CN108320016B - Short-term prediction method for building energy consumption - Google Patents
Short-term prediction method for building energy consumption Download PDFInfo
- Publication number
- CN108320016B CN108320016B CN201810191237.0A CN201810191237A CN108320016B CN 108320016 B CN108320016 B CN 108320016B CN 201810191237 A CN201810191237 A CN 201810191237A CN 108320016 B CN108320016 B CN 108320016B
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- data
- building
- model
- term
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005265 energy consumption Methods 0.000 title claims abstract description 245
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000005286 illumination Methods 0.000 claims abstract description 62
- 238000013135 deep learning Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims description 62
- 238000012360 testing method Methods 0.000 claims description 22
- 230000000694 effects Effects 0.000 claims description 17
- 210000002569 neuron Anatomy 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000012795 verification Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000012544 monitoring process Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000000630 rising effect Effects 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 3
- 238000012300 Sequence Analysis Methods 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000004378 air conditioning Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 230000007177 brain activity Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The application discloses a short-term prediction method of building energy consumption, which comprises the steps of collecting historical data of building sub-term energy consumption, and determining and collecting historical data of main influencing factors influencing the prediction of the building energy consumption sub-term; analyzing and determining to construct an illumination energy consumption prediction model based on the time sequence autoregressive model; constructing an energy consumption prediction model based on a deep learning DBN network; and predicting the energy consumption, the power energy consumption and the special energy consumption of the air conditioner according to the items. The application has the beneficial effects that: the short-term prediction method for building energy consumption provided by the application can be used for more accurately and effectively predicting each item of energy consumption in the building energy consumption.
Description
Technical Field
The application relates to the technical field of building energy consumption prediction, in particular to a building energy consumption prediction method based on a time sequence AR model and a deep confidence network.
Background
In recent years, as the power consumption of buildings rises year by year, the building energy consumption has become a main object of building energy conservation supervision and reconstruction. The main prediction methods currently include: multiple linear regression, artificial neural network, bayes theory, gray theory, etc. Most of the prediction methods belong to shallow structure algorithms, the learning effect on complex nonlinear relations in a high-dimensional data sample is poor, deep learning is an algorithm for simulating human brain activities to analyze, the network topology structure is continuously optimized through shallow-to-deep progressive learning, and ideal learning parameters are selected, so that the problem that the training effect of the shallow structure algorithm in a multi-hidden-layer network is not ideal is effectively avoided. Unlike many other machine learning methods, deep learning can automatically learn data valid features from a large amount of unidentified historical data, and has strong data classification recognition and data prediction capabilities.
The deep belief network (deep belief network, DBN) is one of the most widely used learning models in deep learning, while combining a time series of auto-regressive (AR) models with the DBN network to predict the fractional energy consumption of large buildings has not been involved.
Disclosure of Invention
This section is intended to outline some aspects of embodiments of the application and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section as well as in the description of the application and in the title of the application, which may not be used to limit the scope of the application.
The present application has been made in view of the above-mentioned problems with the existing short-term prediction method of building energy consumption.
It is therefore an object of the present application to provide a method which enables a more accurate and efficient prediction of the individual sub-term energy consumption in the building energy consumption.
In order to solve the technical problems, the application provides the following technical scheme: a short-term prediction method of building energy consumption comprises the steps of collecting historical data of building sub-term energy consumption, determining and collecting historical data of main influencing factors influencing the prediction of the building energy consumption sub-term, wherein the main influencing factors comprise historical data of building illumination energy consumption;
dividing the collected building illumination energy consumption historical data into input data and verification data, analyzing and determining to construct an illumination energy consumption prediction model based on a time sequence autoregressive model, taking the input data as input parameters of the prediction model, and verifying the short-term prediction of the illumination energy consumption and passing through the verification data;
comprehensively analyzing according to the building sub-term energy consumption historical data, establishing a building sub-term energy consumption database, carrying out normalization processing on the building sub-term energy consumption database and the collected main influencing factors of the building sub-term energy consumption, dividing the preprocessed data into training data and test data, and constructing an energy consumption prediction model based on a deep learning DBN network through training and testing by utilizing the training data and the test data;
and taking the illumination energy consumption predicted by the time sequence model and the main influence factors of actual monitoring as input parameters of the DBN model after training, and predicting the air conditioner energy consumption, the power energy consumption and the special energy consumption according to terms.
As a preferable scheme of the short-term prediction method for building energy consumption of the present application, wherein: the building sub-term energy consumption comprises air conditioner energy consumption, power energy consumption and special energy consumption, and the main influencing factors comprise seven characteristics including illumination energy consumption, outdoor average temperature, outdoor average humidity, weather characteristic values, holidays, average wind speed and 24 whole-point moments in the day, and are taken as main influencing factors for the building energy consumption sub-term prediction.
As a preferable scheme of the short-term prediction method for building energy consumption of the present application, wherein: the illumination energy consumption prediction model predicts illumination energy consumption based on a time sequence analysis method, and comprises definition of illumination energy consumption autocorrelation coefficients and partial autocorrelation coefficients;
autocorrelation coefficients: the 7-step difference value of the illumination energy consumption value of 24 hours on a certain day is given asdf i+k and dfi The degree of linear dependence between is defined as:
wherein r (k) =cov (df i ,df i+k ) As auto-covariance, var (df i ) The variance represents the k hysteresis number. Since the variance of the stationary time series is equal, the
Var(df i+k )=Var(df i )=L=Var(0) (2)
Let Var (0) =r (0) =σ, again, from the definition of variance 2 The recurrence of equation (1) is therefore:
therefore, p (k) can be calculated by the recurrence formula (3);
partial autocorrelation coefficients: given intermediate k-1 random differential variables [ df ] i ,df i+1 Λ,df i-k+1 ]Df is then i+k For df i-1 The correlation of (2) is defined as:
wherein ,
as a preferable scheme of the short-term prediction method for building energy consumption of the present application, wherein: the DBN network model includes a DBN network model of an unsupervised restrictive boltzmann machine RBM in combination with a supervised back propagation BP neural network to predict building subentry energy consumption, wherein:
the RBM energy function, also called expert product system, is defined as:
wherein n represents the number of visual units and m represents the number of implicit units;
under the energy function represented by equation (5), the probability that the hidden layer neuron hj is activated is:
since RBM is a bi-directional connection, visible layer neurons can be activated by hidden layer neurons as well, and the probability is:
wherein S (x) is a tanh activation function, the output is between [ -1,1], which is equivalent to adjusting the average value of the input to 0, so as to facilitate the subsequent processing, and the expression thereof
The neurons in the same layer have independence, so the probability density satisfies the independence, and the following formula is obtained:
training RBM in a layer-by-layer iterative mode to obtain values of learning parameters theta= { W, b and c }, wherein bi is bias of a visible node i, cj is bias of an hidden layer node j, wij is a connection matrix between the visible node i and the hidden layer node j, and given training data is fitted through the learning parameters.
Training the training data by adopting a random gradient rising method to maximize a log likelihood function, wherein the updating formula of the final weight is as follows:
VW ij =η(<v i h j > data -<v i h j > recon ) (10)
Vb i =η(<v i > data -<v i > recon ) (11)
Vc j =η(<h j > data -<h j > recon ) (12)
wherein, eta is the learning rate,<~> data for the distribution defined by the training samples,<~> recon a distribution defined for the samples after model reconstruction.
As a preferable scheme of the short-term prediction method for building energy consumption of the present application, wherein: the normalization process may include the steps of,
building energy consumption data (including meteorological data) are converted into [ -1,1] for normalization processing. The normalized formula is:
wherein x' is normalized data, x is sample original input data, xmin and xmax are minimum and maximum values of the original data.
As a preferable scheme of the short-term prediction method for building energy consumption of the present application, wherein: the deep belief network structure further includes the training steps of the restricted boltzmann machine:
inputting the normalized and preprocessed historical data into a first RBM to start unsupervised training, determining the weight and bias of the historical data, wherein the first RBM is used as a second RBM input to train the second RBM, and repeatedly training the three RBMs in sequence for a plurality of times to initialize model parameters; secondly, using a traditional supervised learning mode of BP neural network, back propagation propagates errors from top to bottom to each layer of RBM, and adjusts model parameters of all RBMs, so that the DBN can learn the intrinsic law of complex data and is used for establishing a DBN model; and finally, carrying out data prediction by using the trained network.
As a preferable scheme of the short-term prediction method for building energy consumption of the present application, wherein: the deep belief network further comprises the following training steps:
data acquisition and pretreatment: collecting influence factors related to building sub-term energy consumption, wherein the influence factors comprise artificial activity factors and non-artificial activity factors, the artificial activity factors comprise working and working time, holidays, illumination and the like, and the non-artificial activity factors comprise temperature, humidity, wind speed and the like;
comprehensively analyzing according to the building sub-term energy consumption historical data, establishing a building sub-term energy consumption database, carrying out normalization processing together with the collected influence factors of the building sub-term energy consumption, and finally dividing the preprocessed data into training data and test data;
constructing a DBN network prediction model, and optimally setting parameters of the DBN network prediction model, wherein the parameters comprise the number of layers of an RBM model hidden layer, the number of nodes of a BP network hidden layer and the like;
training a DBN model by using training data, training by adopting a contrast divergence algorithm, and calculating errors of actual output and target output;
and inputting the test data into the trained DBN model for testing to obtain the prediction results of air conditioner energy consumption, power energy consumption and special energy consumption.
As a preferable scheme of the short-term prediction method for building energy consumption of the present application, wherein: and the average absolute error (MAE) and the Root Mean Square Error (RMSE) are adopted as evaluation measurement values of the performance of the prediction model, the MAE better reflects the actual situation of the error of the prediction value, and the RMSE reflects the discrete degree of the error distribution. Smaller RMSE, for example, indicates smaller errors and better prediction. The expression is:
wherein n represents the number of samples, i represents the predicted sequence, y i The actual value of the i-th sequence is indicated,the predicted value of the i-th sequence is represented.
The application has the beneficial effects that: the short-term prediction method for building energy consumption provided by the application can be used for more accurately and effectively predicting each item of energy consumption in the building energy consumption.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. Wherein:
FIG. 1 is a topological diagram of a construction energy consumption sub-term prediction DBN network flow according to a construction energy consumption short-term prediction method according to a first embodiment of the application;
FIG. 2 is a high-level differential graph of illumination energy consumption of a short-term prediction method of building energy consumption according to a first embodiment of the present application;
FIG. 3 is a view showing an RBM network structure of a method for short-term prediction of building energy consumption according to a first embodiment of the present application;
FIG. 4 is a deep belief network architecture diagram of a short-term prediction method of building energy consumption according to a first embodiment of the present application;
FIG. 5 is an analysis chart of autocorrelation coefficients of a short-term prediction method of building energy consumption according to a second embodiment of the present application;
FIG. 6 is a partial autocorrelation coefficient analysis diagram of a short-term prediction method of building energy consumption according to a second embodiment of the present application;
FIG. 7 is a graph showing the result of illumination energy consumption prediction according to the short-term prediction method of building energy consumption according to the second embodiment of the present application;
FIG. 8 is a graph showing the air conditioner energy consumption prediction comparison of the short-term prediction method for building energy consumption according to the second embodiment of the present application;
FIG. 9 is a graph showing a comparison of power consumption prediction of a short-term prediction method of building energy consumption according to a second embodiment of the present application;
FIG. 10 is a graph showing a specific energy consumption prediction comparison of a short-term prediction method for building energy consumption according to a second embodiment of the present application;
FIG. 11 is a graph showing absolute error of air conditioner energy consumption in a short-term prediction method of building energy consumption according to a second embodiment of the present application;
FIG. 12 is a graph showing absolute error of power consumption of a short-term prediction method of building energy consumption according to a second embodiment of the present application;
fig. 13 is a graph showing absolute error of specific energy consumption of the short-term prediction method for building energy consumption according to the second embodiment of the present application.
Detailed Description
So that the manner in which the above recited objects, features and advantages of the present application can be understood in detail, a more particular description of the application, briefly summarized above, may be had by reference to the embodiments, some of which are illustrated in the appended drawings. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present application without making any inventive effort, shall fall within the scope of the present application.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application, but the present application may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present application is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the application. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Further, in describing the embodiments of the present application in detail, the cross-sectional view of the device structure is not partially enlarged to a general scale for convenience of description, and the schematic is only an example, which should not limit the scope of protection of the present application. In addition, the three-dimensional dimensions of length, width and depth should be included in actual fabrication.
Also in the description of the present application, it should be noted that the orientation or positional relationship indicated by the terms "upper, lower, inner and outer", etc. are based on the orientation or positional relationship shown in the drawings, are merely for convenience of describing the present application and simplifying the description, and do not indicate or imply that the apparatus or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present application. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
The terms "mounted, connected, and coupled" should be construed broadly in this disclosure unless otherwise specifically indicated and defined, such as: can be fixed connection, detachable connection or integral connection; it may also be a mechanical connection, an electrical connection, or a direct connection, or may be indirectly connected through an intermediate medium, or may be a communication between two elements. The specific meaning of the above terms in the present application will be understood in specific cases by those of ordinary skill in the art.
Example 1
In this embodiment, aiming at the problems that the existing total energy consumption prediction method cannot accurately distinguish the consumption direction of the building energy consumption and the prediction accuracy is low, the total energy consumption is divided into four major terms according to the energy consumption application, and a new method for short-term prediction of the building energy consumption is provided. Firstly, constructing an illumination energy consumption prediction method based on a time sequence Autoregressive (AR) model, and performing short-term prediction on illumination energy consumption of a building; and secondly, constructing an energy consumption prediction model based on a deep learning DBN network, predicting air conditioner energy consumption, power energy consumption (an elevator, pressurized water supply and the like) and special energy consumption (a computer room, a kitchen and the like) of the building according to terms, and comparing the energy consumption prediction model with energy consumption prediction results of other network models. The result shows that the prediction method provided by the application can more accurately and effectively predict the energy consumption of each sub-term in the building energy consumption. Specifically, referring to fig. 1, the method includes the following steps:
collecting building sub-term energy consumption historical data, and determining and collecting main influence factor historical data affecting building energy consumption sub-term prediction, wherein the main influence factors comprise building illumination energy consumption historical data;
dividing the collected building illumination energy consumption historical data into input data and verification data, analyzing and determining to construct an illumination energy consumption prediction model based on a time sequence autoregressive model, taking the input data as input parameters of the prediction model, and verifying the short-term prediction of the illumination energy consumption and passing through the verification data;
comprehensively analyzing according to the building sub-term energy consumption historical data, establishing a building sub-term energy consumption database, carrying out normalization processing on the building sub-term energy consumption database and the collected main influencing factors of the building sub-term energy consumption, dividing the preprocessed data into training data and test data, and constructing an energy consumption prediction model based on a deep learning DBN network through training and testing by utilizing the training data and the test data;
and taking the illumination energy consumption predicted by the time sequence model and the main influence factors of actual monitoring as input parameters of the DBN model after training, and predicting the air conditioner energy consumption, the power energy consumption and the special energy consumption according to terms.
What needs to be explained here is: the illumination energy consumption reflects the office number in the building, and the office number directly affects the air conditioner energy consumption, the power energy consumption (elevator use and secondary water supply use) and the special energy consumption (such as machine room electricity and kitchen electricity). The outdoor average temperature is closely related to the energy consumption of the air conditioner, 24 whole-point moments are closely related to the number of people in office (working hours and working hours), meanwhile, weather characteristic values and holidays can influence the energy consumption of the air conditioner, so that 7 characteristics of illumination energy consumption, outdoor average temperature, outdoor average humidity, weather characteristic values, holidays, average wind speed and 24 whole-point moments in one day are selected as main influence factors for the energy consumption item prediction of the building in summary theoretical analysis.
Because the illumination energy consumption is closely related to the office number in the building and reflects the traffic, but the office number cannot be counted and is only reflected by the illumination energy consumption, the illumination energy consumption is predicted by using a time series autoregressive model. The illumination energy consumption is therefore a factor of influence of the input parameters which is predicted first. The data to be collected in the present application is building energy consumption including air conditioning energy consumption, power energy consumption and special energy consumption, and the main influencing factors include seven characteristics of illumination energy consumption, outdoor average temperature, outdoor average humidity, weather characteristic values, holidays, average wind speed and 24 whole-point moments in the day, which are taken as main influencing factors for building energy consumption sub-term prediction.
Further, in this embodiment, the illumination energy consumption always appears as a random fluctuation based on the past energy consumption in a short time, belonging to a non-stationary time series, but the high-order difference of the illumination energy consumption is stationary. For example, the illumination energy consumption value of 24 hours of a certain day is [ x ] 1 ,x 2 Λ,x 24 ]=[49.8,51.6,Λ,218.2](unit: kwh), where x i Referring to fig. 2, the 7-step difference of the illumination energy consumption is shown to be stable, and then the illumination energy consumption can be predicted by using a time series analysis method:
the application uses a P-order Autoregressive (AR) model to construct a prediction model of illumination energy consumption, and definition of an autocorrelation coefficient and a partial autocorrelation coefficient of the illumination energy consumption is given below so as to determine the value of the order P.
Definition 1: autocorrelation of illumination energy consumption, given an illumination energy consumption value of 24 hours on a certain day7-order difference value of [ df ] 1 ,df 2 ,L,df 24 ],df i+k and dfi The degree of linear dependence between is defined as:
wherein r (k) =cov (df i ,df i+k ) As auto-covariance, var (df i ) The variance represents the k hysteresis number. Since the variance of the stationary time series is equal, the
Var(df i+k )=Var(df i )=L=Var(0) (2)
Let Var (0) =r (0) =σ, again, from the definition of variance 2 The recurrence of equation (1) is therefore:
therefore, p (k) can be calculated by the recurrence formula (3).
Definition 2: partial autocorrelation coefficients of illumination energy consumption, given intermediate k-1 random differential variables [ df ] i ,df i+1 Λ,df i-k+1 ]Df is then i+k For df i-1 The correlation of (2) is defined as:
wherein ,
the above is to determine the illumination energy consumption of the building predicted by using a time series autoregressive model, construct a prediction model of the illumination energy consumption by using a P-order Autoregressive (AR) model, take the collected input data as the input parameters of the prediction model, and verify the short-term prediction of the illumination energy consumption and the verification of the result by verification data.
Based on the collected historical energy consumption data and the predicted illumination energy consumption value, an energy consumption prediction model based on a deep learning DBN network is constructed, and training, testing and prediction are performed by utilizing the data, and specifically the method comprises the following steps:
step one: deep belief network selection
In this embodiment, a DBN network model consisting of 3 unsupervised restricted boltzmann machines RBM in combination with 1 supervised back propagation BP neural network is used to predict building subentry energy consumption.
As shown in fig. 3, the RBM serves as a neural sensor. The full-bidirectional connection is realized by taking the hidden layer h and the visible layer v as connection weights through the connection matrix w.
The RBM energy function, also called expert product system, can be defined as:
where n represents the number of visual elements and m represents the number of implicit elements.
Under the energy function represented by equation (5), the probability that the hidden layer neuron hj is activated is:
since RBM is a bi-directional connection, visible layer neurons can be activated by hidden layer neurons as well, and the probability is:
wherein S (x) is a tanh activation function, and the output is [ -1,1]The average value of the input is adjusted to 0, so that the subsequent processing is convenient, and the expression is that
The neurons in the same layer have independence, so the probability density satisfies the independence, and the following formula is obtained:
training RBM in a layer-by-layer iterative mode to obtain values of learning parameters theta= { W, b, c }, wherein bi is bias of a visible node i, cj is bias of an hidden layer node j, and Wij is a connection matrix between the visible node i and the hidden layer node j. The given training data is fitted by learning parameters.
Training the training data by adopting a random gradient rising method to maximize a log likelihood function, wherein the updating formula of the final weight is as follows:
VW ij =η(<v i h j > data -<v i h j > recon ) (10)
Vb i =η(<v i > data -<v i > recon ) (11)
Vc j =η(<h j > data -<h j > recon ) (12)
wherein, eta is the learning rate,<~> data for the distribution defined by the training samples,<~> recon a distribution defined for the samples after model reconstruction.
Step two: preprocessing, i.e. normalizing, of data
In order to ensure that the data are in a unified scale and prevent modeling pathological problems caused by different orders of magnitude, the application converts all building energy consumption data (including meteorological data) into the range of-1, performs normalization processing, and the preprocessed original data comprise 7 influencing factors and annual historical energy consumption values of air conditioner, power and special energy consumption. The normalized formula is:
wherein x' is normalized data, x is sample original input data, xmin and xmax are minimum and maximum values of the original data.
Step two: deep confidence network structure
The training of RBM (Boltzmann machine limited), specifically, as shown in figure 4, firstly inputting the preprocessed original data into a first RBM to start unsupervised training, determining the weight and bias of the first RBM, taking the output of the first RBM as the input of a second RBM, training the second RBM, repeating training for 3 RBMs in turn, and repeating training for a plurality of times to initialize model parameters; secondly, using a traditional supervised learning mode of BP neural network, back propagation propagates errors from top to bottom to each layer of RBM, and adjusts model parameters of all RBMs, so that the DBN can learn the intrinsic law of complex data and is used for establishing a DBN model; and finally, carrying out data prediction by using the trained network. It is not difficult to see that training a DBN includes two 2 processes, i.e., unsupervised pre-training and supervised fine tuning, automatically extracts the change features between the input data based on its strong nonlinear mapping capability, and inputs the test data to the DBN model to obtain its predicted values.
Step three: deep confidence network training method
This step is training of the DBN model, and referring again to fig. 1, the deep belief network training steps are as follows:
data acquisition and pretreatment: collecting influence factors related to building sub-term energy consumption, wherein the influence factors comprise artificial activity factors and non-artificial activity factors, the artificial activity factors comprise working and working time, holidays, illumination and the like, and the non-artificial activity factors comprise temperature, humidity, wind speed and the like;
comprehensively analyzing according to the building sub-term energy consumption historical data, establishing a building sub-term energy consumption database, carrying out normalization processing together with the collected influence factors of the building sub-term energy consumption, and finally dividing the preprocessed data into training data and test data;
constructing a DBN network prediction model, and optimally setting parameters of the DBN network prediction model, wherein the parameters comprise the number of layers of an RBM model hidden layer, the number of BP network hidden layer nodes and the like;
training a DBN model by using training data, training by adopting a contrast divergence algorithm for accelerating the training process, and calculating errors of actual output and target output;
and inputting the test data into the trained DBN model for testing to obtain the prediction results of air conditioner energy consumption, power energy consumption and special energy consumption.
Step four: evaluation of model Performance metric
An average absolute error (MAE) and a Root Mean Square Error (RMSE) are adopted as evaluation metric values of model performance, the MAE better reflects the actual situation of the predicted value error, and the RMSE reflects the discrete degree of error distribution. Smaller RMSE, for example, indicates smaller errors and better prediction. The expression is:
wherein n represents the number of samples, i represents the predicted sequence, y i The actual value of the i-th sequence is indicated,the predicted value of the i-th sequence is represented.
Example 2
In the present embodiment, a method for predicting illumination energy consumption based on a time-series Autoregressive (AR) model is constructed based on the first embodiment, and short-term prediction is performed on illumination energy consumption of a building; secondly, an energy consumption prediction model based on a deep learning DBN network is constructed, air conditioning energy consumption, power energy consumption (an elevator, pressurized water supply and the like) and special energy consumption (a computer room, a kitchen and the like) of the building are predicted in terms, and the energy consumption prediction results of the constructed model and other network models are compared. The result shows that the prediction method provided by the application can more accurately and effectively predict the energy consumption of each sub-term in the building energy consumption. Specifically, the method comprises the following steps:
step one: short-term prediction of illumination energy consumption
The method comprises the steps of collecting the hourly illumination energy consumption values of a certain building from 8 months, 7 days, 8 months, 6 days, and one whole year, predicting by using a time sequence AR model, and selecting the illumination energy consumption values of 7 times, 31 days, 8 months, 6 days, 7 multiplied by 24=168 whole-point moments of 2017 to 8 months to verify the model.
As can be seen from fig. 5, the decay of the value of the autocorrelation coefficient to a small value fluctuation is relatively slow and continuous, with a long tail being dragged, conforming to the tailing property; as can be seen from fig. 6, the value of the partial autocorrelation coefficient is damped, and suddenly converges to the threshold value 0 when the hysteresis order is 6, and the partial autocorrelation coefficient values are all 0 after the 6 th order, which accords with the tail-biting property, so that the order p=6 of the autoregressive model AR (P), and it is further verified that the illumination energy consumption prediction accords with the autoregressive model AR (6).
As shown in fig. 7, the predicted value and actual value curves of the illumination energy consumption at the whole point time from 31 days of 7 months of 2017 to 6 days of 8 months are higher in fitting degree, the average absolute error mae= 9.9591 and the root mean square error rmse= 13.5182, and the expected prediction effect is achieved.
Step two: short-term prediction of air conditioning, power and specific energy consumption
The experimental sample data is derived from the subitem monitoring data of air conditioning energy consumption, power energy consumption and special energy consumption of a building in 2016, 8, 7, and 2017, 8, 6, and one year each hour. Wherein 7296 sets of data of 10 months/hour from 8 th month 7 th year to 6 th month 6 th year 2017 are used as training data, 1296 sets of data from 7 th month 6 th month 7 th year 2017 th month 7 th month 30 th year 2017 are used as test data.
And taking the illumination energy consumption predicted by the time sequence model and the actually monitored outdoor average temperature, outdoor average humidity, weather characteristic values, holidays, average wind speed and 24 whole-point moments in a day as input parameters, and establishing a DBN model to predict air conditioning energy consumption, power energy consumption and special energy consumption from 7 months, 31 days and 8 months and 6 days in 2017.
For the DBN network model, the higher the hidden layer and the number of hidden layer neurons, the better the performance of the algorithm, but the computational complexity also increases. Therefore, under the condition of comprehensively considering algorithm performance and calculated amount, the RBM layer number is set to be 3 layers, the number of neurons of each hidden layer is 400, 200 and 100, the iteration number is 200, and the learning rate is 0.001; when back propagation is fine-tuned, the learning rate is changed to 0.1, and the iteration number is 500.
Step three: comparison of prediction results
Based on the same historical data, the data predicted by the building energy consumption prediction model provided by the application are compared with the existing 2 building energy consumption prediction models: data based on the BP neural network and predicted based on the iPSO-BP neural network are compared. The 3 models are compared by selecting a predicted value and an actual value of the energy consumption for one week, which are shown in fig. 8, 9 and 10, for example.
As can be seen from fig. 8, 9 and 10, the DBN model predicted curve fits better than the iPSO-BP model and the BP model predicted curve to the actual curve. To further verify the predicted results, 12 sets of data were randomly selected from fig. 8, 9 and 10, and the three model predicted values were compared with the actual values, respectively, as shown in table 1. The prediction precision of the DBN model of the building energy consumption fractional prediction is higher than that of the iPSO-BP model and the BP model.
TABLE 1 comparison of predicted values and actual values of energy consumption by DBN model, iPSO-BP model and BP model
Step four: air conditioner, power and special energy consumption prediction result analysis
Absolute error pairs for the predictions of air conditioning energy consumption, power energy consumption and specific energy consumption for the DBN model, the iPSO-BP model and the BP neural network model are shown in fig. 11, fig. 12 and fig. 13.
It can be seen from fig. 11, 12 and 13 that the DBN model, the iPSO-BP model and the BP model all fluctuate around 0 value, but the DBN model has smaller fluctuation of the prediction error than the iPSO-BP model and the BP model, and is relatively stable. Therefore, the accuracy and the stability of the DBN model in the training and learning process are higher.
According to the performance indexes of the formula (14) and the formula (15), the test data are respectively analyzed and compared to obtain three energy consumption error comparison results of 3 models, and the three energy consumption error comparison results are shown in the table 2. Average absolute errors of the BP model for air conditioner, power and special energy consumption prediction are 36.2153, 9.0127 and 6.6296 respectively, average absolute errors of the iPSO-BP model for air conditioner, power and special energy consumption prediction are 10.0822, 3.3848 and 2.7347 respectively, and average absolute errors of the DBN model for air conditioner, power and special energy consumption prediction are 5.4707, 1.7916 and 1.6075 respectively; the root mean square errors of the BP model for air conditioner, power and special energy consumption prediction are 42.7680, 12.3035 and 8.3477 respectively, the root mean square errors of the iPSO-BP model for air conditioner, power and special energy consumption prediction are 14.2698, 5.1325 and 2.9536 respectively, and the average relative errors of the DBN model for air conditioner, power and special energy consumption prediction are 8.6575, 2.9536 and 2.2178 respectively. It can be seen that the average absolute error and the root mean square error of the DBN model energy consumption prediction are smaller than those of the iPSO-BP model, and are much smaller than those of the BP model, so that the DBN prediction model provided by the method has a better effect on the building energy consumption item prediction, and the accuracy of the building energy consumption prediction is improved.
TABLE 2 comparison of Performance indicators for energy consumption predictions for DBN model, iPSO-BP model and BP model
It should be noted that the above embodiments are only for illustrating the technical solution of the present application and not for limiting the same, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present application may be modified or substituted without departing from the spirit and scope of the technical solution of the present application, which is intended to be covered in the scope of the claims of the present application.
Claims (7)
1. A short-term prediction method for building energy consumption is characterized by comprising the following steps: comprising the steps of (a) a step of,
collecting building sub-term energy consumption historical data, and determining and collecting main influence factor historical data affecting building energy consumption sub-term prediction, wherein the main influence factors comprise building illumination energy consumption historical data;
dividing the collected building illumination energy consumption historical data into input data and verification data, analyzing and determining to construct an illumination energy consumption prediction model based on a time sequence autoregressive model, taking the input data as input parameters of the prediction model, and verifying the short-term prediction of the illumination energy consumption and passing through the verification data;
comprehensively analyzing according to the building sub-term energy consumption historical data, establishing a building sub-term energy consumption database, carrying out normalization processing on the building sub-term energy consumption database and the collected main influencing factors of the building sub-term energy consumption, dividing the preprocessed data into training data and test data, and constructing an energy consumption prediction model based on a deep learning DBN network through training and testing by utilizing the training data and the test data;
taking the illumination energy consumption predicted by the time sequence model and the main influence factors of actual monitoring as input parameters of the DBN model after training, and predicting the air conditioner energy consumption, the power energy consumption and the special energy consumption according to terms;
the DBN network model includes a DBN network model of an unsupervised restrictive boltzmann machine RBM in combination with a supervised back propagation BP neural network to predict building subentry energy consumption, wherein:
the RBM energy function, also called expert product system, is defined as:
wherein n represents the number of visual units and m represents the number of implicit units;
hidden layer neuron h under the energy function represented by equation (5) j The probability of being activated is:
since RBM is a bi-directional connection, visible layer neurons can be activated by hidden layer neurons as well, and the probability is:
wherein S (x) is a tanh activation function, the output is between [ -1,1], which is equivalent to adjusting the average value of the input to 0, so as to facilitate the subsequent processing, and the expression thereof
The neurons in the same layer have independence, so the probability density satisfies the independence, and the following formula is obtained:
training RBM in a layer-by-layer iterative manner to obtain values of learning parameters theta= { W, b, c }, wherein b i For biasing of visible node i, c j To bias hidden layer node j, W ij Fitting given training data for a connection matrix between the visible node i and the hidden layer node j through learning parameters;
training the training data by adopting a random gradient rising method to maximize a log likelihood function, wherein the updating formula of the final weight is as follows:
VW ij =η(<v i h j > data -<v i h j > recon ) (10)
Vb i =η(<v i > data -<v i > recon ) (11)
Vc j =η(<h j > data -<h j > recon ) (12)
wherein, eta is the learning rate,<~> data for the distribution defined by the training samples,<~> recon a distribution defined for the samples after model reconstruction.
2. The short-term prediction method of building energy consumption according to claim 1, wherein: the building sub-term energy consumption comprises air conditioner energy consumption, power energy consumption and special energy consumption, and the main influencing factors comprise seven characteristics including illumination energy consumption, outdoor average temperature, outdoor average humidity, weather characteristic values, holidays, average wind speed and 24 whole-point moments in the day, and are taken as main influencing factors for the building energy consumption sub-term prediction.
3. The short-term prediction method of building energy consumption according to claim 1, wherein: the illumination energy consumption prediction model predicts illumination energy consumption based on a time sequence analysis method, and comprises definition of illumination energy consumption autocorrelation coefficients and partial autocorrelation coefficients;
autocorrelation coefficients: the 7-order difference value given the illumination energy consumption value of 24 hours on a certain day is [ df 1 ,df 2 ,L,df 24 ],df i+k and dfi The degree of linear dependence between is defined as:
wherein r (k) =cov (df i ,df i+k) As auto-covariance, var (df i ) As variance, k represents the hysteresis number; since the variance of the stationary time series is equal, the
Var(df i+k )=Var(df i )=L=Var(0) (2)
Let Var (0) =r (0) =σ, again, from the definition of variance 2 The recurrence of equation (1) is therefore:
therefore, p (k) can be calculated by the recurrence formula (3);
partial autocorrelation coefficients: given intermediate k-1 random differential variables [ df ] i ,df i+1 Λ,df i-k+1 ]Df is then i+k For df i-1 The correlation of (2) is defined as:
wherein ,the bias autocorrelation coefficients, D (k) the variance of the hysteresis order k, D the total variance, ρk the autocorrelation coefficients of the hysteresis order k.
4. The short-term prediction method of building energy consumption according to claim 1, wherein: the normalization process may include the steps of,
building energy consumption data are converted into [ -1,1] and normalized; the normalized formula is:
wherein x' is normalized data, x is sample original input data, x min and xmax Is the minimum and maximum of the original data.
5. The short-term prediction method of building energy consumption according to claim 1, wherein: the deep belief network structure further includes the training steps of the restricted boltzmann machine:
inputting the normalized and preprocessed historical data into a first RBM to start unsupervised training, determining the weight and bias of the historical data, wherein the first RBM is used as a second RBM input to train the second RBM, and repeatedly training the three RBMs in sequence for a plurality of times to initialize model parameters; secondly, using a traditional supervised learning mode of BP neural network, back propagation propagates errors from top to bottom to each layer of RBM, and adjusts model parameters of all RBMs, so that the DBN can learn the intrinsic law of complex data and is used for establishing a DBN model; and finally, carrying out data prediction by using the trained network.
6. The short-term prediction method for building energy consumption according to claim 5, wherein: the deep belief network further comprises the training steps of the DBN network:
data acquisition and pretreatment: collecting influence factors related to building sub-term energy consumption, wherein the influence factors comprise artificial activity factors and non-artificial activity factors, the artificial activity factors comprise working and working time, holidays, illumination and the like, and the non-artificial activity factors comprise temperature, humidity, wind speed and the like;
comprehensively analyzing according to the building sub-term energy consumption historical data, establishing a building sub-term energy consumption database, carrying out normalization processing together with the collected influence factors of the building sub-term energy consumption, and finally dividing the preprocessed data into training data and test data;
constructing a DBN network prediction model, and optimally setting parameters of the DBN network prediction model, wherein the parameters comprise the number of layers of an RBM model hidden layer, the number of nodes of a BP network hidden layer and the like;
training a DBN model by using training data, training by adopting a contrast divergence algorithm, and calculating errors of actual output and target output;
and inputting the test data into the trained DBN model for testing to obtain the prediction results of air conditioner energy consumption, power energy consumption and special energy consumption.
7. The short-term prediction method of building energy consumption according to claim 1, wherein: the method also comprises the step of adopting an average absolute error MAE and a root mean square error RMSE as evaluation metric values of the prediction model performance, wherein the MAE better reflects the actual situation of the prediction value error, the RMSE reflects the discrete degree of error distribution, the smaller the RMSE is, the smaller the error is, the better the prediction effect is, and the expression is:
where n represents the number of samples, i represents the predicted sequence, yi represents the actual value of the ith sequence,the predicted value of the i-th sequence is represented.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810191237.0A CN108320016B (en) | 2018-03-08 | 2018-03-08 | Short-term prediction method for building energy consumption |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810191237.0A CN108320016B (en) | 2018-03-08 | 2018-03-08 | Short-term prediction method for building energy consumption |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108320016A CN108320016A (en) | 2018-07-24 |
CN108320016B true CN108320016B (en) | 2023-09-19 |
Family
ID=62900506
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810191237.0A Active CN108320016B (en) | 2018-03-08 | 2018-03-08 | Short-term prediction method for building energy consumption |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108320016B (en) |
Families Citing this family (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109255472A (en) * | 2018-08-20 | 2019-01-22 | 天津科技大学 | A kind of Office Building Energy Consumption prediction technique based on Multivariate adaptive regression splines batten |
CN109214572A (en) * | 2018-09-06 | 2019-01-15 | 南京工业职业技术学院 | One kind is based on EPSO-BP to commercial building lamp decoration energy consumption prediction technique |
CN109118020A (en) * | 2018-09-06 | 2019-01-01 | 南京工业职业技术学院 | A kind of subway station energy consumption short term prediction method and its forecasting system |
CN109376939B (en) * | 2018-11-01 | 2021-08-10 | 三峡大学 | Power grid stability real-time prediction method based on adaptive neural network |
CN109582912B (en) * | 2018-11-19 | 2021-02-02 | 北京理工大学 | Public building power consumption monitoring data parameter estimation method |
CN111199305A (en) * | 2018-11-20 | 2020-05-26 | 上海烟草集团有限责任公司 | Neural network-based production energy consumption prediction method and system, electronic terminal and storage medium |
CN109695944B (en) * | 2018-11-29 | 2021-06-18 | 中国汽车工业工程有限公司 | Control method for coating fresh air conditioner based on multi-model deep learning |
CN110009156A (en) * | 2019-04-08 | 2019-07-12 | 北京博锐尚格节能技术股份有限公司 | Building energy consumption budget management method and storage medium |
CN109934422B (en) * | 2019-04-25 | 2021-06-29 | 中国人民解放军国防科技大学 | Neural network wind speed prediction method based on time series data analysis |
CN110309560B (en) * | 2019-06-13 | 2020-12-22 | 清华大学 | Multi-level migration simulation method and device |
CN110435009B (en) * | 2019-07-15 | 2020-11-10 | 中国科学院重庆绿色智能技术研究院 | Intelligent design method for concrete production mix proportion |
CN110400018B (en) * | 2019-07-29 | 2023-06-27 | 上海电力大学 | Operation control method, system and device for coal-fired power plant pulverizing system |
CN110533243A (en) * | 2019-08-27 | 2019-12-03 | 西安建筑科技大学 | A kind of building energy consumption prediction technique based on fractal theory |
CN110942228A (en) * | 2019-10-25 | 2020-03-31 | 万翼科技有限公司 | Energy consumption control method and device, computer equipment and storage medium |
CN111753470B (en) * | 2020-06-30 | 2022-09-16 | 桂林电子科技大学 | Building energy consumption prediction method based on rough set and deep belief neural network |
CN112712213B (en) * | 2021-01-15 | 2023-07-04 | 上海交通大学 | Method and system for predicting deep migration learning energy consumption of concentrated air conditioning house |
CN113344192B (en) * | 2021-05-31 | 2022-01-11 | 中国标准化研究院 | Enterprise-level motor system energy-saving optimization automatic control method and system |
CN117077854B (en) * | 2023-08-15 | 2024-04-16 | 广州视声智能科技有限公司 | Building energy consumption monitoring method and system based on sensor network |
CN117371624B (en) * | 2023-12-07 | 2024-02-20 | 邢台职业技术学院 | Building energy-saving transformation optimization management method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106874581A (en) * | 2016-12-30 | 2017-06-20 | 浙江大学 | A kind of energy consumption of air conditioning system in buildings Forecasting Methodology based on BP neural network model |
CN106991504A (en) * | 2017-05-09 | 2017-07-28 | 南京工业大学 | Building energy consumption Forecasting Methodology, system and building based on metering separate time series |
-
2018
- 2018-03-08 CN CN201810191237.0A patent/CN108320016B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106874581A (en) * | 2016-12-30 | 2017-06-20 | 浙江大学 | A kind of energy consumption of air conditioning system in buildings Forecasting Methodology based on BP neural network model |
CN106991504A (en) * | 2017-05-09 | 2017-07-28 | 南京工业大学 | Building energy consumption Forecasting Methodology, system and building based on metering separate time series |
Non-Patent Citations (1)
Title |
---|
基于深度结构多任务学习的园区型综合能源系统多元负荷预测;史佳琪等;《电网技术》;20180305(第3期);第698-706页 * |
Also Published As
Publication number | Publication date |
---|---|
CN108320016A (en) | 2018-07-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108320016B (en) | Short-term prediction method for building energy consumption | |
Tian | Short-term wind speed prediction based on LMD and improved FA optimized combined kernel function LSSVM | |
CN112949945B (en) | Wind power ultra-short-term prediction method for improving bidirectional long-term and short-term memory network | |
CN109214566A (en) | Short-term wind power prediction method based on shot and long term memory network | |
Ding et al. | Point and interval forecasting for wind speed based on linear component extraction | |
CN106951611A (en) | A kind of severe cold area energy-saving design in construction optimization method based on user's behavior | |
CN113283588B (en) | Near-shore single-point wave height forecasting method based on deep learning | |
CN112633604B (en) | Short-term power consumption prediction method based on I-LSTM | |
CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
CN114676822B (en) | Multi-attribute fusion air quality forecasting method based on deep learning | |
Liu et al. | Heating load forecasting for combined heat and power plants via strand-based LSTM | |
CN106778838A (en) | A kind of method for predicting air quality | |
CN114580545A (en) | Wind turbine generator gearbox fault early warning method based on fusion model | |
Suryo et al. | Improved time series prediction using LSTM neural network for smart agriculture application | |
CN112396152A (en) | Flood forecasting method based on CS-LSTM | |
CN115115125A (en) | Photovoltaic power interval probability prediction method based on deep learning fusion model | |
CN108647817B (en) | Energy consumption load prediction method and system | |
CN110222910A (en) | A kind of active power distribution network Tendency Prediction method and forecasting system | |
CN117096860A (en) | Overhead transmission line current-carrying capacity interval prediction method and equipment based on LSSVM model | |
CN115877483A (en) | Typhoon path forecasting method based on random forest and GRU | |
CN115759343A (en) | E-LSTM-based user electric quantity prediction method and device | |
CN115496264A (en) | Method for predicting generated power of wind turbine generator | |
CN111435471A (en) | Heat supply gas consumption prediction model based on L STM | |
CN114897204A (en) | Method and device for predicting short-term wind speed of offshore wind farm | |
CN115293406A (en) | Photovoltaic power generation power prediction method based on Catboost and Radam-LSTM |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |