Air conditioning system operation strategy modeling method based on mechanism framework and AI

Abstract

An air conditioning system operation strategy modeling method based on a mechanism framework + AI comprises the steps that climate data are obtained and preprocessed, standardized climate data are obtained, and preprocessing comprises missing value filling processing and normalization processing; and iteratively clustering the standardized climate data through a quantum harmonic oscillator model to generate similar climate characteristic region groups. And a convolutional neural network is called to be combined with a YOLO algorithm to perform training learning according to the climate data samples in the similar climate characteristic region group and the control strategy corresponding to each climate data sample, and an operation strategy model of the air conditioning system is established. Wherein the climate data at least comprises temperature data, humidity data and wind speed data, the convolutional neural network is divided into a teacher model and a student model, the teacher model is a pre-trained model obtained by pre-training based on the climate data, and the student model is a new task model for interactive learning with the teacher model.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control of air-conditioning systems, and more specifically, relates to a method for modeling an air-conditioning system operation strategy based on a mechanism framework + AI. Background Art

[0002] With the development of air-conditioning system control technology, different climatic conditions have a greater impact on the equipment life of the air-conditioning system. Especially in extreme weather conditions, the air-conditioning system often needs to operate at high load for a long time, which to a certain extent accelerates the aging and wear of the equipment.

[0003] At present, it is difficult for the existing air-conditioning system control methods to dynamically adjust the air-conditioning load according to real-time climate data, so it is difficult to effectively manage the service life of air-conditioning system equipment. Excessive operation and unreasonable working mode of air-conditioning system equipment have greatly shortened the service life of air-conditioning equipment, resulting in increased maintenance and replacement costs of air-conditioning equipment. Therefore, if the system load can be reasonably regulated according to climate data to avoid overload operation of equipment, the life of the air-conditioning system will be significantly extended. However, if this problem is not solved, the aging of air-conditioning equipment will be aggravated, that is, the high-intensity working mode under extreme weather will accelerate the aging of air-conditioning equipment and shorten its service life. At the same time, it will also increase maintenance costs. This is because the maintenance and replacement costs of the air-conditioning system will increase significantly due to frequent equipment failures, which will affect the overall economic efficiency of operation. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the purpose of the present invention is to solve the above-mentioned defects and further propose an air-conditioning system operation strategy modeling method based on mechanism framework + AI.

[0005] The present invention adopts the following technical solution.

[0006] The first aspect of the present invention discloses a method for modeling an air-conditioning system operation strategy based on a mechanism framework + AI, the method comprising:

[0007] Acquire climate data, and preprocess the climate data to obtain standardized climate data, wherein the preprocessing includes missing value filling processing and normalization processing;

[0008] Iteratively clustering the standardized climate data through a quantum oscillator model to generate regional groups with similar climate characteristics;

[0009] Calling a convolutional neural network in combination with a YOLO algorithm to perform training and learning based on the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample, and establishing an operation strategy model for the air-conditioning system;

[0010] Among them, the climate data includes at least temperature data, humidity data and wind speed data, and the convolutional neural network is divided into a teacher model and a student model. The teacher model is a pre-trained model obtained by pre-training based on the climate data, and the student model is a new task model that interacts and learns with the teacher model.

[0011] Furthermore, the step of acquiring climate data and preprocessing the climate data to obtain standardized climate data includes:

[0012] Arranging the climate data in chronological order to obtain an initial climate data set consisting of time series data, wherein each climate data sample in the initial climate data set is represented in the form of a data sequence, including a temperature data time series, a humidity data time series, and a wind speed data time series;

[0013] Determine the time points where the data missing values ​​in the initial climate data set are located to form an index set, and fill the data missing values ​​by using a K-nearest neighbor filling method; and

[0014] A time series prediction model is constructed based on each climate data time series through an autoregressive moving average model, and all climate data samples are normalized to obtain a preprocessed standardized climate data set;

[0015] Among them, the modeling expression of the autoregressive moving average model is:

[0016]

[0017] In the formula, The index number is The data is missing values, is the autoregression coefficient, indicating the The weight of the lag term, is the moving average coefficient, indicating the The weight of the error lag term, and Respectively indicate time With time The random error between and are the orders of the autoregressive term and the moving average term, respectively.

[0018] Furthermore, the standardized climate data is iteratively clustered by the quantum oscillator model to generate regional groups with similar climate characteristics, including:

[0019] The L-ISODATA algorithm is called to randomly select multiple standardized climate data samples from the standardized climate data set as initial cluster centers. The expression of the initial cluster center is:

[0020] ;

[0021] In the formula, is the number of standardized climate data samples, For the Cluster centers, , The value of is a data sequence of standardized climate data samples;

[0022] Performing quantum energy assignment based on the initial cluster center, and calling a Gaussian sampling function to generate energy disturbance, so as to update the position of the cluster center according to the energy disturbance;

[0023] Among them, the expression of the quantum energy assignment is:

[0024] ;

[0025] In the formula, For the Cluster Centers The quantum energy of is the proportionality factor, used to control the scale of the energy value, Represents the cluster center data density.

[0026] Furthermore, the expression of the data density is:

[0027]

[0028] In the formula, is the cluster center The number of climate data samples is a smoothing parameter used to control the decay rate of data density. Respectively represent The temperature data time series, humidity data time series and wind speed data time series of the climate data samples are Represents the cluster center No. row vector, ;

[0029] The expression for updating the position of the cluster center is:

[0030] ,

[0031] ;

[0032] In the formula, is the updated cluster center position, is the energy disturbance and conforms to the normal distribution. is the learning rate constant and is a unit random direction vector satisfying .

[0033] Furthermore, the iterative clustering of the standardized climate data by the quantum oscillator model to generate regional groups with similar climate characteristics also includes:

[0034] Calculating the data density and quantum energy assignment of the candidate location points of the updated cluster center, and comparing the data density and quantum energy assignment of the candidate location points with the data density and quantum energy assignment of the cluster center before the update, to determine whether to update the location and quantum energy assignment of the cluster center again;

[0035] When the number of iterative updates of the cluster center reaches a set maximum number of iterations or the absolute value of the difference between the quantum energy assignment of the cluster center after the update and the quantum energy assignment of the cluster center before the update is less than a first threshold, the optimized updated cluster center set is output to obtain the group of regions with similar climate characteristics;

[0036] Among them, the quantum energy assignment of the cluster center after update is , the quantum energy of the cluster center before updating is assigned to ,when season , update the location of the cluster center and the quantum energy assignment.

[0037] Furthermore, the calling of the convolutional neural network in combination with the YOLO algorithm is used to train and learn the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample to establish an operation strategy model of the air-conditioning system, including:

[0038] Inputting each standardized climate data sample in the group of regions with similar climate characteristics into the teacher model to extract a first climate feature vector corresponding to each standardized climate data sample;

[0039] Taking the first climate feature vector and the control strategy corresponding to each standardized climate data sample as input of the YOLO algorithm, calling the YOLO algorithm to predict the control weight vector corresponding to each standardized climate data sample, and taking the first climate feature vector and the control strategy as input and the control weight vector as output, training the YOLO algorithm;

[0040] Among them, the length of the control weight vector is the number of the control strategies, and the loss function expression of the YOLO algorithm is:

[0041]

[0042] In the formula, is the first elements, is the control weight vector of the cluster center corresponding to the climate data sample. elements, Indicates the number of climate data samples corresponding to the cluster center.

[0043] Furthermore, the calling of the convolutional neural network in combination with the YOLO algorithm is used to train and learn the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample to establish an operation strategy model of the air-conditioning system, and further includes:

[0044] The second climate feature vector of each standardized climate data sample in the group of regions with similar climate features is extracted by the teacher model, and a mapping function is called to map each element in the second climate feature vector. The expression of the mapping function is:

[0045]

[0046] In the formula, For each element in the second climate characteristic vector, is a constant;

[0047] Using the first climate feature vector and the second climate feature vector as inputs of the trained YOLO algorithm to output a first result and a second result, wherein the first result and the second result are control weight vectors corresponding to the first climate feature vector and the second climate feature vector, respectively;

[0048] Obtaining a first prediction result of the teacher model based on the standardized climate data sample and a second prediction result of the student model based on the same standardized climate data sample, and calculating a loss value between the teacher model and the student model according to the first prediction result and the second prediction result;

[0049] The calculation formula of the loss value is:

[0050]

[0051] In the formula, is the first prediction result, is the second prediction result.

[0052] The second aspect of the present invention discloses an air conditioning system operation strategy modeling device based on a mechanism framework + AI, the device comprising:

[0053] A data preprocessing module, used to obtain climate data and preprocess the climate data to obtain standardized climate data, wherein the preprocessing includes missing value filling processing and normalization processing;

[0054] An iterative clustering module, used for iteratively clustering the standardized climate data through a quantum oscillator model to generate regional groups with similar climate characteristics;

[0055] An operation strategy model establishment module is used to call a convolutional neural network in combination with a YOLO algorithm to train and learn the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample, so as to establish an operation strategy model for the air conditioning system;

[0056] Among them, the climate data includes at least temperature data, humidity data and wind speed data, the similar climate feature area group is a standardized climate data classification group with multiple clustering centers, and the convolutional neural network is divided into a teacher model and a student model. The teacher model is a pre-trained model obtained by pre-training based on the climate data, and the student model is a new task model that interacts with the teacher model for learning.

[0057] The third aspect of the present invention discloses a terminal, including a processor and a storage medium; the characteristics are:

[0058] The storage medium is used to store instructions;

[0059] The processor is used to operate according to the instructions to execute the steps of the method described in the first aspect.

[0060] The fourth aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0061] The beneficial effects of the present invention are that, compared with the prior art, the present invention has the following advantages:

[0062] By obtaining climate data such as temperature data, humidity data, and wind speed data, and preprocessing the obtained climate data with missing value filling and normalization, standardized climate data is obtained, and then the obtained standardized climate data is iteratively clustered through the quantum oscillator model to generate similar climate feature regional groups of standardized climate data classification groups containing multiple cluster centers after clustering. Finally, the convolutional neural network of the teacher-student model is called in combination with the YOLO algorithm to train and learn according to the climate data samples in the similar climate feature regional groups and the control strategy corresponding to each climate data sample to complete the modeling of the air-conditioning system operation strategy model. This method can dynamically adjust the modeling of the air-conditioning system operation strategy model according to real-time climate data, so that the final generated model can dynamically and reasonably adjust the load of each device during the operation of the air-conditioning based on the real-time climate data combined with the characteristics of the teacher-student model, which not only extends the service life of the air-conditioning system equipment, but also its intelligent control process also reduces the cost of operation and maintenance and replacement to a certain extent, and improves the economy of the overall operation of the air-conditioning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flow chart of an air-conditioning system operation strategy modeling method based on a mechanism framework + AI provided by the present invention. DETAILED DESCRIPTION

[0064] The present application is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present application.

[0065] like Figure 1 As shown, in one embodiment, a method for modeling an air conditioning system operation strategy based on a mechanism framework + AI comprises the following steps:

[0066] Step S110, acquiring climate data, and preprocessing the climate data to obtain standardized climate data, the preprocessing including missing value filling processing and normalization processing.

[0067] The climate data includes at least temperature data, humidity data and wind speed data.

[0068] In some embodiments, the present invention provides an air conditioning system operation strategy modeling method based on a mechanism framework + AI, which obtains climate data and preprocesses the climate data to obtain standardized climate data, specifically including the following steps:

[0069] Step S111, arrange the climate data in chronological order to obtain an initial climate data set consisting of time series data, wherein each climate data sample in the initial climate data set is represented in the form of a data series, including a temperature data time series, a humidity data time series, and a wind speed data time series.

[0070] Step S112, determining the time points where the data missing values ​​in the initial climate data set are located to form an index set, and filling the missing data values ​​by using the K nearest neighbor filling method.

[0071] Step S113, constructing a time series prediction model based on each climate data time series through an autoregressive moving average model, and normalizing all climate data samples to obtain a preprocessed standardized climate data set.

[0072] Among them, the modeling expression of the autoregressive moving average model is:

[0073]

[0074] In the formula, The index number is The data is missing values, is the autoregression coefficient, indicating the The weight of the lag term, is the moving average coefficient, indicating the The weight of the error lag term, and Respectively indicate time With time The random error between and are the orders of the autoregressive term and the moving average term, respectively.

[0075] Step S120, iteratively clustering the standardized climate data through a quantum oscillator model to generate groups of regions with similar climate characteristics.

[0076] In some embodiments, the present invention provides an air conditioning system operation strategy modeling method based on a mechanism framework + AI, which iteratively clusters standardized climate data through a quantum oscillator model to generate regional groups with similar climate characteristics, specifically including the following steps:

[0077] Step S121, calling the L-ISODATA algorithm to randomly select multiple standardized climate data samples from the standardized climate data set as initial cluster centers, the expression of the initial cluster center is:

[0078]

[0079] In the formula, is the number of standardized climate data samples, For the Cluster centers, , The values ​​are data sequences of standardized climate data samples.

[0080] Step S122, performing quantum energy assignment based on the initial cluster center, and calling the Gaussian sampling function to generate energy disturbance, so as to update the position of the cluster center according to the energy disturbance.

[0081] Among them, the expression for quantum energy assignment is:

[0082]

[0083] In the formula, For the Cluster Centers The quantum energy of is the proportionality factor, used to control the scale of the energy value, Represents the cluster center data density.

[0084] In some embodiments, the present invention provides an air conditioning system operation strategy modeling method based on a mechanism framework + AI, and the expression of data density is:

[0085]

[0086] In the formula, is the cluster center The number of climate data samples is a smoothing parameter used to control the decay rate of data density. Respectively represent The temperature data time series, humidity data time series and wind speed data time series of the climate data samples are Represents the cluster center No. row vector, .

[0087] The expression for updating the position of the cluster center is:

[0088]

[0089]

[0090] In the formula, is the updated cluster center position, is the energy disturbance and conforms to the normal distribution. is the learning rate constant and is a unit random direction vector satisfying .

[0091] In some embodiments, the present invention provides an air conditioning system operation strategy modeling method based on a mechanism framework + AI, which iteratively clusters standardized climate data through a quantum oscillator model to generate regional groups with similar climate characteristics, and specifically includes the following steps:

[0092] Step S123, calculate the data density and quantum energy assignment of the candidate location points of the updated cluster center, and compare the data density and quantum energy assignment of the candidate location points with the data density and quantum energy assignment of the cluster center before the update to determine whether to update the location and quantum energy assignment of the cluster center again.

[0093] Step S124, when the number of iterative updates of the cluster center reaches the set maximum number of iterations or the absolute value of the difference between the quantum energy assignment of the cluster center after the update and the quantum energy assignment of the cluster center before the update is less than the first threshold, the optimized updated cluster center set is output to obtain a group of regions with similar climate characteristics.

[0094] Among them, the quantum energy assignment of the cluster center after update is , the quantum energy of the cluster center before updating is assigned to ,when season , update the location of the cluster center and the quantum energy assignment.

[0095] Step S130, calling a convolutional neural network in combination with a YOLO algorithm to perform training and learning based on climate data samples in a group of regions with similar climate characteristics and a control strategy corresponding to each climate data sample, and establishing an operation strategy model for the air-conditioning system.

[0096] Among them, the convolutional neural network is divided into a teacher model and a student model. The teacher model is a pre-trained model obtained by pre-training based on climate data, and the student model is a new task model that interacts and learns with the teacher model.

[0097] In some embodiments, the present invention provides an air-conditioning system operation strategy modeling method based on a mechanism framework + AI, calling a convolutional neural network combined with a YOLO algorithm to train and learn according to climate data samples in a group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample, and establishing an air-conditioning system operation strategy model, specifically including the following steps:

[0098] Step S131, input each standardized climate data sample in the group of regions with similar climate characteristics into the teacher model to extract the first climate feature vector corresponding to each standardized climate data sample.

[0099] Step S132, taking the first climate characteristic vector and the control strategy corresponding to each standardized climate data sample as input of the YOLO algorithm, calling the YOLO algorithm to predict the control weight vector corresponding to each standardized climate data sample, and training the YOLO algorithm with the first climate characteristic vector and the control strategy as input and the control weight vector as output.

[0100] Among them, the length of the control weight vector is the number of control strategies, and the loss function expression of the YOLO algorithm is:

[0101]

[0102] In the formula, is the first elements, is the control weight vector of the cluster center corresponding to the climate data sample. elements, Indicates the number of climate data samples corresponding to the cluster center.

[0103] In some embodiments, the present invention provides an air-conditioning system operation strategy modeling method based on a mechanism framework + AI, calling a convolutional neural network combined with a YOLO algorithm to train and learn according to climate data samples in a group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample, and establishing an air-conditioning system operation strategy model, which specifically also includes the following steps:

[0104] Step S133, extracting the second climate feature vector of each standardized climate data sample in the group of regions with similar climate features through the teacher model, and calling the mapping function to map each element in the second climate feature vector. The expression of the mapping function is:

[0105]

[0106] In the formula, For each element in the second climate eigenvector, is a constant.

[0107] Step S134, using the first climate feature vector and the second climate feature vector as inputs of the trained YOLO algorithm to output a first result and a second result, wherein the first result and the second result are control weight vectors corresponding to the first climate feature vector and the second climate feature vector, respectively.

[0108] Step S135, obtain the first prediction result of the teacher model based on the standardized climate data sample and the second prediction result of the student model based on the same standardized climate data sample, and calculate the loss value between the teacher model and the student model according to the first prediction result and the second prediction result.

[0109] The calculation formula of the loss value is:

[0110]

[0111] In the formula, is the first prediction result, is the second prediction result.

[0112] The above-mentioned air-conditioning system operation strategy modeling method based on mechanism framework + AI obtains climate data such as temperature data, humidity data and wind speed data, and performs missing value filling and normalization preprocessing on the obtained climate data to obtain standardized climate data, and then iteratively clusters the obtained standardized climate data through the quantum oscillator model to generate clustered similar climate feature regional groups. Finally, the convolutional neural network of the teacher-student model is called in combination with the YOLO algorithm to train and learn according to the climate data samples in the similar climate feature regional groups and the control strategy corresponding to each climate data sample to complete the modeling of the air-conditioning system operation strategy model. This method can dynamically adjust the modeling of the air-conditioning system operation strategy model according to real-time climate data, so that the final generated model can dynamically and reasonably adjust the load of each device during the operation of the air-conditioning based on the real-time climate data combined with the characteristics of the teacher-student model, which not only extends the service life of the air-conditioning system equipment, but also its intelligent control process also reduces the cost of operation and maintenance and replacement to a certain extent, and improves the economy of the overall operation of the air-conditioning system.

[0113] In a specific embodiment, the present invention provides an air conditioning system operation strategy modeling method based on a mechanism framework + AI, comprising steps 1 to 3:

[0114] Step 1: Climate data acquisition and preprocessing.

[0115] Specifically, based on the climate data such as temperature, humidity, wind speed, etc. in the region, the climate data is filled with missing values ​​(such as the K nearest neighbor filling method) and normalized to obtain standardized climate data.

[0116] It should be noted that when the climate data corresponds to different parameters (temperature, humidity, wind speed), the control strategy of the air conditioning system is also different. The control strategies include:

[0117] Temperature control strategies: adjust compressor operating frequency, adjust fan speed, utilize free cooling, ventilate at night, intelligent defrost, adjust refrigerant flow, cold and heat storage technology, frequency conversion technology, and adjust air circulation path.

[0118] Humidity control strategy: adjust the operating frequency of the compressor, dynamically set the humidity target value, intelligently control the start and stop of the dehumidification function, intelligent defrosting, adjust the refrigerant flow, cold and heat storage technology, and frequency conversion technology.

[0119] Wind speed control strategy: dynamically set fresh air speed target value, frequency conversion technology, and adjust air circulation path.

[0120] Therefore, in this embodiment, the climate data should be divided into kind.

[0121] In this embodiment, step 1 includes steps 1.1 to 1.3:

[0122] Step 1.1, meteorological data acquisition.

[0123] Specifically, the acquired climate data are recorded in chronological order to form time series data, i.e., the initial climate data set. Each climate data sample It can be expressed as a data sequence, such as .

[0124] in, T=[ T 1 , T 2 ,…, T n ] , represents the temperature data time series, similarly, They represent the humidity data time series and the wind speed data time series respectively.

[0125] Step 1.2, fill in missing values.

[0126] Specifically, it includes steps 1.2.1~1.2.2:

[0127] Step 1.2.1, first determine the time point where the missing value is located and form an index set .

[0128] Step 1.2.2: Build a time series prediction model. For each time series ( ), using the autoregressive moving average model, the expression is:

[0129]

[0130] In the formula, The index number is The data is missing values, is the autoregression coefficient, indicating the The weight of the lag term, is the moving average coefficient, The two coefficients are determined based on maximum likelihood estimation. and Respectively indicate time With time The random error, and are the orders of the autoregressive term and the moving average term, respectively.

[0131] Step 1.3: Normalize the data.

[0132] Step 2: Cluster the climate data based on the quantum oscillator model.

[0133] Specifically, the L-ISODATA (Local Iterative Self-Organizing Data Analysis Technique) algorithm can be used. It is an unsupervised data clustering algorithm that iteratively clusters climate data to generate regional groups with similar climate characteristics.

[0134] In this embodiment, step 2 includes steps 2.1 to 2.5:

[0135] Step 2.1, random selection Climate data samples are used as the initial cluster centers, and the expression is:

[0136]

[0137] In the formula, For the Cluster centers, ; The value of is the data sequence of the sample. For a column number , a matrix with 3 rows, The value of is equal to 7.

[0138] Step 2.2, quantum energy assignment, the expression of quantum energy assignment is:

[0139]

[0140] In the formula, is the cluster center The quantum energy of is the proportional coefficient, which is used to control the scale of the capability value and can be set to 1. Represents the cluster center The data density is expressed as:

[0141]

[0142] In the formula, is the cluster center is the number of climate data samples and is a smoothing parameter used to control the decay rate of data density, which can be set to 1. Respectively represent The temperature data time series, humidity data time series and wind speed data time series of samples are It means No. row vector, .

[0143] Step 2.3, set the Gaussian sampling function and form the sampling area.

[0144] Specifically, the energy perturbation is generated using the Gaussian sampling function, and the cluster center position is updated according to the energy perturbation. The new cluster center position The update expression is:

[0145]

[0146]

[0147] In the formula, It is energy disturbance and conforms to normal distribution; is the learning rate constant, which can be set to 0.1, and is a unit random direction vector satisfying .

[0148] Step 2.4, compare the new cluster center energies and update.

[0149] Specifically, based on the formula in step 2.2, calculate the data density and quantum energy of the new cluster center candidate points If satisfied , then update the cluster center and quantum energy, that is, .

[0150] Step 2.5, repeat steps 2.3 to 2.4 until the maximum number of iterations is reached or quantum convergence is reached, that is , output the optimized updated cluster center set .

[0151] Step 3: For each sample of climate data The corresponding control strategy uses a convolutional neural network and combines it with the YOLO algorithm for learning to establish an operation strategy model for the air-conditioning system.

[0152] It should be noted that the core architecture of YOLO is based on convolutional neural networks (CNN) to process two-dimensional image data. However, in this embodiment, YOLO's grid division and local features are mainly used to predict the data time series, so as to better explore the influence of the corresponding type of operation strategy on the corresponding parameters. In addition, the reason why the historical data of different cluster centers are mixed together for learning is to unify the learning parameters in the operation strategy model.

[0153] In this embodiment, the convolutional neural network selects a teacher-student model, and step 3 includes steps 3.1 to 3.6:

[0154] Step 3.1, input each climate data sample into the teacher model and extract the corresponding first climate feature vector.

[0155] It should be noted that, whether it is a teacher model or a student model, it mainly includes a multi-layer convolutional neural network, and the convolutional neural network includes a residual feature extraction neural network.

[0156] Step 3.2, input the first climate feature vector and the control strategy corresponding to the sample of climate data into the YOLO algorithm, predict the corresponding control weight vector, and then complete the training of the YOLO algorithm.

[0157] Among them, the control weight vector is a length of A vector of is the number of control strategies, and each element in the control weight vector is used to feedback the contribution of the corresponding control strategy to the prediction result.

[0158] The loss function of the YOLO algorithm is shown as follows:

[0159]

[0160] In the formula, is the first elements, is the control weight vector of the cluster center corresponding to the climate data. elements, Indicates the number of samples of the cluster center.

[0161] In step 3.3, each sample of climate data is input into the teacher model to extract the corresponding second climate feature vector.

[0162] Specifically, each element in the second climate characteristic vector is mapped by the following mapping function, so as to prevent the first climate characteristic vector from damaging the second climate characteristic vector.

[0163] Among them, the mapping function must correspond one to one, and the mapping function The expression is:

[0164]

[0165] In the formula, Refers to each element in the second climate feature vector.

[0166] Step 3.4, input the first climate feature vector and the second climate feature vector into the trained YOLO algorithm to obtain a first result and a second result.

[0167] It should be noted that the first climate feature vector in step 3.1 (or the second climate feature vector in step 3.3) undergoes convolution operations, activation functions, pooling and other operations in the YOLO algorithm, and is finally output as the first result (or second result) in step 3.4.

[0168] Step 3.5, calculate the loss value of the teacher-student model based on the first detection result of the teacher model and the second detection result of the student model.

[0169] Specifically, the loss value As shown below:

[0170]

[0171] In the formula, is the first prediction result, is the second prediction result.

[0172] Step 3.6, determine whether the student model converges.

[0173] Specifically, if the sum of the predicted loss value and the value of the loss function of the Scaled-YOLO algorithm is less than the preset threshold, it is determined that the student model has converged and the training is stopped; otherwise, it is determined that the student model has not converged and returns to step 3.1 to continue iteration.

[0174] The following describes an air-conditioning system operation strategy modeling device based on a mechanism framework + AI provided by the present invention. The air-conditioning system operation strategy modeling device based on a mechanism framework + AI described below and the air-conditioning system operation strategy modeling method based on a mechanism framework + AI described above can be referenced to each other.

[0175] In one embodiment, an air conditioning system operation strategy modeling device based on a mechanism framework + AI includes a data preprocessing module, an iterative clustering module, and an operation strategy model building module.

[0176] The data preprocessing module is used to obtain climate data and preprocess the climate data to obtain standardized climate data. The preprocessing includes missing value filling and normalization.

[0177] The iterative clustering module is used to iteratively cluster the standardized climate data through the quantum oscillator model to generate regional groups with similar climate characteristics.

[0178] The operation strategy model establishment module is used to call the convolutional neural network combined with the YOLO algorithm to train and learn the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample, and establish the operation strategy model of the air-conditioning system.

[0179] Among them, the climate data includes at least temperature data, humidity data and wind speed data. The convolutional neural network is divided into a teacher model and a student model. The teacher model is a pre-trained model obtained by pre-training based on the climate data, and the student model is a new task model that interacts with the teacher model for learning.

[0180] In this embodiment, the present invention provides an air conditioning system operation strategy modeling device based on a mechanism framework + AI, and the data preprocessing module is specifically used for:

[0181] The climate data are arranged in chronological order to obtain an initial climate data set consisting of time series data. Each climate data sample in the initial climate data set is represented in the form of a data series, including a temperature data time series, a humidity data time series, and a wind speed data time series.

[0182] The time points where the data missing values ​​are located in the initial climate dataset are determined to form an index set, and the missing values ​​are filled in using the K nearest neighbor filling method.

[0183] A time series prediction model is constructed based on each climate data time series through the autoregressive moving average model, and all climate data samples are normalized to obtain the preprocessed standardized climate data set.

[0184] Among them, the modeling expression of the autoregressive moving average model is:

[0185]

[0186] In the formula, The index number is The data is missing values, is the autoregression coefficient, indicating the The weight of the lag term, is the moving average coefficient, indicating the The weight of the error lag term, and Respectively indicate time With time The random error between and are the orders of the autoregressive term and the moving average term, respectively.

[0187] In this embodiment, the present invention provides an air conditioning system operation strategy modeling device based on a mechanism framework + AI, and the iterative clustering module is specifically used for:

[0188] The L-ISODATA algorithm is called to randomly select multiple standardized climate data samples from the standardized climate data set as the initial clustering centers. The expression of the initial clustering center is:

[0189]

[0190] In the formula, is the number of standardized climate data samples, For the Cluster centers, , The values ​​are data sequences of standardized climate data samples.

[0191] Quantum energy is assigned based on the initial cluster center, and the Gaussian sampling function is called to generate energy perturbations to update the positions of the cluster centers according to the energy perturbations.

[0192] Among them, the expression for quantum energy assignment is:

[0193]

[0194] In the formula, For the Cluster Centers The quantum energy of is the proportionality factor, used to control the scale of the energy value, Represents the cluster center data density.

[0195] In this embodiment, the present invention provides an air conditioning system operation strategy modeling device based on a mechanism framework + AI, and the expression of data density is:

[0196]

[0197] In the formula, is the cluster center The number of climate data samples is a smoothing parameter used to control the decay rate of data density. Respectively represent The temperature data time series, humidity data time series and wind speed data time series of the climate data samples are Represents the cluster center No. row vector, .

[0198] The expression for updating the position of the cluster center is:

[0199]

[0200]

[0201] In the formula, is the updated cluster center position, is the energy disturbance and conforms to the normal distribution. is the learning rate constant and is a unit random direction vector satisfying .

[0202] In this embodiment, the present invention provides an air conditioning system operation strategy modeling device based on a mechanism framework + AI, and the iterative clustering module is specifically used for:

[0203] The data density and quantum energy assignment of the candidate location points of the updated cluster center are calculated, and the data density and quantum energy assignment of the candidate location points are compared with the data density and quantum energy assignment of the cluster center before the update to determine whether to update the location and quantum energy assignment of the cluster center again.

[0204] When the number of iterative updates of the cluster center reaches the set maximum number of iterations or the absolute value of the difference between the quantum energy assignment of the cluster center after the update and the quantum energy assignment of the cluster center before the update is less than the first threshold, the optimized updated cluster center set is output to obtain a group of regions with similar climate characteristics.

[0205] Among them, the quantum energy assignment of the cluster center after update is , the quantum energy of the cluster center before updating is assigned to ,when season , update the location of the cluster center and the quantum energy assignment.

[0206] In this embodiment, the present invention provides an air conditioning system operation strategy modeling device based on a mechanism framework + AI, and the operation strategy model building module is specifically used for:

[0207] Each standardized climate data sample in the group of regions with similar climate characteristics is input into the teacher model to extract the first climate feature vector corresponding to each standardized climate data sample.

[0208] The first climate feature vector and the control strategy corresponding to each standardized climate data sample are used as input of the YOLO algorithm to call the YOLO algorithm to predict the control weight vector corresponding to each standardized climate data sample, and the YOLO algorithm is trained with the first climate feature vector and the control strategy as input and the control weight vector as output.

[0209] Among them, the length of the control weight vector is the number of control strategies, and the loss function expression of the YOLO algorithm is:

[0210]

[0211] In the formula, is the first elements, is the control weight vector of the cluster center corresponding to the climate data sample. elements, Indicates the number of climate data samples corresponding to the cluster center.

[0212] In this embodiment, the present invention provides an air conditioning system operation strategy modeling device based on a mechanism framework + AI, and the operation strategy model building module is specifically used for:

[0213] The second climate feature vector of each standardized climate data sample in the group of regions with similar climate characteristics is extracted through the teacher model, and the mapping function is called to map each element in the second climate feature vector. The expression of the mapping function is:

[0214]

[0215] In the formula, For each element in the second climate eigenvector, is a constant.

[0216] The first climate feature vector and the second climate feature vector are used as inputs of the trained YOLO algorithm to output a first result and a second result, where the first result and the second result are control weight vectors corresponding to the first climate feature vector and the second climate feature vector, respectively.

[0217] Obtain a first prediction result of the teacher model based on a standardized climate data sample and a second prediction result of the student model based on the same standardized climate data sample, and calculate the loss value between the teacher model and the student model according to the first prediction result and the second prediction result.

[0218] The calculation formula of the loss value is:

[0219]

[0220] In the formula, is the first prediction result, is the second prediction result.

[0221] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0222] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the above. The computer-readable storage medium used herein is not to be interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0223] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0224] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0225] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0226] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0227] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0228] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0229] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for modeling air-conditioning system operation strategy based on mechanism framework + AI, characterized in that: The method comprises: Acquire climate data, and preprocess the climate data to obtain standardized climate data, wherein the preprocessing includes missing value filling processing and normalization processing; Iteratively clustering the standardized climate data through a quantum oscillator model to generate regional groups with similar climate characteristics; Calling a convolutional neural network in combination with a YOLO algorithm to perform training and learning based on the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample, and establishing an operation strategy model for the air-conditioning system; Among them, the climate data includes at least temperature data, humidity data and wind speed data, and the convolutional neural network is divided into a teacher model and a student model. The teacher model is a pre-trained model obtained by pre-training based on the climate data, and the student model is a new task model that interacts and learns with the teacher model.

2. The air conditioning system operation strategy modeling method based on mechanism framework + AI according to claim 1 is characterized in that: The step of acquiring climate data and preprocessing the climate data to obtain standardized climate data includes: Arranging the climate data in chronological order to obtain an initial climate data set consisting of time series data, wherein each climate data sample in the initial climate data set is represented in the form of a data sequence, including a temperature data time series, a humidity data time series, and a wind speed data time series; Determine the time points where the data missing values ​​in the initial climate data set are located to form an index set, and fill the data missing values ​​by using a K-nearest neighbor filling method; and A time series prediction model is constructed based on each climate data time series through an autoregressive moving average model, and all climate data samples are normalized to obtain a preprocessed standardized climate data set; Among them, the modeling expression of the autoregressive moving average model is: In the formula, The index number is The data is missing values, is the autoregression coefficient, indicating the The weight of the lag term, is the moving average coefficient, indicating the The weight of the error lag term, and Respectively indicate time With time The random error between and are the orders of the autoregressive term and the moving average term, respectively.

3. The air conditioning system operation strategy modeling method based on mechanism framework + AI according to claim 2 is characterized in that: The method of iteratively clustering the standardized climate data by using a quantum oscillator model to generate regional groups with similar climate characteristics comprises: The L-ISODATA algorithm is called to randomly select multiple standardized climate data samples from the standardized climate data set as initial cluster centers. The expression of the initial cluster center is: ； In the formula, is the number of standardized climate data samples, For the Cluster centers, , The value of is a data sequence of standardized climate data samples; Performing quantum energy assignment based on the initial cluster center, and calling a Gaussian sampling function to generate energy disturbance, so as to update the position of the cluster center according to the energy disturbance; Among them, the expression of the quantum energy assignment is: ； In the formula, For the Cluster Centers The quantum energy of is the proportionality factor, used to control the scale of the energy value, Represents the cluster center data density.

4. The air conditioning system operation strategy modeling method based on mechanism framework + AI according to claim 3 is characterized in that: The expression of the data density is: In the formula, is the cluster center The number of climate data samples is a smoothing parameter used to control the decay rate of data density. Respectively represent The temperature data time series, humidity data time series and wind speed data time series of the climate data samples are Represents the cluster center No. row vector, ; The expression for updating the position of the cluster center is: ， ； In the formula, is the updated cluster center position, is the energy disturbance and conforms to the normal distribution. is the learning rate constant and is a unit random direction vector satisfying .

5. The air conditioning system operation strategy modeling method based on mechanism framework + AI according to claim 4 is characterized in that: The iterative clustering of the standardized climate data by the quantum oscillator model to generate regional groups with similar climate characteristics also includes: Calculating the data density and quantum energy assignment of the candidate location points of the updated cluster center, and comparing the data density and quantum energy assignment of the candidate location points with the data density and quantum energy assignment of the cluster center before the update, to determine whether to update the location and quantum energy assignment of the cluster center again; When the number of iterative updates of the cluster center reaches a set maximum number of iterations or the absolute value of the difference between the quantum energy assignment of the cluster center after the update and the quantum energy assignment of the cluster center before the update is less than a first threshold, the optimized updated cluster center set is output to obtain the group of regions with similar climate characteristics; Among them, the quantum energy assignment of the cluster center after update is , the quantum energy of the cluster center before updating is assigned to ,when season , update the location of the cluster center and the quantum energy assignment.

6. The air conditioning system operation strategy modeling method based on mechanism framework + AI according to claim 5 is characterized in that: The calling of the convolutional neural network in combination with the YOLO algorithm is used to train and learn the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample to establish an operation strategy model of the air conditioning system, including: Inputting each standardized climate data sample in the group of regions with similar climate characteristics into the teacher model to extract a first climate feature vector corresponding to each standardized climate data sample; Taking the first climate feature vector and the control strategy corresponding to each standardized climate data sample as input of the YOLO algorithm, calling the YOLO algorithm to predict the control weight vector corresponding to each standardized climate data sample, and taking the first climate feature vector and the control strategy as input and the control weight vector as output, training the YOLO algorithm; Among them, the length of the control weight vector is the number of the control strategies, and the loss function expression of the YOLO algorithm is: In the formula, is the first elements, is the control weight vector of the cluster center corresponding to the climate data sample. elements, Indicates the number of climate data samples corresponding to the cluster center.

7. The air conditioning system operation strategy modeling method based on mechanism framework + AI according to claim 6 is characterized in that: The calling of the convolutional neural network in combination with the YOLO algorithm to perform training and learning based on the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample to establish an operation strategy model of the air conditioning system also includes: The second climate feature vector of each standardized climate data sample in the group of regions with similar climate features is extracted by the teacher model, and a mapping function is called to map each element in the second climate feature vector. The expression of the mapping function is: In the formula, For each element in the second climate characteristic vector, is a constant; Using the first climate feature vector and the second climate feature vector as inputs of the trained YOLO algorithm to output a first result and a second result, wherein the first result and the second result are control weight vectors corresponding to the first climate feature vector and the second climate feature vector, respectively; Obtaining a first prediction result of the teacher model based on the standardized climate data sample and a second prediction result of the student model based on the same standardized climate data sample, and calculating a loss value between the teacher model and the student model according to the first prediction result and the second prediction result; The calculation formula of the loss value is: In the formula, is the first prediction result, is the second prediction result.

8. An air conditioning system operation strategy modeling device based on mechanism framework + AI, characterized in that: The device comprises: A data preprocessing module, used to obtain climate data and preprocess the climate data to obtain standardized climate data, wherein the preprocessing includes missing value filling processing and normalization processing; An iterative clustering module, used for iteratively clustering the standardized climate data through a quantum oscillator model to generate regional groups with similar climate characteristics; An operation strategy model establishment module is used to call a convolutional neural network in combination with a YOLO algorithm to train and learn the climate data samples in the group of regions with similar climate characteristics and the control strategy corresponding to each climate data sample, so as to establish an operation strategy model for the air conditioning system; Among them, the climate data includes at least temperature data, humidity data and wind speed data, and the convolutional neural network is divided into a teacher model and a student model. The teacher model is a pre-trained model obtained by pre-training based on the climate data, and the student model is a new task model that interacts and learns with the teacher model.

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

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images