] The contribution rates are summed from large to small until the sum of the contribution rates is greater than a preset value, and the eigenvectors whose sum of contribution rates is greater than the preset value are screened from the eigenvector matrix to form a screening matrix.

[0095] Specifically, the contribution rate of the kth principal component is defined as the eigenvalue corresponding to the principal component divided by the sum of the total eigenvalues, expressed as:

[0096]

[0097] Among them, λ i is the eigenvalue, n is the number of features, V k is the contribution rate.

[0098] The cumulative contribution rate of the kth principal component is defined as the sum of the eigenvalues ​​of the first k principal components divided by the sum of the total eigenvalues;

[0099]

[0100] Where C is the eigenvector matrix.

[0101] When the cumulative contribution rate is greater than or equal to 95% (preset value), the first k principal components can form a new matrix as the criterion for statistical analysis. The matrix composed of the principal components arranged in order according to the contribution rate is called the screening matrix Score. The final selected matrix is ​​the part of the screening matrix Score with a cumulative contribution rate higher than 95%.

[0102] S7. Set the initial training data weight and the number of sub-predictors.

[0103] S8. Train each sub-predictor according to the screening matrix, the initial training data weight, and the number of predictors to obtain a training weight for each sub-predictor.

[0104] Each sub-predictor is trained according to the screening matrix, the initial training data weight, and the number of predictors, and the training weight of each sub-predictor is obtained, including:

[0105] Specifically, a BP neural network is established based on the dimensionality of the training data, and the weights and thresholds of each node in the network are initialized. PSO, as a swarm intelligence search algorithm, has the advantages of simple implementation principles and strong global search capabilities. Using the PSO algorithm to optimize the network's initial weights and thresholds can improve the performance of the BP network.

[0106] M groups of data are randomly selected from the screening matrix as training data.

[0107] Randomly extract m groups of training data [x1, x2, ..., x m ] From the processed training data set, m groups of training data weights w i :

[0108]

[0109] Where w i is the initial weight; m is the number of training data.

[0110] Specifically, weights are assigned to the training data, and PSO is used to optimize the initial weights and thresholds of the BP neural network.

[0111] The training data is input into each of the predictors to obtain a predicted value, and a prediction error is obtained based on the predicted value and the actual value in the training data.

[0112] According to the prediction error and the adjustment formula, the initial training data weight is adjusted to obtain the adjusted weight.

[0113] The adjustment formula is:

[0114]

[0115] Where i = 1, 2…, m, e c is the absolute value of the prediction error, e e is the predetermined error limit.

[0116] Specifically, adjust the sub-predictor weight w i , and normalized.

[0117] The prediction error rate is calculated based on the adjustment weight.

[0118] Specifically, calculate the prediction error rate e of the t-th sub-predictor t .

[0119]

[0120] Where N is the number of weak predictors.

[0121] According to the prediction error rate and the weight formula, the training weight of each sub-predictor is obtained.

[0122] The weight formula is:

[0123]

[0124] Among them, α t is the training weight, e t is the prediction error rate, and N is the number of sub-predictors.

[0125] S9. Obtain the training error of each sub-predictor.

[0126] Obtaining the training error of each sub-predictor includes:

[0127] A training error is obtained according to the training weights, the predicted values ​​of the training data, the true values ​​of the training data, and an error formula.

[0128] The error formula is:

[0129]

[0130] Where N is the number of training data, is the predicted value of the i-th training data, y i is the true value of the i-th training data, Generate True or False according to the judgment result, that is, the value is 0 or 1, ω mi is the weight corresponding to the i-th training data on the m-th classifier.

[0131] Specifically, traditional AdaBoost models treat all misclassifications equally. When updating weights, they only determine whether the current training data is correctly classified, without defining the degree of misclassification. A mild misclassification can be misclassified to a different degree of absurdity than a severe misclassification, leading to significant misclassification of the true case. Therefore, significant misclassifications should be considered more carefully when updating sample weights to ensure they receive the necessary attention.

[0132] S10. Update the training weight of each sub-predictor according to the training error to obtain an updated weight of each sub-predictor.

[0133] The updating of the training weight of each sub-predictor according to the training error to obtain the updated weight of each sub-predictor includes:

[0134] Get the weight update formula.

[0135] The training error is input into the weight update formula to obtain the updated weight of each sub-predictor.

[0136] The weight update formula is:

[0137]

[0138] Among them, ω mi is the weight corresponding to the i-th training data on the m-th classifier, Z m is the weight sum, is the predicted value of the i-th training data, y i The true value of the i-th training data, α m is the weight of each sub-predictor. Formula Used to keep the sum of weights equal to 1. Cosine similarity is used to define the gap between two categories. To ensure the similarity is in (0, 1), an absolute value is added and the result is positive. The difference between categories is fully distinguished by the equation.

[0139]

[0140] Among them, e m is the training error, and L is the number of categories.

[0141] S11. Perform weighted summation on the updated weights of each sub-predictor to obtain a strong predictor.

[0142] Specifically, after N trainings, N groups of sub-predictors f are combined according to their weights. t (x) are weighted and combined into a strong prediction function F(x).

[0143]

[0144] S12: Input the actual air-conditioning energy consumption data into the strong predictor to obtain the predicted air-conditioning energy consumption.

[0145] Specifically, by accurately predicting energy consumption, factories can plan energy use in advance and avoid energy waste. For example, if it is predicted that air conditioning energy consumption will be high during a certain period, but production tasks will be light, the production plan can be adjusted in advance to schedule some production tasks during the period when air conditioning energy consumption is lower, thereby reducing the factory's overall energy consumption.

[0146] Based on energy consumption forecasts, energy supply is rationally allocated, such as optimizing the energy supply to different workshops and different equipment to ensure that while meeting production needs, redundant energy supply is minimized and energy waste is avoided.

[0147] An air conditioning energy consumption prediction system, comprising:

[0148] A first acquisition module is used to acquire training air-conditioning energy consumption data, wherein the training air-conditioning energy consumption data includes date, time, meteorological parameters, control information and energy consumption value;

[0149] A processing module, configured to normalize the training air-conditioning energy consumption data to obtain a standard matrix;

[0150] A conversion module, used for converting the standard matrix into a covariance matrix;

[0151] A decomposition module, configured to perform eigendecomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0152] A sorting module, configured to sort the corresponding eigenvectors according to the eigenvalues ​​to obtain an eigenvector matrix;

[0153] a calculation module, configured to calculate a contribution rate of each eigenvector according to the eigenvalue, and screen the eigenvector matrix according to the contribution rate to obtain a screening matrix;

[0154] The setting module is used to set the initial training data weight and the number of sub-predictors;

[0155] a training module, configured to train each sub-predictor according to the screening matrix, the initial training data weight, and the number of predictors to obtain a training weight for each sub-predictor;

[0156] The second acquisition module is used to obtain the training error of each sub-predictor;

[0157] an updating module, configured to update the training weight of each sub-predictor according to the training error to obtain an updated weight of each sub-predictor;

[0158] A summation module is used to perform weighted summation on the updated weights of each sub-predictor to obtain a strong predictor;

[0159] The prediction module is used to input the actual air-conditioning energy consumption data into the strong predictor to obtain the predicted air-conditioning energy consumption.

[0160] An embodiment of the present application also discloses a terminal device, including a memory and a processor. The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, an air conditioning energy consumption prediction method is adopted.

[0161] Among them, the terminal device can be a computer device such as a desktop computer, a laptop computer or a cloud server, and the terminal device includes but is not limited to a processor and a memory. For example, the terminal device can also include input and output devices, network access devices and buses, etc.

[0162] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0163] Among them, the memory can be an internal storage unit of the terminal device, such as the hard disk or memory of the terminal device, or it can be an external storage device of the terminal device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD) or flash memory card (FC) equipped on the terminal device, etc., and the memory can also be a combination of the internal storage unit and the external storage device of the terminal device. The memory is used to store computer programs and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or is to be output. This application does not impose any restrictions on this.

[0164] Among them, through this terminal device, an air conditioning energy consumption prediction method in the above embodiment is stored in the memory of the terminal device, and is loaded and executed on the processor of the terminal device for easy use.

[0165] An embodiment of the present application further discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, an air conditioning energy consumption prediction method in the above embodiment is adopted.

[0166] Among them, the computer program can be stored in a computer-readable medium, the computer program includes computer program code, the computer program code can be in the form of source code, object code, executable file or certain middleware, etc. The computer-readable medium includes any entity or device that can carry computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that computer-readable medium includes but is not limited to the above-mentioned components.

[0167] Among them, through this computer-readable storage medium, an air conditioning energy consumption prediction method in the above embodiment is stored in a computer-readable storage medium, and is loaded and executed on a processor to facilitate the storage and application of the above method.

[0168] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of protection of the present application is limited to these examples. In line with the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0169] The one or more embodiments of this application are intended to encompass all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of this application should be included in the scope of protection of this application.

Claims

1. A method for predicting air conditioning energy consumption, characterized in that: include: Acquire training air conditioning energy consumption data, wherein the training air conditioning energy consumption data includes date, time, meteorological parameters, control information, and energy consumption value; Normalizing the training air conditioning energy consumption data to obtain a standard matrix; Converting the standard matrix into a covariance matrix; Performing eigendecomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors; Sort the corresponding eigenvectors according to the eigenvalues ​​to obtain an eigenvector matrix; Calculating the contribution rate of each eigenvector according to the eigenvalue, and screening the eigenvector matrix according to the contribution rate to obtain a screening matrix; Set the initial training data weight and the number of sub-predictors; Training each sub-predictor according to the screening matrix, the initial training data weight, and the number of sub-predictors to obtain a training weight for each sub-predictor; Get the training error of each sub-predictor; updating the training weight of each sub-predictor according to the training error to obtain an updated weight of each sub-predictor; Perform weighted summation on the updated weights of each sub-predictor to obtain a strong predictor; The actual air-conditioning energy consumption data is input into the strong predictor to obtain the predicted air-conditioning energy consumption.

2. The air conditioning energy consumption prediction method according to claim 1, wherein: Converting the standard matrix into a covariance matrix includes: Transposing the standard matrix to obtain a transposed matrix; Get the conversion formula; Input the transposed matrix and the standard matrix into the conversion formula to obtain a covariance matrix; The conversion formula is: Among them, X is the standard matrix, X T is the transposed matrix, and n is the number of training data.

3. The air conditioning energy consumption prediction method according to claim 1, wherein: The step of calculating the contribution rate of each eigenvector according to the eigenvalue, and screening the eigenvector matrix according to the contribution rate to obtain a screening matrix includes: According to the eigenvalue corresponding to each eigenvector, the total eigenvalue is calculated; According to the eigenvalue corresponding to each eigenvector and the total eigenvalue, the contribution rate corresponding to each eigenvector is obtained; The contribution rates are summed from large to small until the sum of the contribution rates is greater than a preset value, and the eigenvectors whose sum of contribution rates is greater than the preset value are screened from the eigenvector matrix to form a screening matrix.

4. The air conditioning energy consumption prediction method according to claim 1, wherein: Each sub-predictor is trained according to the screening matrix, the initial training data weight, and the number of sub-predictors, and the training weight of each sub-predictor is obtained, including: Randomly extracting m groups of data from the screening matrix as training data; Inputting the training data into each of the predictors to obtain a predicted value, and obtaining a prediction error based on the predicted value and the actual value in the training data; Adjusting the initial training data weights according to the prediction error and the adjustment formula to obtain adjusted weights; The adjustment formula is: Where i = 1, 2…, m, e c is the absolute value of the prediction error, e e is the predetermined error limit; Calculating a prediction error rate based on the adjustment weights; According to the prediction error rate and the weight formula, the training weight of each sub-predictor is obtained.

5. The air conditioning energy consumption prediction method according to claim 4, wherein: The weight formula is: Among them, α t is the training weight, e t is the prediction error rate, and N is the number of sub-predictors.

6. The air conditioning energy consumption prediction method according to claim 4, wherein: Obtaining the training error of each sub-predictor includes: Obtaining a training error based on the training weights, the predicted values ​​of the training data, the true values ​​of the training data, and an error formula; The error formula is: Where N is the number of training data, is the predicted value of the i-th training data, y i is the true value of the i-th training data, Generate True or False according to the judgment result, that is, the value is 0 or 1, ω mi is the weight corresponding to the i-th training data on the m-th classifier.

7. The air conditioning energy consumption prediction method according to claim 6, wherein: The updating of the training weight of each sub-predictor according to the training error to obtain the updated weight of each sub-predictor includes: Get the weight update formula; Inputting the training error into the weight update formula to obtain the updated weight of each sub-predictor; The weight update formula is: Among them, ω mi is the weight corresponding to the i-th training data on the m-th classifier, Z m is the weight sum, is the predicted value of the i-th training data, y i The true value of the i-th training data, α m is the weight of each sub-predictor; Among them, e m is the training error, and L is the number of categories.

8. An air conditioning energy consumption prediction system, characterized in that: include: A first acquisition module is used to acquire training air-conditioning energy consumption data, wherein the training air-conditioning energy consumption data includes date, time, meteorological parameters, control information and energy consumption value; A processing module, configured to normalize the training air-conditioning energy consumption data to obtain a standard matrix; A conversion module, used for converting the standard matrix into a covariance matrix; A decomposition module, configured to perform eigendecomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors; A sorting module, configured to sort the corresponding eigenvectors according to the eigenvalues ​​to obtain an eigenvector matrix; a calculation module, configured to calculate a contribution rate of each eigenvector according to the eigenvalue, and screen the eigenvector matrix according to the contribution rate to obtain a screening matrix; The setting module is used to set the initial training data weight and the number of sub-predictors; a training module, configured to train each sub-predictor according to the screening matrix, the initial training data weight, and the number of predictors to obtain a training weight for each sub-predictor; The second acquisition module is used to obtain the training error of each sub-predictor; an updating module, configured to update the training weight of each sub-predictor according to the training error to obtain an updated weight of each sub-predictor; A summation module is used to perform weighted summation on the updated weights of each sub-predictor to obtain a strong predictor; The prediction module is used to input the actual air-conditioning energy consumption data into the strong predictor to obtain the predicted air-conditioning energy consumption.

9. A terminal device comprising a memory and a processor, characterized in that: The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, the method according to any one of claims 1 to 7 is adopted.

10. A computer-readable storage medium storing a computer program, wherein: When the computer program is loaded and executed by a processor, the method according to any one of claims 1 to 7 is adopted.

Images