WO2022052547A1

WO2022052547A1 - Method for predicting energy efficiency of air-conditioning system, and air-conditioning system

Info

Publication number: WO2022052547A1
Application number: PCT/CN2021/100427
Authority: WO
Inventors: 盛凯; 矫晓龙; 任兆亭
Original assignee: 青岛海信日立空调系统有限公司
Priority date: 2020-09-14
Filing date: 2021-06-16
Publication date: 2022-03-17
Also published as: CN114186469A

Abstract

The present application provides a method for predicting energy efficiency of an air-conditioning system, comprising: acquiring past data of an air conditioning system to form a data set, and preprocessing the data set; on the basis of the processed data set, establishing an air conditioning energy efficiency prediction model based on an integrated CNN-LSTM; and acquiring real-time operating data of the air-conditioning system and preprocessing the data, and inputting the processed real-time data into the established air-conditioning energy efficiency prediction model to obtain an air-conditioning energy efficiency prediction value at time t+1.

Description

Air conditioning system energy efficiency prediction method and air conditioning system

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the priority of the Chinese patent application filed on September 14, 2020 with the application number 202010959889.1 and the title of the invention is "air conditioning system energy efficiency prediction method and air conditioning system", the entire contents of which are incorporated into this application by reference middle.

technical field

The present application relates to the technical field of air conditioning, and in particular, to a method for predicting energy efficiency of an air conditioning system and an air conditioning system.

Background technique

With the improvement of people's living standards, air conditioners have increasingly become one of the indispensable electrical appliances in people's lives, and at the same time, there are higher requirements for the intelligent level of air conditioners. The intelligent level of air-conditioning is not only reflected in the control, but also in the control of air-conditioning energy consumption. Manufacturers are striving to improve the energy efficiency ratio of air conditioners, so that air conditioners can operate at higher energy efficiency and meet the requirements of energy saving.

SUMMARY OF THE INVENTION

In some embodiments of the present application, a method for predicting energy efficiency of an air conditioning system is proposed, including:

S1: Collect historical data of the air conditioning system to form a data set, and preprocess the data set;

S2: Based on the processed data set, establish an air conditioning energy efficiency prediction model based on a comprehensive CNN-LSTM;

S3: Collect real-time operating data of the air-conditioning system and preprocess the data, input the processed real-time data into the established air-conditioning energy efficiency prediction model, and obtain the air-conditioning energy efficiency prediction value at time t+1.

In some embodiments of the present application, an air conditioning system is also proposed, including:

parameter sensor;

The controller controls the parameter sensor to measure the parameter value of the air-conditioning system, and can execute any one of the above-mentioned methods for predicting the energy efficiency of the air-conditioning system.

Description of drawings

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

FIG. 1 is a flowchart of a method for predicting energy efficiency of an air conditioning system according to some embodiments of the present application;

FIG. 2 is a partial flowchart of a method for predicting energy efficiency of an air conditioning system according to some embodiments of the present application;

Fig. 3 is the loss function diagram of the CNN-LSTM model of some embodiments of the present application;

Fig. 4 is the loss function diagram of the LSTM model of some embodiments of the present application;

5 is a comparison diagram of the predicted value and the test value of the energy consumption ratio of the CNN-LSTM model of some embodiments of the present application;

6 is a comparison diagram of the predicted value and the test value of the energy consumption ratio of the LSTM model of some embodiments of the present application;

7 is a comparison diagram of the predicted value and the test value of the energy consumption ratio of the CNN-LSTM model of some embodiments of the present application;

FIG. 8 is a comparison diagram of the predicted value and the test value of the energy consumption ratio of the LSTM model according to some embodiments of the present application.

detailed description

There are mainly two ways to improve the energy efficiency ratio. One is to start with the forward control of the air conditioner itself, obtain the detailed operating parameters of the air conditioner through experiments, and adjust the control algorithm to achieve energy-saving control. The other aspect is to predict the energy efficiency ratio of the air conditioner. A feedback input from the control system to complete more precise control of the air conditioner. Linear regression, Bayesian estimation algorithm, genetic algorithm, etc. can predict the energy efficiency of air conditioners, but the energy efficiency of air conditioners is affected by many factors and parameters, and it is a very complex nonlinear system. prediction effect. For the recurrent neural network and its variant long short-term memory network LSTM, the LSTM algorithm has achieved good engineering results, and its algorithm logic is very similar to the operating mechanism of the air conditioner, but it can only be predicted from the time series, ignoring some characteristics of the reference variable itself.

The method for predicting the energy efficiency of an air conditioning system and the air conditioning system of the present application will be described below with reference to FIGS. 1-8 . FIGS. 1 and 2 are flowcharts of methods for predicting the energy efficiency of an air conditioning system according to some embodiments of the present application.

S1: Collect historical data of the air-conditioning system and form a data set, and preprocess the data set;

S2: Establish an air conditioning energy efficiency prediction model based on a comprehensive CNN-LSTM;

S3: Collect real-time data of the air-conditioning system and preprocess the data, input the processed real-time data into the established air-conditioning energy efficiency prediction model, and obtain the air-conditioning energy efficiency prediction value at time t+1.

By combining the convolutional neural network CNN in deep learning and the long short-term memory network LSTM, a prediction model combining high-dimensional vector space and time series is established to more accurately predict the energy efficiency of the air conditioning system, and then take timely measures. Control, in order to meet the premise of comfort, save energy consumption.

In step S1, when collecting the historical data of the air-conditioning system, it is necessary to determine the parameters that affect the energy efficiency ratio of the air-conditioning system. When selecting parameters, the parameters that have a greater impact on the energy efficiency ratio are mainly selected, including evaporator outlet water temperature (TEI), evaporation Outlet temperature (TEO), condenser inlet temperature (TCI), condenser outlet temperature (TCO), heat exchanger inlet temperature (TSI), condenser loop temperature (TS0), heat exchanger outlet temperature (TBI) ), evaporator loop temperature (TBO), building inlet water temperature (Cond Tons), building outlet water temperature (Cooling Tons), steam heating capacity (kW), evaporator flow (TEA), condenser flow (TCA), condensation Temperature difference characterization value (TRE), refrigerant temperature (TRC), the above 15 parameter values can be obtained through sensors.

The parameter values and energy efficiency ratios of the air-conditioning system at a time point are collected to form a parameter vector Xi = [x ₁ , x ₂ ,..., _{x n} _, y], where x ₁ , x ₂ ,..., x _n is the influence The parameter values corresponding to the n parameters of the energy efficiency ratio of the air conditioner, n indicates that a total of n parameters are collected, and y is the energy efficiency ratio at this time point. The air conditioner can be obtained by collecting the parameter values and energy efficiency ratios of the air conditioning system at different time points. The historical data set X of the system, X=[X ₁ , X ₂ ,..., X _t ], where t represents the time point of data collection, data collection is carried out at certain time intervals, and the collection forms data with certain time characteristics set. Specifically, the parameter values of the above-mentioned 15 parameters that affect the energy efficiency ratio can be selected, and the energy efficiency ratio COP can be obtained by calculation, COP=cooling capacity/loss of power, X _i =[TEI,TEO,TCI,TCO,TSI,TSO,TBI ,TBO,Cond Tons, Cooling Tons,Kw,TEA,TCA,TRE,TRC,COP], the dimension of this vector is 16 dimensions.

In step S1, preprocessing the data set includes:

One-hot encoding the data of each independent variable;

Then normalize the data set: normalize each data in the data set to the range of [0,1], remove the unit limit of the data, and convert it into a dimensionless value. The conversion function is:

x*=(xx _min )/(x _max -x _min ),

where x is the value before conversion, x _max is the maximum value in each column of data, x _min is the minimum value in each column of data, and x* is the value after conversion.

Among them, each column of parameters in the data set is processed during normalization.

The normalized data set is divided into training set and test set, which can be split according to the ratio of A:B, and the split ratio can be 8:2. Specifically, if the collected data is 5000 vectors, then 4000 One is the training set and 1000 is the test set.

In step S2, based on the processed data set, the specific steps for establishing an air conditioning energy efficiency prediction model based on a comprehensive CNN-LSTM are:

S21: Input the preprocessed dataset into the CNN network to obtain a time series with high-dimensional feature information;

S22: Input the time series with high-dimensional feature information into the LSTM network for training to obtain a CNN-LSTM model.

In step S21, the specific steps of obtaining a time series with high-dimensional feature information are:

Select samples from the preprocessed training set and input the samples into the convolution function

Extract high-dimensional feature information; extract important features through the pooling layer, and flatten the input into a one-dimensional vector through the Flatten layer to obtain a time series with high-dimensional feature information.

Specifically, the principle of data prediction and sample selection is to use the data of the first N rows to predict the last COP data of the N+1th row, and so on to slide according to the time window, the sliding step is 1, and the N rows are equivalent to the sliding window. The width is N.

The LSTM network includes a memory unit, and the memory unit includes a forgetting gate, an input gate and an output gate, which can selectively memorize the correction parameters of the feedback loss function with the gradient descent. In step S22, the time series with high-dimensional feature information is input into the LSTM The network is trained, and the input time series is calculated as follows in the LSTM network:

Forgetting gate: f _t =σ(W _f ·[h _t-1 ,x _t ]+b _f );

f _t is the output value of the forget gate, W _f is the weight of the forget gate neural network, h _t-1 is the output of the node at time t-1, x _t is the input of the node at time t, b _f is the input of the forget gate neural network Bias.

Input gate: i _t =σ(W _i ·[h _t-1 ,x _t ]+ _bi );

_i _t is the output value of the input gate, Wi is the weight of the input gate neural network, h _t-1 is the output of the node at time t-1, x _t is the input of the node at time t, and b _i is the input gate neural network. Bias.

Output gate: o _t =σ(W _o ·[h _t-1 ,x _t ]+b _o );

o _t is the output value of the output gate, W _o is the weight of the output gate neural network, h _t-1 is the output of the node at time t-1, x _t is the input of the node at time t, and b _o is the output gate neural network. Bias.

Unit input:

W _c is the weight of the input of the unit, h _t-1 is the output of the node at time t-1, x _t is the input of the node at time t, and b _c is the bias of the input of the unit.

Unit output:

• Represents an element-wise multiplication operation of a matrix.

Node output value: h _{t =} o _t ·tanh(c _t );

h _t is the output of the node at time t, and represents the element-wise multiplication operation of the matrix.

Fusion activation function in CNN-LSTM model:

During the training process, the LSTM network is converted into the corresponding output value through the fully connected layer Dense, the loss function selects the mean square error loss function, and the optimizer selects the Adam optimizer.

After establishing the CNN-LSTM model, select samples in the test set to test the accuracy of the CNN-LSTM model, and calculate the error of the CNN-LSTM model. If the error is greater than the set value, increase the number of training times to obtain a new CNN-LSTM model. If it is less than or equal to the set value, the accuracy of the CNN-LSTM model is evaluated to meet the conditions. Specifically, the root mean square error RSME is used to evaluate the prediction accuracy of the model.

Among them, n is the total number of evaluation samples, y _i ^obj is the actual value of the ith sample, y _i ^model is the model predicted value of the ith sample, and the RMSE model predicts the root mean square error between the data and the real data.

The principle of sample selection during data testing is to use the data of the first N rows to predict the last COP data of the N+1th row, and so on to slide according to the time window, the sliding step is 1, and the N rows are equivalent to the width of the sliding window N. .

After the CNN-LSTM model is determined, the real-time data of the air-conditioning system is collected and the data is preprocessed, and the processed real-time data is input into the CNN-LSTM model, and the air-conditioning COP at time t+1 can be obtained. Processing is the same as for historical data.

Specifically, more than 5,000 sets of data are collected for the air-conditioning system to obtain an initial data set. Some samples of the collected data set are shown in Table 1.

Table 1

The deep learning network based on CNN and LSTM consists of input layer, 1D-CNN layer, pooling layer, LSTM layer, fully connected layer and output layer. The loss function selects the mean square error loss function, and the optimizer selects the Adam optimizer. Among them, the deep learning network model based on 1D-CNN and LSTM is shown in Table 2.

Table 2

层名称layer name	输出张量形状output tensor shape	参数个数Number of parameters
dense_1(Dense)dense_1(Dense)	(None,5,128)(None,5,128)	21762176
conv1d_1(Conv1D)conv1d_1(Conv1D)	(None,5,80)(None,5,80)	1032010320
max_pooling1d_1max_pooling1d_1	(None,2,80)(None,2,80)	00
conv1d_2(Conv1D)conv1d_2(Conv1D)	(None,2,48)(None,2,48)	38883888
max_pooling1d_2max_pooling1d_2	(None,2,48)(None,2,48)	00
dropout_1(Dropout)dropout_1 (Dropout)	(None,2,48)(None,2,48)	00
lstm_1(LSTM)lstm_1 (LSTM)	(None,2,32)(None,2,32)	1036810368
lstm_2(LSTM)lstm_2 (LSTM)	(None,16)(None,16)	31363136
dense_2(Dense)dense_2(Dense)	(None,32)(None,32)	544544
dense_3(Dense)dense_3(Dense)	(None,1)(None,1)	3333

FIG. 3 is a loss function diagram of the CNN-LSTM model of some embodiments of the present application, and FIG. 4 is a loss function diagram of the LSTM model of some embodiments of the present application. It can be seen that the CNN-LSTM model converges quickly and the loss value is small, indicating that the CNN - The LSTM model is more accurate in prediction than the LSTM model.

FIG. 5 is a comparison diagram of the predicted value and test value of the energy consumption ratio of the CNN-LSTM model according to some embodiments of the present application, and FIG. 6 is a comparison diagram of the predicted value and test value of the energy consumption ratio of the LSTM model according to some embodiments of the present application. Compared with the 1000 time series of the test set, it can be seen from Figure 5 and Figure 6 that the predicted value of the CNN-LSTM model is closer to the test value, indicating that the CNN-LSTM model is more accurate than the LSTM model. FIG. 7 is a comparison diagram of the predicted value and the test value of the energy consumption ratio of the CNN-LSTM model of some embodiments of the present application, and FIG. 8 is a comparison diagram of the predicted value and the test value of the energy consumption ratio of the LSTM model of some embodiments of the present application, where is Compared with the 5000 time series of the total data set, it can be seen from Figure 7 and Figure 8 that the predicted value of the CNN-LSTM model is closer to the test value, indicating that the CNN-LSTM model is more accurate than the LSTM model. Among them, the COP Value in Figure 5-Figure 8 is the normal COP value reduced by 100 times.

Some embodiments of the present application further provide an air conditioning system, including a parameter sensor and a controller, the controller controls the parameter sensor to measure the parameter value of the air conditioning system, and can execute the above-mentioned method for predicting the energy efficiency of the air conditioning system.

By setting the controller of the air-conditioning system, the energy-efficiency prediction method of the air-conditioning system can be run, so that the air-conditioning system can take precise control according to the prediction result, improve the user experience, and save energy consumption.

The air-conditioning system also includes structures such as a compressor, a condenser, an expansion valve, and an evaporator, which can perform the cooling and/or heating functions of the air-conditioning system.

The above are only specific embodiments of the present application, but the protection scope of the present application is not limited to this. should be covered within the scope of protection of this application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

A method for predicting energy efficiency of an air conditioning system, comprising:

S1: Collect historical data of the air conditioning system to form a data set, and preprocess the data set;

S2: Based on the processed data set, establish an air conditioning energy efficiency prediction model based on a comprehensive CNN-LSTM;

S3: Collect real-time operating data of the air-conditioning system and preprocess the data, input the processed real-time data into the established air-conditioning energy efficiency prediction model, and obtain the air-conditioning energy efficiency prediction value at time t+1.
The method for predicting energy efficiency of an air-conditioning system according to claim 1, wherein in step S2, based on the processed data set, the specific steps of establishing an air-conditioning energy efficiency prediction model based on a comprehensive CNN-LSTM are:

S21: Input the preprocessed dataset into the CNN network to obtain a time series with high-dimensional feature information;

S22: Input the time series with high-dimensional feature information into the LSTM network for training to obtain a CNN-LSTM model.
The method for predicting energy efficiency of an air-conditioning system according to claim 2, wherein in step S21, the specific steps of obtaining a time series with high-dimensional feature information are:

Select samples from the preprocessed dataset and input the samples into the convolution function
Extract high-dimensional feature information; extract important features through the pooling layer, and flatten the input into a one-dimensional vector through the Flatten layer to obtain a time series with high-dimensional feature information.
The method for predicting energy efficiency of an air-conditioning system according to claim 2, wherein the LSTM network includes a memory unit, and the memory unit includes a forget gate, an input gate and an output gate, and in step S22, the input time series is performed in the LSTM network as follows calculate:

Forgetting gate: f t =σ(W f ·[h t-1 ,x t ]+b f );

Input gate: i t =σ(W i ·[h t-1 ,x t ]+ bi );

Output gate: o t =σ(W o ·[h t-1 ,x t ]+b o );

Unit input:

Unit output:

Node output value: h t = o t ·tanh(c t );

W f , Wi , W o , W c are weights, b f , b i , b o , b c are biases, h t-1 is the output value of the node at time t-1, and x t is the node at time t the input value.
The method for predicting energy efficiency of an air-conditioning system according to claim 2, characterized in that, in the training process, the loss function selects the mean square error loss function, and the optimizer selects the Adam optimizer.
The method for predicting energy efficiency of an air-conditioning system according to claim 1, wherein in step S1, the data set formed by collecting historical data of the air-conditioning system is X,

X=[X 1 , X 2 ,...,X t ], where t represents the time point of data collection;

X t =[x 1 , x 2 ,..., x n , y], where x 1 , x 2 ,..., x n are the parameter values corresponding to the n parameters that affect the energy efficiency ratio of the air conditioner, and y is the time point energy efficiency ratio.
The method for predicting energy efficiency of an air-conditioning system according to claim 6, wherein in the step S1, preprocessing the data set includes:

One-hot encoding the data of each independent variable;

Then normalize the data set: convert each data in the data set to the range of [0,1], remove the unit limit of the data, and convert it into a dimensionless value. The conversion function is:

x*=(xx min )/(x max -x min ),

where x is the value before conversion, x max is the maximum value in each column of data, x min is the minimum value in each column of data, and x* is the value after conversion.
The method for predicting energy efficiency of an air conditioning system according to claim 7, wherein the normalized data set is classified and divided into a training set and a test set according to the ratio of A:B.
The method for predicting energy efficiency of an air-conditioning system according to claim 8, wherein the model prediction accuracy is evaluated by using the root mean square error (RSME),

Among them, n is the total number of evaluation samples, y i obj is the actual value of the ith sample, and y i model is the model predicted value of the ith sample.
An air conditioning system, comprising:

parameter sensor;

The controller controls the parameter sensor to measure the parameter value of the air conditioning system, and can execute the method for predicting the energy efficiency of the air conditioning system according to any one of claims 1-9.