CN117574310A - Building central air conditioner load prediction method and system based on multitask deep learning - Google Patents
Building central air conditioner load prediction method and system based on multitask deep learning Download PDFInfo
- Publication number
- CN117574310A CN117574310A CN202311363269.1A CN202311363269A CN117574310A CN 117574310 A CN117574310 A CN 117574310A CN 202311363269 A CN202311363269 A CN 202311363269A CN 117574310 A CN117574310 A CN 117574310A
- Authority
- CN
- China
- Prior art keywords
- load
- data
- central air
- air conditioner
- air conditioning
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000013135 deep learning Methods 0.000 title claims abstract description 27
- 238000004378 air conditioning Methods 0.000 claims abstract description 52
- 238000000605 extraction Methods 0.000 claims abstract description 37
- 230000004927 fusion Effects 0.000 claims abstract description 10
- 230000008569 process Effects 0.000 claims abstract description 10
- 238000010276 construction Methods 0.000 claims abstract description 6
- 239000013598 vector Substances 0.000 claims description 20
- 238000004590 computer program Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 9
- 238000003860 storage Methods 0.000 claims description 9
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 230000003190 augmentative effect Effects 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000009191 jumping Effects 0.000 claims description 2
- 238000012549 training Methods 0.000 claims description 2
- 238000001914 filtration Methods 0.000 claims 1
- 238000007781 pre-processing Methods 0.000 claims 1
- 230000009466 transformation Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/30—Control or safety arrangements for purposes related to the operation of the system, e.g. for safety or monitoring
- F24F11/46—Improving electric energy efficiency or saving
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
- F24F11/63—Electronic processing
- F24F11/64—Electronic processing using pre-stored data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2131—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on a transform domain processing, e.g. wavelet transform
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/096—Transfer learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- Signal Processing (AREA)
- Fuzzy Systems (AREA)
- Air Conditioning Control Device (AREA)
Abstract
The invention belongs to the technical field of load prediction, and provides a method and a system for predicting the load of a central air conditioner of a building based on multi-task deep learning, wherein the method and the system are combined with large building data provided with an air conditioning system and a large building central air conditioner load prediction model to obtain a load prediction value of the air conditioning system; the construction process of the large building central air conditioner load prediction model comprises the following steps: the method comprises the steps of performing feature extraction by adopting a corresponding network structure aiming at different types of data, extracting load short-period features by adopting a first extraction network model, extracting load long-period features by adopting a second extraction network, extracting features of other factors influencing the load by adopting a third extraction network, performing feature fusion on all features obtained by network extraction of different structures, extracting advanced features, and performing prediction based on the advanced features to obtain load predicted values of all central air conditioning systems at the next moment. The problem of mutual influence among a plurality of systems is solved, and simultaneously, the load of all the systems is predicted.
Description
Technical Field
The invention belongs to the technical field of load prediction, and particularly relates to a method and a system for predicting the load of a building central air conditioner based on multi-task deep learning.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the development of society, the central air conditioning system is increasingly widely used in buildings. Meanwhile, the central air conditioner is also a main energy consumption device, so that effective energy management measures are needed, and the energy utilization efficiency is improved under the condition of ensuring the balance of system load and energy supply. This requires the future load information of the central air conditioner, i.e. the load of the central air conditioner to be predicted. With the development of intelligent sensor technology and computer technology, a large amount of data of a central air conditioning system is accumulated, so that data-based central air conditioning load prediction is possible.
The central air conditioner load prediction method based on the data comprises a method based on statistical analysis and a method based on machine learning. The former has relatively poor prediction accuracy for nonlinear systems. The neural network model is one of machine learning algorithms, and becomes a popular load prediction method due to the excellent performance of the neural network model in dealing with nonlinear problems. The existing prediction method based on the neural network mainly solves the problem of single system prediction. In some buildings with multiple sets of central air conditioning systems, the systems can be mutually influenced, and only a single task learning frame is adopted, so that some related characteristics among the systems can be lost, and the prediction accuracy is reduced;
the air conditioning load is affected by a number of factors. Human activity is a major factor affecting the air conditioning load. Human activities can be largely divided into short-period (in days) and long-period (in weeks) activities, so that the air conditioning load also has a rule of a change in the long-and-short periods. In addition, various external uncertainty factors can interfere with the air conditioning load, reducing the prediction accuracy of the model.
Disclosure of Invention
In order to solve at least one technical problem in the background art, the invention provides a method and a system for predicting the load of a building central air conditioner based on multi-task deep learning, which are used for predicting the load of the building central air conditioner by a method for simultaneously solving a plurality of tasks so as to solve the problem of mutual influence among a plurality of systems and simultaneously predicting the loads of all the systems.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the first aspect of the invention provides a building central air conditioner load prediction method based on multi-task deep learning, comprising the following steps:
acquiring large building data provided with an air conditioning system;
combining the large building data with the air conditioning system and a large building central air conditioning load prediction model to obtain a load prediction value of the air conditioning system; the construction process of the large building central air conditioner load prediction model comprises the following steps:
the method comprises the steps of performing feature extraction by adopting a corresponding network structure aiming at different types of data, extracting load short-period features by adopting a first extraction network model, extracting load long-period features by adopting a second extraction network, extracting features of other factors influencing the load by adopting a third extraction network, performing feature fusion on all features obtained by network extraction of different structures, extracting advanced features, and performing prediction based on the advanced features to obtain load predicted values of all central air conditioning systems at the next moment.
A second aspect of the present invention provides a building central air conditioning load prediction system based on multitasking deep learning, comprising:
a data acquisition module for acquiring large building data equipped with an air conditioning system;
the load prediction module is used for combining the large building data with the air conditioning system and a large building central air conditioning load prediction model to obtain a load prediction value of the air conditioning system; the construction process of the large building central air conditioner load prediction model comprises the following steps:
the method comprises the steps of performing feature extraction by adopting a corresponding network structure aiming at different types of data, extracting load short-period features by adopting a first extraction network model, extracting load long-period features by adopting a second extraction network, extracting features of other factors influencing the load by adopting a third extraction network, performing feature fusion on all features obtained by network extraction of different structures, extracting advanced features, and performing prediction based on the advanced features to obtain load predicted values of all central air conditioning systems at the next moment.
A third aspect of the present invention provides a computer-readable storage medium.
A computer readable storage medium having stored thereon a computer program which when executed by a processor implements the steps in the method for predicting load of a building central air conditioner based on multitasking deep learning as described in the first aspect.
A fourth aspect of the invention provides a computer device.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in the method for predicting building central air conditioning load based on multitasking deep learning as described in the first aspect when the program is executed.
Compared with the prior art, the invention has the beneficial effects that:
1. aiming at the prediction problem of a single system, the invention adopts a method of multiplexing deep learning (MTL) to fuse multiple aspects of information and simultaneously solve multiple tasks, which can process different types of data by using different types of neural network models and fuse the characteristics of each data so as to ensure that the prediction is more accurate.
2. Aiming at the condition that the air conditioner load is influenced by various factors, the invention uses an LSTM model to extract the load short period characteristic and uses ConvGRU to extract the load long period characteristic; aiming at other factors influencing the load, such as indoor and outdoor temperature, building area and the like, the BPNN is used for extracting the characteristics, the network with different structures is used for fully extracting the characteristics of the data, and after the characteristics are fully extracted, all the characteristics are fused through the characteristic fusion part, so that the advanced characteristics are extracted. Finally, at the prediction output part, the load predicted value of each central air conditioning system at the next moment is output.
3. Aiming at the problem that various external uncertain factors can cause interference to the air conditioner load, a variational modal decomposition algorithm is adopted to extract signal components containing rich characteristic information from a noise-containing original load sequence, and after high-frequency components are filtered, high-frequency noise can be filtered, so that the prediction accuracy is improved.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a flow chart of a method for predicting the load of a central air conditioner of a building based on multi-task deep learning provided by an embodiment of the invention;
FIG. 2 is a block diagram of an LSTM node provided by an embodiment of the invention;
fig. 3 is a structure diagram of a convglu node according to an embodiment of the present invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
As the prediction problem of a single system mentioned in the background art of the invention, the invention adopts multi-task deep learning (MTL) which is a method capable of fusing multiple aspects of information and simultaneously solving multiple tasks, and can process different types of data by utilizing different types of neural network models and fuse the characteristics of each data so as to ensure that the prediction is more accurate.
According to the invention, the LSTM model is used for extracting the load short period characteristic, the ConvGRU is used for extracting the load long period characteristic, the BPNN is used for extracting the characteristics aiming at other factors influencing the load, such as indoor and outdoor temperature, building area and the like, the network with different structures is used for fully extracting the characteristics of data, and then all the characteristics are fused through the characteristic fusion part, so that the advanced characteristics are extracted. Finally, at the prediction output part, the load predicted value of each central air conditioning system at the next moment is output.
Aiming at the problem that various external uncertain factors can cause interference to the air conditioner load, a variational modal decomposition algorithm is adopted to extract signal components containing rich characteristic information from a noise-containing original load sequence, and after high-frequency components are filtered, high-frequency noise can be filtered, so that the prediction accuracy is improved.
Example 1
As shown in fig. 1, the embodiment provides a method for predicting the load of a central air conditioner of a building based on multi-task deep learning, which comprises the following steps:
step 1: historical data of a large building provided with M air conditioning systems is obtained.
The data includes the load of each systemIndoor temperature->Outdoor temperature->Building areaWhere i=1.m is an index of the air conditioning system and T is a sequence length.
Step 2: the data is preprocessed.
Since the air conditioning load is disturbed by various factors, there may be an abnormal value in the data. Therefore, the data set needs to be preprocessed first, and abnormal values in the data are deleted. The difference in the dimensions of the input variables affects the prediction effect, and therefore normalization of the data set is required. Raw data is normalized to a range between 0 and 1 using the linear normalization method (1).
Wherein x is max And x min For maximum and minimum values of various data, x and x * Raw data and normalized data, respectively.
And adopting a noise reduction algorithm based on variation modal decomposition to carry out noise reduction treatment on the air conditioner load.
The noise reduction processing step includes:
step 1: the sequence f (t) is first split into K modal components.
Step 2: when splitting f (t), the constraint expression (2) is firstly converted into (3) through an augmented Lagrangian function,
wherein f (t) is a sequence requiring noise reduction; u (u) k 、w k Is the kth modal component expression and center frequency; the parameter K is the number of modes; delta (t) represents a dirac function; * Is a convolution operation.
Step 3: fourier transforming f (t) of the time domain into f (ω) of the frequency domain, and then initializing
Entering a loop, updating u according to equation (4) k (ω)、w k And (omega) updating lambda (w) according to the formula (5) until the number of split reaches K, and judging whether the formula (6) is satisfied. If yes, jumping out of the circulation, and if not, continuing the circulation.
Where f (t) represents the time domain form of the signal to be split, f (ω) represents the frequency domain form of f (t), ω represents the frequency, w k Represents the center frequency, u k (ω) represents a modal component, λ (w) represents a Lagrangian multiplier, and α representsSecond order penalty factors.
u k (omega) and w k (ω) represents the kth modal component after splitting the sequence f (t) and the center frequency corresponding to the modal component,
given a precision ε >0, it is determined whether the following conditions are met:
wherein lambda is Lagrange operator; alpha is a secondary penalty factor; gamma is noise tolerance.
After the cycle is completed, the u of the frequency domain is converted by Fourier inversion k (ω)、w k (omega) conversion to time domain u k (t)、w k (t)。
Finally, the high frequency part is filtered (7) using a low pass filter to remove high frequency noise.
Wherein a is a filter coefficient, 0< a <1.
Load L of air conditioning system i by using noise reduction algorithm based on variation modal decomposition i (t) performing a treatment. The components of the processed M air conditioning system loads at the moment t are written into an input vector form (8) to serve as inputs of LSTM and ConvGRU nodes:
since the sequence length of the air conditioning load is T, it is necessary to perform normalization processing on the data set, unlike the lengths of LSTM and convglu.
For short period data, a section of shorter historical load data (with the length of D and the same length as that of LSTM) is intercepted as the actual input value of LSTM, namely { D (t), D (t-1),. The third party.d (t-D+1) }, and the corresponding actual output value is the actual load value L (t+1) ∈R of M systems at the moment of t+1 M 。
For long period data, a longer historical load data (length S (S)>D) The same length as convglu) as the actual input value of convglu, i.e., { d (t), d (t-1),..d (t-s+1) }. Vector composed of indoor and outdoor temperatures and building area at time tAs the actual input to the BPNN.
The original dataset may produce a T-s+1 pair of standard inputs and outputs. 90% of the pre-standard dataset was used as training set and the last 10% was used as test set.
Step 3: and constructing a large building central air conditioner load prediction model based on multi-task deep learning. The constructed multi-task deep nerve model mainly comprises a feature extraction part, a feature fusion part and an output part.
A feature extraction section: three network structures are used for feature extraction for different types of data.
Vector X consisting of indoor temperature, outdoor temperature and building area bp Input to the BPNN, the vector is subjected to the operation of the formula (9) by the two-layer BPNN, and Y is output bp I.e. the abstract features of this vector.
Y bp =f(W 2 f(W 1 X bp +b 1 )+b 2 ) (9)
Wherein the input vector X bp The vector is composed of indoor and outdoor temperatures and building areas; the network output is Y bp The method comprises the steps of carrying out a first treatment on the surface of the W, b is a network parameter; f (g) is the activation function.
Inputting short period data into LSTM with length D, hidden state output h of last node of LSTM D (=Y lstm ) I.e. the abstract features of the short period data to be extracted.
Specifically, LSTM is a time Recurrent Neural Network (RNN) commonly used to process time series data. For short period data, LSTM is used for feature extraction, FIG. 2 is a block diagram of a node of LSTM, and the LSTM mathematical model is as in formula (10):
wherein, the input of the node at the moment j is a vector formed by decomposing the air conditioner load, namely x j D (t-d+j), j=1, 2..d; the network output is Y lstm =h D ;i t 、f t 、o t 、h t Respectively representing an input door, a forgetting door, an output door and a hidden unit; w (W) i 、U i 、b i 、W f 、U f 、b f 、W c 、U c 、b c 、W o 、U o 、b o Is a network parameter; sigma (g), g (g) represent sigmoid and tanh functions; e represents vector element dot product.
Inputting long period data into ConvGRU with length S (11), and hiding state output h of last node of ConvGRU S (=Y cgru ) I.e. the abstract features of the long period data to be extracted. The outputs of the three networks are each represented by a one-dimensional vector.
The GRU model is faster in calculation number compared with LSTM, is suitable for processing longer sequences, convGRU is a variation of the GRU model, and the convolution operation is fused into GRU, so that network parameters are reduced, and the operation efficiency is improved. For long period data, feature extraction was performed using convglu, fig. 3 is a block diagram of one node of convglu, and the convglu mathematical model is as follows:
wherein, the input of the node at the moment j is a vector formed by decomposing the air conditioner load, namely x j D (t-s+j), t=1, 2..s; the network output is Y cgru =h S ;z t 、r t 、h t '、h t Respectively representing an update gate, a reset gate, a candidate state and a current state; w (W) z 、U z 、W r 、U r 、W h 、U h Is a network parameter; * Representing a convolution operation.
Feature fusion part: three one-dimensional vectors of the output of the feature extraction section are synthesized into one-dimensional vector X using a concat (g) function con As in formula (12).
An output section: to be synthesized into one-dimensional vector X con Input into the multi-layer BPNN, and output vector Y after operation out The load predicted value of M air conditioning systems is shown as formula (13).
After various data are processed differently, information needs to be fused to extract advanced features in the data. Because three different networks are used for processing data in the input stage of the model, the structures of the obtained outputs are inconsistent, and therefore, the outputs of the networks are needed to be fused into a vector X con :
X con =concat(Y bp ,Y lstm ,Y cgru ) (12)
And taking the fused data as input, and extracting advanced features by using the multi-layer BPNN.
Y out =F(WX con +b) (13)
Wherein Y is out ∈R M Load forecast values of M air conditioning systems; f (g) is the equivalent function of the multi-layer BPNN.
The parameter optimization algorithm employed is to optimize model parameters using a back-propagation algorithm with the goal of minimizing the loss function (14).
Wherein, parameter set θ= { W, b, W i ,U i ,b i ,W f ,U f ,b f ,W c ,U c ,b c ,W o ,U o ,b o ,W z ,U z ,W r ,U r ,W h ,U h -all parameters in the model;y is the predicted output value and the actual output value.
Example two
The embodiment provides a building central air conditioning load prediction system based on multitasking deep learning, comprising:
a data acquisition module for acquiring large building data equipped with an air conditioning system;
the load prediction module is used for combining the large building data with the air conditioning system and a large building central air conditioning load prediction model to obtain a load prediction value of the air conditioning system; the construction process of the large building central air conditioner load prediction model comprises the following steps:
the method comprises the steps of performing feature extraction by adopting a corresponding network structure aiming at different types of data, extracting load short-period features by adopting a first extraction network model, extracting load long-period features by adopting a second extraction network, extracting features of other factors influencing the load by adopting a third extraction network, performing feature fusion on all features obtained by network extraction of different structures, extracting advanced features, and performing prediction based on the advanced features to obtain load predicted values of all central air conditioning systems at the next moment.
Example III
The present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps in a method for predicting a load of a building central air conditioner based on multitasking deep learning as described above.
Example IV
The embodiment provides a computer device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the steps in the building central air conditioning load prediction method based on the multi-task deep learning when executing the program.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random access Memory (Random AccessMemory, RAM), or the like.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. The method for predicting the load of the building central air conditioner based on the multi-task deep learning is characterized by comprising the following steps of:
acquiring large building data provided with an air conditioning system;
combining the large building data with the air conditioning system and a large building central air conditioning load prediction model to obtain a load prediction value of the air conditioning system; the construction process of the large building central air conditioner load prediction model comprises the following steps:
the method comprises the steps of performing feature extraction by adopting a corresponding network structure aiming at different types of data, extracting load short-period features by adopting a first extraction network model, extracting load long-period features by adopting a second extraction network, extracting features of other factors influencing the load by adopting a third extraction network, performing feature fusion on all features obtained by network extraction of different structures, extracting advanced features, and performing prediction based on the advanced features to obtain load predicted values of all central air conditioning systems at the next moment.
2. The method for predicting the load of the building central air conditioner based on the multi-task deep learning as set forth in claim 1, wherein the feature extraction is performed by adopting a corresponding network structure for different types of data, specifically:
inputting a vector formed by other factors influencing the load into the BPNN, performing operation processing on the vector through the two layers of the BPNNs, and outputting abstract features of the vector;
the method comprises the steps of inputting short-period data into an LSTM with a first length, wherein the hidden state output of the last node of the LSTM is the abstract feature of the short-period data to be extracted, inputting long-period data into a ConvGRU with a second length, and the hidden state output of the last node of the ConvGRU is the abstract feature of the long-period data to be extracted.
3. The method for predicting the load of a building central air conditioner based on multi-task deep learning as set forth in claim 1, wherein the other factors affecting the load include: indoor temperature, outdoor temperature, and building area.
4. The method for predicting the load of a central air conditioner of a building based on multi-task deep learning as set forth in claim 1, wherein the preprocessing of data after the data of a large building with an air conditioning system is obtained, comprises: and carrying out normalization operation, noise reduction treatment and standardization treatment on the data.
5. The method for predicting the load of the central air conditioner of the building based on the multi-task deep learning as set forth in claim 4, wherein the noise reduction processing of the data adopts a noise reduction algorithm based on a variation modal decomposition to perform the noise reduction processing of the load of the air conditioner, and the method comprises the following steps:
splitting the sequence f (t) into K modal components; when splitting f (t), firstly converting the constraint expression through an augmented Lagrangian function; f (t) of the time domain is converted into f (omega) of the frequency domain through Fourier transformation, and then the f (omega) enters a cycle to update u k (ω)、w k (omega) updating lambda (w) until the number of split reaches K, judging whether the set precision is reached, if so, jumping out of the circulation, otherwise, continuing the circulation; after the cycle is completed, the u of the frequency domain is converted by Fourier inversion k (ω)、w k (omega) conversion to time domain u k (t)、w k (t) finally, filtering the high frequency part by using a low pass filter to remove high frequency noise, wherein u k (ω)、w k (ω) represents the kth modal component after splitting the sequence f (t) and the center frequency corresponding to the modal component; lambda (w) represents the lagrange multiplier.
6. The method for predicting the load of a central air conditioner in a building based on multi-task deep learning as set forth in claim 4, wherein the normalizing the data comprises:
for short period data, intercepting a section of historical load data with length D as an actual input value of LSTM, wherein the corresponding actual output value is the actual load value L (t+1) epsilon R of M systems at time t+1 M ;
For long period data, intercepting a section of historical load data with the length of S, S > D as an actual input value of ConvGRU;
the vector consisting of the indoor and outdoor temperatures and the building area at the time t is taken as the actual input of the BPNN.
7. The method for predicting the load of a building central air conditioner based on multi-task deep learning according to claim 1, wherein the model parameters are optimized by adopting a back propagation algorithm with the aim of minimizing a loss function when training a large building central air conditioner load prediction model.
8. The utility model provides a building central air conditioning load prediction system based on multitask deep learning which characterized in that includes:
a data acquisition module for acquiring large building data equipped with an air conditioning system;
the load prediction module is used for combining the large building data with the air conditioning system and a large building central air conditioning load prediction model to obtain a load prediction value of the air conditioning system; the construction process of the large building central air conditioner load prediction model comprises the following steps:
the method comprises the steps of performing feature extraction by adopting a corresponding network structure aiming at different types of data, extracting load short-period features by adopting a first extraction network model, extracting load long-period features by adopting a second extraction network, extracting features of other factors influencing the load by adopting a third extraction network, performing feature fusion on all features obtained by network extraction of different structures, extracting advanced features, and performing prediction based on the advanced features to obtain load predicted values of all central air conditioning systems at the next moment.
9. A computer readable storage medium having stored thereon a computer program, which when executed by a processor, implements the steps of the method for predicting the load of a building central air conditioner based on multitasking deep learning as claimed in any one of claims 1 to 7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method for predicting the load of a building central air conditioner based on multi-task deep learning as claimed in any one of claims 1-7 when executing the program.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311363269.1A CN117574310A (en) | 2023-10-19 | 2023-10-19 | Building central air conditioner load prediction method and system based on multitask deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311363269.1A CN117574310A (en) | 2023-10-19 | 2023-10-19 | Building central air conditioner load prediction method and system based on multitask deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117574310A true CN117574310A (en) | 2024-02-20 |
Family
ID=89890671
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311363269.1A Pending CN117574310A (en) | 2023-10-19 | 2023-10-19 | Building central air conditioner load prediction method and system based on multitask deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117574310A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117909665A (en) * | 2024-03-18 | 2024-04-19 | 青岛哈尔滨工程大学创新发展中心 | Ship motion envelope forecast data processing method and system based on Fourier filtering |
-
2023
- 2023-10-19 CN CN202311363269.1A patent/CN117574310A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117909665A (en) * | 2024-03-18 | 2024-04-19 | 青岛哈尔滨工程大学创新发展中心 | Ship motion envelope forecast data processing method and system based on Fourier filtering |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112364975B (en) | Terminal running state prediction method and system based on graph neural network | |
Abdi et al. | Forecasting of short-term traffic-flow based on improved neurofuzzy models via emotional temporal difference learning algorithm | |
de Jesus Rubio | SOFMLS: online self-organizing fuzzy modified least-squares network | |
CN111582571A (en) | Power grid operation situation sensing method and system with model driving and data driving integrated | |
Mandal et al. | In search of suitable fuzzy membership function in prediction of time series data | |
US8620631B2 (en) | Method of identifying Hammerstein models with known nonlinearity structures using particle swarm optimization | |
CN113325721B (en) | Model-free adaptive control method and system for industrial system | |
CN110570035B (en) | People flow prediction system for simultaneously modeling space-time dependency and daily flow dependency | |
CN117574310A (en) | Building central air conditioner load prediction method and system based on multitask deep learning | |
CN112272074B (en) | Information transmission rate control method and system based on neural network | |
CN112733458A (en) | Engineering structure signal processing method based on self-adaptive variational modal decomposition | |
CN114494783A (en) | Pre-training method based on dynamic graph neural network | |
Kupwiwat et al. | Deep deterministic policy gradient and graph convolutional network for bracing direction optimization of grid shells | |
Biswas et al. | Priority based fuzzy goal programming technique to fractional fuzzy goals using dynamic programming | |
CN116737521A (en) | HPC operation power consumption prediction method and system based on self-supervision comparison learning | |
CN116822722A (en) | Water level prediction method, system, device, electronic equipment and medium | |
Milad et al. | Expanded neo-fuzzy adaptive decayed brain emotional learning network for online time series predication | |
KR20160056068A (en) | Wavelet-based Time delayed adaptive neuro fuzzy inference system for predicting nonlinear behavior of smart concrete structure equipped with MR-dampers | |
Hadjistoykov et al. | Remark on intuitionistic fuzzy cognitive maps | |
CN114911157A (en) | Robot navigation control method and system based on partial observable reinforcement learning | |
Słoń | The use of fuzzy numbers in the process of designing relational fuzzy cognitive maps | |
Albelwi | A Robust Energy Consumption Forecasting Model using ResNet-LSTM with Huber Loss | |
Yu et al. | Short-Term Traffic Flow Prediction for Hybrid Time Series Decomposition Analysis and LSTM Neural Network | |
CN116894390A (en) | Wind speed prediction method and system in extreme weather | |
Ltaief et al. | PID-PSO control for Takagi-Sugeno Fuzzy model |
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 |