CN112183826A - Building energy consumption prediction method based on deep cascade generation countermeasure network and related product - Google Patents

Building energy consumption prediction method based on deep cascade generation countermeasure network and related product Download PDF

Info

Publication number
CN112183826A
CN112183826A CN202010964722.4A CN202010964722A CN112183826A CN 112183826 A CN112183826 A CN 112183826A CN 202010964722 A CN202010964722 A CN 202010964722A CN 112183826 A CN112183826 A CN 112183826A
Authority
CN
China
Prior art keywords
energy consumption
building
time sequence
consumption data
network
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.)
Granted
Application number
CN202010964722.4A
Other languages
Chinese (zh)
Other versions
CN112183826B (en
Inventor
胡书山
王鹏
余日季
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hubei University
Original Assignee
Hubei University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hubei University filed Critical Hubei University
Priority to CN202010964722.4A priority Critical patent/CN112183826B/en
Publication of CN112183826A publication Critical patent/CN112183826A/en
Application granted granted Critical
Publication of CN112183826B publication Critical patent/CN112183826B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • 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
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
    • 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
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/06—Power analysis or power optimisation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Geometry (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computer Hardware Design (AREA)
  • Human Resources & Organizations (AREA)
  • Mathematical Physics (AREA)
  • Economics (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Tourism & Hospitality (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Structural Engineering (AREA)
  • Civil Engineering (AREA)
  • Architecture (AREA)
  • General Business, Economics & Management (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)

Abstract

The embodiment of the application discloses a building energy consumption prediction method based on a deep cascade generation countermeasure network and a related product, wherein the method comprises the following steps: inputting the space factor of the building into a symmetrical residual error network, and extracting the space characteristic of the building through the symmetrical residual error network; performing local characteristic quantization on the spatial characteristic through a spatial attention mechanism to obtain a quantized spatial characteristic; inputting the quantized spatial features and the time sequence factors of the building into a bidirectional long and short term memory network to obtain the time sequence features of the building; performing local characteristic quantization on the time sequence characteristics through a time sequence attention mechanism to obtain space-time factor combined characteristics; and predicting and obtaining the energy consumption data of the building according to the space-time joint characteristics.

Description

Building energy consumption prediction method based on deep cascade generation countermeasure network and related product
Technical Field
The application relates to the technical field of computers, in particular to a building energy consumption prediction method based on a deep cascade generation countermeasure network and a related product.
Background
The building energy consumption (namely the operation energy consumption of the building) is the sum of the energy consumption of various devices (such as heating, ventilating, air conditioning, lighting and office equipment) in the operation process of the building. Building energy consumption is the first major energy consumption of the modern day, for example, the energy consumption accounts for about 35% in europe and the united states, and about 30% in china, but most building heating, ventilating and air conditioning systems have high energy consumption and carbon emission. The total energy consumption of the buildings in 2015 years in China is up to 9.64 hundred million tons of standard coal, and the total energy consumption of the buildings in 2030 years predicted by the international energy agency will reach 15.2 hundred million tons of standard coal.
The development of computer technology brings a new opportunity for building energy consumption research, and the crossed academic research of the two has become a research hotspot in recent years, and is concerned by a plurality of researchers at home and abroad, and research subjects comprise building modeling, energy consumption optimization and the like. The building energy consumption prediction takes a building form coefficient, a climate state and the like as input, and a computer is used for calculating the building thermal state and energy consumption dynamically and accurately time by time. As a multi-factor related complex nonlinear problem, the problem needs to be solved by combining mathematical modeling, engineering background knowledge and computer programming and outputting a result by relying on strong computing power of a computer.
An accurate building energy consumption prediction result plays a crucial role in building energy efficiency optimization. For a single building, energy consumption prediction can help a designer to optimize the design of a building model and the use of building materials, and the energy efficiency of the design model is improved; and the system can also assist a manager in analyzing equipment optimization and building modification strategies, so that a more reasonable energy efficiency optimization strategy is selected. The regional multi-span building is oriented, energy consumption prediction can provide data support for electric energy scheduling decision in the smart grid, energy consumption storage equipment is used for assistance, the phenomenon of power consumption peak is weakened, and the deployment and management cost of regional power supply equipment is reduced.
Disclosure of Invention
The embodiment of the application provides a building energy consumption prediction method based on a deep cascade generation countermeasure network and a related product, and the accuracy of building energy consumption prediction can be improved.
A building energy consumption prediction method based on a deep cascade generation countermeasure network is realized by generating the countermeasure network, wherein the generation countermeasure network comprises a generator and a discriminator, and the generator comprises a symmetrical residual error network, a space attention mechanism, a bidirectional long-short term memory network and a time sequence attention mechanism; the method comprises the following steps:
inputting the space factor of the building into the symmetrical residual error network, and extracting the space characteristic of the building through the symmetrical residual error network;
performing local feature quantization on the spatial features through the spatial attention mechanism to obtain quantized spatial features;
inputting the quantized spatial features and the time sequence factors of the building into the bidirectional long and short term memory network to obtain the time sequence features of the building;
performing local characteristic quantization on the time sequence characteristics through the time sequence attention mechanism to obtain space-time factor joint characteristics;
predicting and obtaining energy consumption data of the building according to the space-time joint characteristics;
wherein the discriminator is used for discriminating the authenticity of the predicted energy consumption data output by the generator when training the generator.
Further, the symmetric residual error network includes a convolution module and a deconvolution module, the convolution module includes K residual error blocks, and the deconvolution module includes K deconvolution blocks, where K is an integer greater than or equal to 1;
the inputting the space factor of the building into a symmetrical residual error network, and extracting the space characteristic of the building through the symmetrical residual error network, includes:
performing convolution processing on the space factor through K residual blocks in the convolution module to obtain intermediate features;
and carrying out deconvolution processing on the intermediate features through K deconvolution blocks in the deconvolution module to obtain the spatial features of the building.
Further, the spatial attention mechanism includes a channel attention module and a spatial attention module; the local feature quantization is performed on the spatial feature through a spatial attention mechanism to obtain a quantized spatial feature, and the method includes:
inputting the spatial features into the channel attention module to obtain channel attention weights, and obtaining first features according to the spatial features and the channel attention weights;
inputting the first feature into the spatial attention module to obtain a spatial attention weight, and obtaining the quantized spatial feature according to the first feature and the spatial attention weight.
Further, the bidirectional long-short term memory network comprises a forward propagation network and a backward propagation network;
inputting the quantized spatial features and the time sequence factors of the building into a bidirectional long-short term memory network to obtain the time sequence features of the building, wherein the method comprises the following steps:
inputting the quantized spatial features and the time sequence factors of the building into the forward propagation network to obtain first time sequence features of the building;
inputting the quantized spatial features and the time sequence factors of the building into the backward propagation network to obtain second time sequence features of the building;
and constructing and obtaining the time sequence characteristics of the building according to the first time sequence characteristics and the second time sequence characteristics.
Further, the method further comprises:
inputting a space factor sample and a time sequence factor sample of a building into a generator for generating a countermeasure network to obtain predicted energy consumption data of the building;
acquiring real energy consumption data, and inputting the predicted energy consumption data and the real energy consumption data into the discriminator for generating the countermeasure network to obtain discrimination results of the real energy consumption data and the predicted energy consumption data;
and adjusting the parameters of the generator and the discriminator according to the discrimination result.
Further, the discriminator comprises t residual blocks and a target function which are connected in sequence;
the inputting the predicted energy consumption data and the real energy consumption data into the discriminator for generating the countermeasure network to obtain the discrimination result of the real energy consumption data and the predicted energy consumption data includes:
inputting the predicted energy consumption data and the real energy consumption data into the t residual blocks to respectively obtain the distinguishing characteristic output by each residual block, wherein the distinguishing characteristic output by the previous-stage residual block is used as the input of the next-stage residual block;
and inputting the distinguishing features output by all the residual blocks into the objective function to obtain the distinguishing results of the real energy consumption data and the predicted energy consumption data.
Further, the discriminant features include a probability distribution of the predicted energy consumption data and a probability distribution of the real energy consumption data, and the objective function includes a first objective function and a second objective function;
the inputting the discrimination characteristics output by all the residual blocks into the objective function to obtain the discrimination results of the real energy consumption data and the predicted energy consumption data includes:
outputting the probability distribution of the predicted energy consumption data and the probability distribution of the real energy consumption data to the first objective function, and calculating to obtain a first divergence and a second divergence;
and inputting the first divergence and the second divergence into a second objective function to obtain a judgment result of the real energy consumption data and the predicted energy consumption data.
A building energy consumption prediction device based on a deep cascade generation countermeasure network is realized by generating the countermeasure network, wherein the generation countermeasure network comprises a generator and a discriminator, and the generator comprises a symmetrical residual error network, a space attention mechanism, a bidirectional long-short term memory network and a time sequence attention mechanism; the device comprises:
the space characteristic extraction module is used for inputting the space factor of the building into the symmetrical residual error network and extracting the space characteristic of the building through the symmetrical residual error network;
the spatial feature quantization module is used for carrying out local feature quantization on the spatial features through the spatial attention mechanism to obtain quantized spatial features;
the time sequence characteristic extraction module is used for inputting the quantized spatial characteristics and the time sequence factors of the building into the bidirectional long and short term memory network to obtain the time sequence characteristics of the building;
the time sequence characteristic quantization module is used for carrying out local characteristic quantization on the time sequence characteristics through the time sequence attention mechanism to obtain space-time factor joint characteristics;
and the energy consumption data prediction module is used for predicting and obtaining the energy consumption data of the building according to the space-time joint characteristics, wherein the discriminator is used for discriminating the authenticity of the predicted energy consumption data output by the generator when the generator is trained.
An electronic device comprises a memory and a processor, wherein the memory stores computer-executable instructions, and the processor realizes the method when executing the computer-executable instructions on the memory.
According to the building energy consumption prediction method based on the deep cascade generation countermeasure network and the related products, the characteristics of the space factors can be extracted through the symmetrical residual error neural network, and the characteristics of the time sequence factors can be extracted through the bidirectional long-short term memory network. And quantizing the spatial local feature weight through a spatial attention mechanism, a time sequence attention mechanism and a time sequence local feature weight to generate the space-time factor joint feature. The extracted features have identification power and robustness, so that the energy consumption data is predicted according to the extracted features, and the obtained energy consumption data is more accurate.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings used in the description of the embodiments will be briefly introduced below.
Fig. 1 is a schematic flow chart of a building energy consumption prediction method for generating a countermeasure network based on deep cascading in one embodiment.
FIG. 2 is a schematic diagram of energy consumption impact factors in one embodiment.
FIG. 3 is a schematic diagram of a spatial feature attention mechanism in one embodiment.
FIG. 4 is a diagram of a spatio-temporal factor joint feature extraction network in one embodiment.
FIG. 5 is a diagram of an arbiter that generates a countermeasure network in one embodiment.
FIG. 6 is a diagram of an energy consumption prediction model in one embodiment.
Fig. 7 is a schematic structural diagram of a building energy consumption prediction device for generating a countermeasure network based on deep cascading in one embodiment.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the present application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the specification of the present application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to a determination" or "in response to a detection". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".
Fig. 1 is a schematic flow chart of a building energy consumption prediction method for generating a countermeasure network based on deep cascading in an embodiment, as shown in fig. 1. The method is realized by generating the countermeasure network, wherein the generated countermeasure network comprises a generator and a discriminator, and the generator comprises a symmetrical residual error network, a space attention mechanism, a bidirectional long-short term memory network and a time sequence attention mechanism. The method includes steps 102 through 110. Wherein:
step 102, inputting the space factor of the building into a symmetrical residual error network, and extracting the space characteristic of the building through the symmetrical residual error network;
104, performing local characteristic quantization on the spatial characteristics through a spatial attention mechanism to obtain quantized spatial characteristics;
step 106, inputting the quantized spatial characteristics and the time sequence factors of the building into a bidirectional long-short term memory network to obtain the time sequence characteristics of the building;
108, performing local characteristic quantization on the time sequence characteristics through a time sequence attention mechanism to obtain space-time factor combined characteristics;
step 110, predicting and obtaining energy consumption data of the building according to space-time joint characteristics; wherein the discriminator is used for discriminating the authenticity of the predicted energy consumption data output by the generator when the generator is trained.
In one embodiment, the building refers to an internal space (e.g., office, corridor) formed by using various main elements and forms of the building to meet the needs of people in production or life. And analyzing influence factors of building energy consumption by taking a building as a unit.
FIG. 2 is a schematic diagram of energy consumption impact factors in one embodiment. As shown in fig. 2, the energy consumption impact factors can be generalized to both spatial and temporal aspects. The space factor refers to the physical characteristics of a building and has important influence on the performances of building heat preservation, illumination and the like, for example, a wall material has important influence on the building heat preservation performance, and the orientation of a window plays a decisive role in the lighting quantity. The Building Information Modeling (BIM) model is a data source of the space factor, and a Building in the BIM model can be split into a plurality of buildings according to a relevant standard. A single building factor is divided into two parts, namely self space parameters and entity component space parameters, wherein the self space parameters of the building comprise area, volume, floors where the building is located and the like. The solid components comprise walls, windows, roofs, floors, equipment and the like, the space factors of the first four components comprise parameters such as construction materials, areas, orientation, heat transfer coefficients and the like, and the equipment parameters comprise parameters such as lighting power, heating, ventilation and air conditioning power and temperature setting.
The time sequence factor refers to the outdoor and indoor time sequence state of a building, and has an important influence on the energy consumption of an indoor lighting and heating ventilation air conditioning system, such as: the hot outdoor environment causes a rapid rise in the indoor refrigeration load. The project takes weather and schedule data as data sources of time sequence factors, and the weather data comprise parameters such as outdoor air temperature, solar radiation quantity, sun angle and air humidity. The schedule data comprises schedules of indoor personnel, light, heating ventilation and air conditioning, other equipment and the like, and parameters of the schedules of the indoor personnel comprise the number and activity of the indoor personnel. The parameters of the heating ventilation air conditioning system schedule comprise on-off states, cold and warm temperature setting and the like.
For the extraction of the building parameters, the BIM model can be analyzed by using the BIMserver, and the most widely applied IFC (industry Foundation class) standard is selected as the input format of the BIM model, so that the universality and expandability of the extraction algorithm are improved. The extraction algorithm analyzes the building structure data and the equipment data, and establishes a tree structure of the building (IFCSPACE) by taking the whole building (IFCBUILDING) as a root node with respect to the structure data. The method comprises the steps of extracting attributes (such as area, orientation and the like) of each building, splitting each building into a plurality of building enclosing entities (such as wall IFCWALLSTANDARDCASE, window IFCFINDOW and the like), and extracting parameters such as three-dimensional geometric coordinate point sequence, materials, heat transfer coefficients and the like of each entity. For the device data, the extraction algorithm processes the air conditioning system (e.g., IFCBOILER, IFCCHILLER, etc.), lighting devices, and other devices in turn, which need only obtain their power and location information. Analysis of the air conditioning system requires analysis of the air circulation path (ifcdisprimotionport) in addition to obtaining basic parameters such as power, temperature settings, etc. to arrive at the building served by the hvac system. The symbols of the building energy consumption space and the time sequence factor are shown in the table 1.
TABLE 1 architectural energy consumption space and time sequence factor notation
Figure BDA0002681834430000081
Figure BDA0002681834430000091
The parameters of the time sequence factors do not need to be extracted, but the data quality needs to be guaranteed through data cleaning. Data cleaning includes missing data filling and outlier detection, and can use the mean of the time point values before and after filling the missing data, i.e.
Figure BDA0002681834430000092
The outlier refers to a data object which is obviously deviated from general data distribution in a time sequence, and due to the fact that the quantity of time sequence data such as weather and schedule related to building energy consumption is not large, the time sequence data outlier detection is completed by adopting an autoregressive integrated moving average line (ARIMA) model. The ARIMA (p, d, q) model can be represented by equation (1) based on a second order difference xt(equation (2)), the predicted value x of the original sequencetRepresented by formula (3).
Figure BDA0002681834430000093
Figure BDA0002681834430000094
Figure BDA0002681834430000095
In one embodiment, the building energy consumption related space and the time sequence factor have heterogeneous interconnectivity, namely, are independent and organically related. And constructing a deep cascade neural network to finish the joint characterization of space and time sequence factors, and designing a double-attention mechanism for the cascade network to realize the local weighting of space and time sequence characteristics. The network model includes two main parts: the device comprises a symmetrical residual error neural network and a bidirectional long and short term memory network, wherein the symmetrical residual error neural network is responsible for representing space factors, and the bidirectional long and short term memory network is responsible for representing time sequence factors. In order to improve the identification power and robustness of the model, a space attention mechanism can be designed for the symmetrical residual error neural network, and the space local characteristic weight is quantized; a time sequence attention mechanism is designed for the bidirectional long-short term memory network, and the time sequence local characteristic weight is quantized.
And after the space-time factor joint characteristics are obtained, the energy consumption data of the building can be predicted and obtained according to the space-time factor joint characteristics.
The building energy consumption prediction method based on the deep cascade generation countermeasure network provided by the embodiment can extract the characteristics of the space factors through the symmetrical residual error neural network and extract the characteristics of the time sequence factors through the bidirectional long-short term memory network. And quantizing the spatial local feature weight through a spatial attention mechanism, a time sequence attention mechanism and a time sequence local feature weight to generate the space-time factor joint feature. The extracted features have identification power and robustness, so that the energy consumption data is predicted according to the extracted features, and the obtained energy consumption data is more accurate.
In one embodiment, the symmetric residual network comprises a convolution module and a deconvolution module, wherein the convolution module comprises K residual blocks, and the deconvolution module comprises K deconvolution blocks, where K is an integer greater than or equal to 1; inputting the space factor of the building into a symmetrical residual error network, and extracting the space characteristics of the building through the symmetrical residual error network, wherein the method comprises the following steps: performing convolution processing on the space factor through K residual blocks in a convolution module to obtain intermediate features; and performing deconvolution processing on the intermediate features through K deconvolution blocks in the deconvolution module to obtain the spatial features of the building.
For example, a symmetric residual neural network consists of two modules, a convolution module and a deconvolution module. The convolution module is composed of five residual blocks, each residual block comprises five layers, the first three layers are residual convolution layers, the fourth layer is Batch Normalization, and the last layer is Dropout. The residual convolution principle is shown in equation (4), and when the number of channels x and F is different, the dimension of x is changed (equation (5)), and RReLU (Rectified Linear Unit) (equation (6)) is used as the convolution activation function, aiThe uniform distribution of U (l, U) is satisfied, l is less than U, l&u∈[0，1)。
F＝W2*RReLU(W1x)，y＝F(x，{Wi})+x (4)
y＝F(x，{Wi})+Wsx (5)
Figure BDA0002681834430000101
The residual block uses Batch Normalization (Batch Normalization) to pull the eigenvalue distribution back to the standard normal distribution, falling in the interval where the activation function is sensitive to the input. Batch normalization first calculates the mini-batch mean and variance (equation (7)), normalizes the calculation results, and makes them in a better nonlinear region by feature transformation (equation (8)).
Figure BDA0002681834430000102
Figure BDA0002681834430000103
The residual block adopts Dropout to relieve the occurrence of overfitting, and achieves the regularization effect to a certain extent, the Dropout principle is shown as formula (9), wherein the Bernoulli function generates a probability r vector, namely a vector of 0 and 1 is generated randomly.
Figure BDA0002681834430000104
The deconvolution module consists of five deconvolution blocks, each of which includes two layers of two-dimensional transpose deconvolution and batch normalization. The two-dimensional transpose deconvolution principle is shown in formula (10), kiRepresenting the ith deconvolution kernel. Combining the deconvolved intermediate feature maps
Figure BDA0002681834430000111
And obtaining a deconvolution final output characteristic diagram.
Figure BDA0002681834430000112
In one embodiment, the spatial attention mechanism includes a channel attention module and a spatial attention module; carrying out local feature quantization on the spatial features through a spatial attention mechanism to obtain quantized spatial features, wherein the method comprises the following steps: inputting the spatial features into a channel attention module to obtain a channel attention weight, and obtaining first features according to the spatial features and the channel attention weight; and inputting the first feature into a spatial attention module to obtain a spatial attention weight, and obtaining a quantized spatial feature according to the first feature and the spatial attention weight.
In one embodiment, the channel attention module includes a first average pooling layer, a first maximum pooling layer, and a full-link layer; inputting the spatial features into a channel attention module to obtain a channel attention weight, comprising: inputting the spatial features into a first average pooling layer and a first maximum pooling layer respectively to obtain averaged pooled spatial features and maximum pooled spatial features; adding the average pooled spatial features and the maximum pooled spatial features through a full connection layer to obtain an attention weight; the space attention module comprises a second average pooling layer, a second maximum pooling layer and a convolution layer; inputting the first feature into a spatial attention module to obtain a spatial attention weight, comprising: inputting the first features into a second average pooling layer and a second maximum pooling layer respectively to obtain first features after average pooling and first features after maximum pooling; and performing convolution processing on the average pooled first features and the maximum pooled first features through a convolution layer to obtain a spatial attention weight.
Specifically, the local feature quantization of the spatial feature is to quantize the channel information and the spatial information of the spatial feature output by the deconvolution module to a local feature weight. As shown in fig. 3, the attention mechanism includes a channel attention module and a spatial attention module, and the channel attention module calculates global average pooling and global maximum pooling information of feature maps, and then obtains a channel attention parameter (equation (11)) through full connection layer post-addition, i.e., a channel attention weight. The spatial attention module performs global maximum and average pooling on the coordinates of each channel feature map to obtain two feature maps, and then convolves the feature maps to obtain a spatial attention parameter (equation (12)), namely a spatial attention weight.
Figure RE-GDA0002761890750000111
Figure RE-GDA0002761890750000113
In one embodiment, the bidirectional long-short term memory network comprises a forward propagation network and a backward propagation network; inputting the quantized spatial characteristics and the time sequence factors of the building into a bidirectional long-short term memory network to obtain the time sequence characteristics of the building, wherein the time sequence characteristics comprise: inputting the quantized spatial features and the time sequence factors of the building into a forward propagation network to obtain first time sequence features of the building; inputting the quantized spatial characteristics and the time sequence factors of the building into a backward propagation network to obtain second time sequence characteristics of the building; and constructing and obtaining the time sequence characteristics of the building according to the first time sequence characteristics and the second time sequence characteristics.
For the time sequence factor, the project adopts the representation of a bidirectional long-short term memory network. The output of the current moment is related to the previous state and possibly related to the future state, the bidirectional long-term and short-term memory network respectively represents the input state through the forward propagation direction and the backward propagation direction, and the results of the two directions are integrated to obtain the final representation result. Inputting the quantized spatial features and the time sequence factors into a bidirectional long and short term memory network, obtaining first time sequence features of the building through forward propagation of the network, obtaining second time sequence features of the building through backward propagation, and constructing and obtaining the time sequence features according to the first time sequence features and the second time sequence features.
In the embodiments provided by the present application, the time sequence factor includes a time factor corresponding to n times, and the time sequence feature includes a time feature corresponding to n times, where n is an integer greater than or equal to 1; inputting the quantized spatial characteristics and the time sequence factors of the building into a bidirectional long-short term memory network to obtain the time sequence characteristics of the building, wherein the time sequence characteristics comprise: inputting the quantized spatial features and the n time factors into a forward propagation network to obtain n first time features; inputting the quantized spatial features and the n time factors into a backward propagation network to obtain n second time features; and respectively combining the first time characteristic and the second time characteristic corresponding to the same time to obtain n time characteristics.
Specifically, the local feature quantization is performed on the time sequence feature through a time sequence attention mechanism to obtain a spatio-temporal factor joint feature, which includes: inputting the n moment characteristics into a time sequence attention mechanism to obtain a time sequence weight corresponding to each moment characteristic; and carrying out weighted summation on the n moment characteristics according to the time sequence weight to obtain the space-time factor combined characteristic.
FIG. 4 is a diagram of a spatio-temporal factor joint feature extraction network in one embodiment, as shown in FIG. 4. The space-time factor combined feature extraction network comprises a symmetrical residual error network, a space attention mechanism, a bidirectional long-short term memory network and a time sequence attention mechanism. Wherein, in the two-way long and short memory network, forward propagation is taken as an example, as shown in formulas (13a) - (13e),each LSTM unit comprises a forgetting gate
Figure BDA0002681834430000131
Input gate
Figure BDA0002681834430000132
Output gate
Figure BDA0002681834430000133
Memory cell
Figure BDA0002681834430000134
Hidden state based on last unit
Figure BDA0002681834430000135
And memory state
Figure BDA0002681834430000136
Hidden state output of current cell
Figure BDA0002681834430000137
(formula (14)). Combining the forward propagation and backward propagation feature information to generate a feature representation of the time point t
Figure BDA0002681834430000138
Figure RE-GDA0002761890750000131
Figure RE-GDA0002761890750000132
Figure RE-GDA0002761890750000133
Figure RE-GDA0002761890750000134
Figure BDA00026818344300001313
Figure BDA00026818344300001314
To further quantify the importance of the local temporal features, the temporal feature attention mechanism is the feature h output at time ttGiven a relative weight αtWeight αtThe calculation principle of (2) is shown in the formulas (15) and (16). After the calculation of the relative weights at all time points is completed, the output characteristics of the bidirectional long-short term memory network can be expressed as
Figure BDA00026818344300001315
Figure BDA00026818344300001316
Figure BDA00026818344300001317
In an embodiment, the building energy consumption prediction method based on the deep cascade generation countermeasure network further includes: inputting a space factor sample and a time sequence factor sample of a building into a generator for generating a countermeasure network to obtain predicted energy consumption data of the building; acquiring real energy consumption data, and inputting the predicted energy consumption data and the real energy consumption data into the discriminator for generating the countermeasure network to obtain discrimination results of the real energy consumption data and the predicted energy consumption data; and adjusting the parameters of the generator and the discriminator according to the discrimination result.
The generation countermeasure network comprises a generator and a discriminator, wherein the generator generates energy consumption data close to real data through input space and time sequence factors, and the discriminator is responsible for discriminating the generated data and the real data. The prediction model enables the generator and the discriminator to reach balance (Nash equilibrium point) through repeated iteration to obtain a high-precision prediction result.
Specifically, a large number of samples (spatial factor samples and time series factor samples) may be input to the generator, and the predicted energy consumption data may be generated by the generator. And inputting the predicted energy consumption data and the real energy consumption data into a discriminator to obtain a discrimination result. And then the generator and the discriminator are adjusted according to the discrimination result, so that the generator and the discriminator are trained more accurately.
In an embodiment provided by the present application, the discriminator includes t residual blocks and an objective function, which are sequentially connected; the inputting the predicted energy consumption data and the real energy consumption data into the discriminator for generating the countermeasure network to obtain the discrimination result of the real energy consumption data and the predicted energy consumption data includes: inputting the predicted energy consumption data and the real energy consumption data into the t residual blocks to respectively obtain the distinguishing characteristic output by each residual block, wherein the distinguishing characteristic output by the previous-stage residual block is used as the input of the next-stage residual block; and inputting the distinguishing features output by all the residual blocks into the objective function to obtain the distinguishing results of the real energy consumption data and the predicted energy consumption data.
Specifically, as shown in fig. 5, the discriminator includes 5 residual blocks and an objective function connected in sequence, each residual block includes five structural layers, the first three layers implement residual convolution by two convolution functions and an RReLU activation function, the fourth layer is a batch normalization function, and the last layer is a Dropout function. Inputting the real energy consumption data and the predicted energy consumption data into a 1 st residual block, and sequentially transmitting the data backwards. In order to reserve more features and improve the discrimination capability, the discrimination feature output by the previous stage of residual block is used as the input of the next stage of residual block, and each residual block outputs one discrimination feature, so that the final discrimination output is obtained.
In one embodiment, the discriminative features include a probability distribution of the predicted energy consumption data and a probability distribution of the real energy consumption data, and the objective function includes a first objective function and a second objective function; inputting the discrimination characteristics output by all the residual blocks into a target function to obtain discrimination results of real energy consumption data and predicted energy consumption data, wherein the discrimination results comprise the following steps: outputting the probability distribution of the predicted energy consumption data and the probability distribution of the real energy consumption data to a first objective function, and calculating to obtain a first divergence and a second divergence; and inputting the first divergence and the second divergence into a second objective function to obtain the discrimination results of the real energy consumption data and the predicted energy consumption data.
In order to improve the accuracy of building energy consumption prediction, an objective function is established for a prediction model by combining a Wasserstein distance (equation (17)) and a gradient penalty strategy (gradient penalty). The Wasserstein distance improvement first objective function (equation (18) and equation (19)) calculates that the gradient of the first divergence (KL divergence) and the second divergence (JS divergence) disappears, and the equivalent optimization objective has the defects of unreasonable and the like. PrIs the probability distribution, P, of the real energy consumption datagTo predict a probability distribution of energy consumption data. The weight clipping mode independently limits the value range of each network parameter, and the optimal strategy leads all parameters to be extreme (namely maximum value or minimum value), possibly causing the problems of gradient disappearance or gradient explosion.
Figure BDA0002681834430000151
Figure BDA0002681834430000152
Figure BDA0002681834430000153
The model adopts a gradient punishment strategy to overcome the deficiency of Wasserstein distance, the selection range of the gradient punishment strategy is not in the whole network, and only sampling processing is carried out between true and false distributions (formula (20)), so that a second objective function (formula (21)) of the building energy consumption prediction model is obtained.
Figure BDA0002681834430000154
Figure BDA0002681834430000155
To evaluate the performance of the energy consumption prediction model, the project is intended to determine the coefficients (R-Square, R) using Root Mean Square Error (RMSE) (equation (22)), Mean Absolute Percentage Error (MAPE) (equation (23)), and the like2) (equation (24)) the three indices quantify the accuracy of the prediction.
Figure BDA0002681834430000156
Figure BDA0002681834430000161
Figure BDA0002681834430000162
FIG. 6 is a diagram of an energy consumption prediction model in one embodiment. As shown in fig. 6, the energy consumption prediction model includes four key steps of data collection, data processing, energy consumption prediction, result summarization, and forensics. The first step (data collection step) is responsible for completing the full data collection work, and the collected data comprises four types: the system comprises a building information model, weather data, schedule data and real building energy consumption monitoring data. And a second step (data processing step) of mainly finishing preprocessing the collected four types of data, including extracting building space parameters from the BIM data, finishing data cleaning on weather and schedule data, and finishing Eenergyplus modeling and correction. A third step (energy consumption prediction step) of mainly constructing a prediction model for building energy consumption and representing energy consumption related factors by combining technologies such as a residual error neural network, a long-term and short-term memory network, an attention mechanism and the like; and a countermeasure network is generated as a theoretical support, a target function suitable for building energy consumption is designed, and the accuracy of energy consumption prediction is improved. And the last step (the result summary and the condensing step) is combined with the building energy consumption prediction model and the prediction result, the condensing intelligent building energy consumption prediction model is further used for outputting the prediction result.
Fig. 7 is a schematic structural diagram of a building energy consumption prediction device for generating a countermeasure network based on deep cascading in one embodiment. As shown in fig. 7, the building energy consumption prediction apparatus based on the deep cascade generation countermeasure network is implemented by generating the countermeasure network, the generation countermeasure network includes a generator and a discriminator, the generator includes a symmetric residual error network, a spatial attention mechanism, a bidirectional long-short term memory network, a time sequence attention mechanism and an energy consumption prediction network; the device includes:
a spatial feature extraction module 702, configured to input a spatial factor of the building into the symmetric residual network, and extract a spatial feature of the building through the symmetric residual network;
a spatial feature quantization module 704, configured to perform local feature quantization on a spatial feature through a spatial attention mechanism to obtain a quantized spatial feature;
the time sequence feature extraction module 706 is used for inputting the quantized spatial features and the time sequence factors of the building into the bidirectional long-short term memory network to obtain the time sequence features of the building;
a time sequence feature quantization module 708, configured to perform local feature quantization on the time sequence features through a time sequence attention mechanism to obtain a spatio-temporal factor joint feature;
and the energy consumption data prediction module 710 is configured to predict energy consumption data of the building according to the spatio-temporal joint features, wherein the discriminator is configured to discriminate authenticity of the predicted energy consumption data output by the generator when the generator is trained.
The building energy consumption prediction device based on the deep cascade generation countermeasure network can extract the characteristics of the space factors through the symmetrical residual error neural network and extract the characteristics of the time sequence factors through the bidirectional long-short term memory network. And quantizing the spatial local feature weight through a spatial attention mechanism, a time sequence attention mechanism and a time sequence local feature weight to generate the space-time factor joint feature. The extracted features have identification power and robustness, so that the energy consumption data is predicted according to the extracted features, and the obtained energy consumption data is more accurate.
In one embodiment, the symmetric residual network comprises a convolution module and a deconvolution module, wherein the convolution module comprises K residual blocks, and the deconvolution module comprises K deconvolution blocks, where K is an integer greater than or equal to 1; the spatial feature extraction module 702 is further configured to perform convolution processing on the spatial factor through K residual blocks in the convolution module to obtain an intermediate feature; and performing deconvolution processing on the intermediate features through K deconvolution blocks in the deconvolution module to obtain the spatial features of the building.
In one embodiment, the spatial attention mechanism includes a channel attention module and a spatial attention module; the spatial feature quantization module 704 is further configured to input the spatial feature to the channel attention module, obtain a channel attention weight, and obtain a first feature according to the spatial feature and the channel attention weight; and inputting the first feature into a spatial attention module to obtain a spatial attention weight, and obtaining a quantized spatial feature according to the first feature and the spatial attention weight.
In one embodiment, the bidirectional long-short term memory network comprises a forward propagation network and a backward propagation network; the time sequence feature extraction module 706 is further configured to input the quantized spatial features and the time sequence factors of the building into a forward propagation network to obtain first time sequence features of the building; inputting the quantized spatial characteristics and the time sequence factors of the building into a backward propagation network to obtain second time sequence characteristics of the building; and constructing and obtaining the time sequence characteristics of the building according to the first time sequence characteristics and the second time sequence characteristics.
In one embodiment, the apparatus further includes a model training module, configured to input the space factor samples and the time sequence factor samples of the building to a generator for generating the countermeasure network, so as to obtain predicted energy consumption data of the building; acquiring real energy consumption data, and inputting the predicted energy consumption data and the real energy consumption data into a discriminator for generating a countermeasure network to obtain discrimination results of the real energy consumption data and the predicted energy consumption data; and adjusting the parameters of the generator and the discriminator according to the discrimination result.
In one embodiment, the discriminator comprises t residual blocks and an objective function which are connected in sequence; the model training module is also used for inputting the predicted energy consumption data and the real energy consumption data into the t residual blocks to respectively obtain the distinguishing characteristics output by each residual block, wherein the distinguishing characteristics output by the previous-stage residual block are used as the input of the next-stage residual block; and inputting the discrimination characteristics output by all the residual blocks into the objective function to obtain discrimination results of the real energy consumption data and the predicted energy consumption data.
In one embodiment, the discriminative features include a probability distribution of the predicted energy consumption data and a probability distribution of the real energy consumption data, the objective function including a first objective function and a second objective function; the model training module is further used for outputting the probability distribution of the predicted energy consumption data and the probability distribution of the real energy consumption data to the first objective function, and calculating to obtain a first divergence and a second divergence; and inputting the first divergence and the second divergence into a second objective function to obtain a judgment result of the real energy consumption data and the predicted energy consumption data.
In one embodiment, an electronic device is provided, which includes a memory having computer-executable instructions stored thereon and a processor that implements the method of the above embodiments when executing the computer-executable instructions on the memory.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, implements the method in the above-described embodiments.
In practical applications, the electronic devices may further include necessary other components, including but not limited to any number of input/output systems, processors, controllers, memories, etc., respectively, and all electronic devices that can implement the method for managing big data across cloud platforms according to the embodiments of the present application are within the scope of the present application.
The memory includes, but is not limited to, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or a portable read-only memory (CD-ROM), which is used for storing instructions and data.
The input system is for inputting data and/or signals and the output system is for outputting data and/or signals. The output system and the input system may be separate devices or may be an integral device.
The processor may include one or more processors, for example, one or more Central Processing Units (CPUs), and in the case of one CPU, the CPU may be a single-core CPU or a multi-core CPU. The processor may also include one or more special purpose processors, which may include GPUs, FPGAs, etc., for accelerated processing.
The memory is used to store program codes and data of the network device.
The processor is used for calling the program codes and data in the memory and executing the steps in the method embodiment. Specifically, reference may be made to the description of the method embodiment, which is not repeated herein.
In the several embodiments provided in the present application, it should be understood that the disclosed system and method may be implemented in other ways. For example, the division of the unit is only one logical function division, and other division may be implemented in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. The shown or discussed mutual coupling, direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, systems or units, and may be in an electrical, mechanical or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. The procedures or functions according to the embodiments of the present application are wholly or partially generated when the computer program instructions are loaded and executed on a computer. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable system. The computer instructions may be stored on or transmitted over a computer-readable storage medium. The computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)), or wirelessly (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that includes one or more of the available media. The usable medium may be a read-only memory (ROM), or a Random Access Memory (RAM), or a magnetic medium, such as a floppy disk, a hard disk, a magnetic tape, a magnetic disk, or an optical medium, such as a Digital Versatile Disk (DVD), or a semiconductor medium, such as a Solid State Disk (SSD).
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive various equivalent modifications or substitutions within the technical scope of the present application, and these modifications or substitutions should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (10)

1. The building energy consumption prediction method is characterized in that the method is realized by generating a countermeasure network, wherein the generated countermeasure network comprises a generator and a discriminator, and the generator comprises a symmetric residual error network, a space attention mechanism, a bidirectional long-short term memory network and a time sequence attention mechanism; the method comprises the following steps:
inputting the space factor of the building into the symmetrical residual error network, and extracting the space characteristic of the building through the symmetrical residual error network;
performing local feature quantization on the spatial features through the spatial attention mechanism to obtain quantized spatial features;
inputting the quantized spatial features and the time sequence factors of the building into the bidirectional long and short term memory network to obtain the time sequence features of the building;
performing local characteristic quantization on the time sequence characteristics through the time sequence attention mechanism to obtain space-time factor joint characteristics;
predicting and obtaining energy consumption data of the building according to the space-time joint characteristics;
wherein the discriminator is used for discriminating the authenticity of the predicted energy consumption data output by the generator when training the generator.
2. The method according to claim 1, wherein the symmetric residual network comprises a convolution module and a deconvolution module, the convolution module comprises K residual blocks, and the deconvolution module comprises K deconvolution blocks, wherein K is an integer greater than or equal to 1;
the inputting the space factor of the building into a symmetrical residual error network, and extracting the space characteristic of the building through the symmetrical residual error network, includes:
performing convolution processing on the space factor through K residual blocks in the convolution module to obtain intermediate features;
and carrying out deconvolution processing on the intermediate features through K deconvolution blocks in the deconvolution module to obtain the spatial features of the building.
3. The method of claim 1, wherein the spatial attention mechanism comprises a channel attention module and a spatial attention module; the local feature quantization is performed on the spatial feature through a spatial attention mechanism to obtain a quantized spatial feature, and the method includes:
inputting the spatial features into the channel attention module to obtain channel attention weights, and obtaining first features according to the spatial features and the channel attention weights;
inputting the first feature into the spatial attention module to obtain a spatial attention weight, and obtaining the quantized spatial feature according to the first feature and the spatial attention weight.
4. The method of claim 1, wherein the bidirectional long-short term memory network comprises a forward propagation network and a backward propagation network;
inputting the quantized spatial features and the time sequence factors of the building into a bidirectional long-short term memory network to obtain the time sequence features of the building, wherein the method comprises the following steps:
inputting the quantized spatial features and the time sequence factors of the building into the forward propagation network to obtain first time sequence features of the building;
inputting the quantized spatial features and the time sequence factors of the building into the backward propagation network to obtain second time sequence features of the building;
and constructing and obtaining the time sequence characteristics of the building according to the first time sequence characteristics and the second time sequence characteristics.
5. The method according to any one of claims 1 to 4, further comprising:
inputting a space factor sample and a time sequence factor sample of a building into a generator for generating a countermeasure network to obtain predicted energy consumption data of the building;
acquiring real energy consumption data, and inputting the predicted energy consumption data and the real energy consumption data into the discriminator for generating the countermeasure network to obtain discrimination results of the real energy consumption data and the predicted energy consumption data;
and adjusting the parameters of the generator and the discriminator according to the discrimination result.
6. The method of claim 5, wherein the discriminator comprises t residual blocks and an objective function connected in sequence;
the inputting the predicted energy consumption data and the real energy consumption data into the discriminator for generating the countermeasure network to obtain the discrimination result of the real energy consumption data and the predicted energy consumption data includes:
inputting the predicted energy consumption data and the real energy consumption data into the t residual blocks to respectively obtain the distinguishing characteristic output by each residual block, wherein the distinguishing characteristic output by the previous-stage residual block is used as the input of the next-stage residual block;
and inputting the distinguishing features output by all the residual blocks into the objective function to obtain the distinguishing results of the real energy consumption data and the predicted energy consumption data.
7. The method of claim 6, wherein the discriminative features include a probability distribution of the predicted energy consumption data and a probability distribution of the true energy consumption data, and wherein the objective function includes a first objective function and a second objective function;
the inputting the discrimination characteristics output by all the residual blocks into the objective function to obtain the discrimination results of the real energy consumption data and the predicted energy consumption data includes:
outputting the probability distribution of the predicted energy consumption data and the probability distribution of the real energy consumption data to the first objective function, and calculating to obtain a first divergence and a second divergence;
and inputting the first divergence and the second divergence into a second objective function to obtain a judgment result of the real energy consumption data and the predicted energy consumption data.
8. The device is characterized in that the device is realized by generating the countermeasure network, the generated countermeasure network comprises a generator and a discriminator, and the generator comprises a symmetrical residual error network, a space attention mechanism, a bidirectional long-short term memory network and a time sequence attention mechanism; the device comprises:
the space characteristic extraction module is used for inputting the space factor of the building into the symmetrical residual error network and extracting the space characteristic of the building through the symmetrical residual error network;
the spatial feature quantization module is used for carrying out local feature quantization on the spatial features through the spatial attention mechanism to obtain quantized spatial features;
the time sequence characteristic extraction module is used for inputting the quantized spatial characteristics and the time sequence factors of the building into the bidirectional long and short term memory network to obtain the time sequence characteristics of the building;
the time sequence characteristic quantization module is used for carrying out local characteristic quantization on the time sequence characteristics through the time sequence attention mechanism to obtain space-time factor joint characteristics;
and the energy consumption data prediction module is used for predicting and obtaining the energy consumption data of the building according to the space-time joint characteristics, wherein the discriminator is used for discriminating the authenticity of the predicted energy consumption data output by the generator when the generator is trained.
9. An electronic device comprising a memory having computer-executable instructions stored thereon and a processor that, when executing the computer-executable instructions on the memory, implements the method of any of claims 1-7.
10. A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the method of any one of claims 1 to 7.
CN202010964722.4A 2020-09-15 2020-09-15 Building energy consumption prediction method based on deep cascade generation countermeasure network and related products Active CN112183826B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010964722.4A CN112183826B (en) 2020-09-15 2020-09-15 Building energy consumption prediction method based on deep cascade generation countermeasure network and related products

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010964722.4A CN112183826B (en) 2020-09-15 2020-09-15 Building energy consumption prediction method based on deep cascade generation countermeasure network and related products

Publications (2)

Publication Number Publication Date
CN112183826A true CN112183826A (en) 2021-01-05
CN112183826B CN112183826B (en) 2023-08-01

Family

ID=73921027

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010964722.4A Active CN112183826B (en) 2020-09-15 2020-09-15 Building energy consumption prediction method based on deep cascade generation countermeasure network and related products

Country Status (1)

Country Link
CN (1) CN112183826B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113343334A (en) * 2021-05-28 2021-09-03 同济大学 Cross-building air conditioner energy consumption prediction method and device based on air conditioner energy consumption sensitive variable
CN113469470A (en) * 2021-09-02 2021-10-01 国网浙江省电力有限公司杭州供电公司 Energy consumption data and carbon emission correlation analysis method based on electric brain center
CN113947186A (en) * 2021-10-13 2022-01-18 广东工业大学 Heat supply energy consumption circulation prediction method based on generation of countermeasure network
CN114021811A (en) * 2021-11-03 2022-02-08 重庆大学 Attention-based improved traffic prediction method and computer medium
CN116011088A (en) * 2023-03-20 2023-04-25 天津大学 Carbon emission calculation method and device for assembled bridge engineering based on IFC expansion
CN118114350A (en) * 2024-04-22 2024-05-31 华南理工大学建筑设计研究院有限公司 GAN and GA-based low-carbon building design decision method for summer heat and winter warm areas

Citations (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120278051A1 (en) * 2011-04-29 2012-11-01 International Business Machines Corporation Anomaly detection, forecasting and root cause analysis of energy consumption for a portfolio of buildings using multi-step statistical modeling
US20150006127A1 (en) * 2013-06-28 2015-01-01 International Business Machines Corporation Constructing and calibrating enthalpy based predictive model for building energy consumption
CN106991504A (en) * 2017-05-09 2017-07-28 南京工业大学 Building energy consumption prediction method and system based on subentry measurement time sequence and building
WO2018153322A1 (en) * 2017-02-23 2018-08-30 北京市商汤科技开发有限公司 Key point detection method, neural network training method, apparatus and electronic device
CN108664687A (en) * 2018-03-22 2018-10-16 浙江工业大学 A kind of industrial control system space-time data prediction technique based on deep learning
CN108959716A (en) * 2018-06-07 2018-12-07 湖北大学 A kind of conversion method and device of Building Information Model and energy simulation model
CN109062956A (en) * 2018-06-26 2018-12-21 湘潭大学 A kind of space-time characteristic method for digging of facing area integrated energy system
CN109063903A (en) * 2018-07-19 2018-12-21 山东建筑大学 A kind of building energy consumption prediction technique and system based on deeply study
CN109948691A (en) * 2019-03-14 2019-06-28 齐鲁工业大学 Iamge description generation method and device based on depth residual error network and attention
CN109961177A (en) * 2019-03-11 2019-07-02 浙江工业大学 A kind of general water cooled central air conditioner energy consumption prediction technique based on shot and long term memory Recognition with Recurrent Neural Network
CN111080002A (en) * 2019-12-10 2020-04-28 华南理工大学 Deep learning-based multi-step prediction method and system for building electrical load
US20200134804A1 (en) * 2018-10-26 2020-04-30 Nec Laboratories America, Inc. Fully convolutional transformer based generative adversarial networks
CN111091045A (en) * 2019-10-25 2020-05-01 重庆邮电大学 Sign language identification method based on space-time attention mechanism
CN111178626A (en) * 2019-12-30 2020-05-19 苏州科技大学 Building energy consumption prediction method and monitoring prediction system based on WGAN algorithm
CN111223301A (en) * 2020-03-11 2020-06-02 北京理工大学 Traffic flow prediction method based on graph attention convolution network
CN111426344A (en) * 2020-03-20 2020-07-17 淮阴工学院 Building energy consumption intelligent detection system
CN111475546A (en) * 2020-04-09 2020-07-31 大连海事大学 Financial time sequence prediction method for generating confrontation network based on double-stage attention mechanism

Patent Citations (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120278051A1 (en) * 2011-04-29 2012-11-01 International Business Machines Corporation Anomaly detection, forecasting and root cause analysis of energy consumption for a portfolio of buildings using multi-step statistical modeling
US20150006127A1 (en) * 2013-06-28 2015-01-01 International Business Machines Corporation Constructing and calibrating enthalpy based predictive model for building energy consumption
WO2018153322A1 (en) * 2017-02-23 2018-08-30 北京市商汤科技开发有限公司 Key point detection method, neural network training method, apparatus and electronic device
CN106991504A (en) * 2017-05-09 2017-07-28 南京工业大学 Building energy consumption prediction method and system based on subentry measurement time sequence and building
CN108664687A (en) * 2018-03-22 2018-10-16 浙江工业大学 A kind of industrial control system space-time data prediction technique based on deep learning
CN108959716A (en) * 2018-06-07 2018-12-07 湖北大学 A kind of conversion method and device of Building Information Model and energy simulation model
CN109062956A (en) * 2018-06-26 2018-12-21 湘潭大学 A kind of space-time characteristic method for digging of facing area integrated energy system
CN109063903A (en) * 2018-07-19 2018-12-21 山东建筑大学 A kind of building energy consumption prediction technique and system based on deeply study
US20200134804A1 (en) * 2018-10-26 2020-04-30 Nec Laboratories America, Inc. Fully convolutional transformer based generative adversarial networks
CN109961177A (en) * 2019-03-11 2019-07-02 浙江工业大学 A kind of general water cooled central air conditioner energy consumption prediction technique based on shot and long term memory Recognition with Recurrent Neural Network
CN109948691A (en) * 2019-03-14 2019-06-28 齐鲁工业大学 Iamge description generation method and device based on depth residual error network and attention
CN111091045A (en) * 2019-10-25 2020-05-01 重庆邮电大学 Sign language identification method based on space-time attention mechanism
CN111080002A (en) * 2019-12-10 2020-04-28 华南理工大学 Deep learning-based multi-step prediction method and system for building electrical load
CN111178626A (en) * 2019-12-30 2020-05-19 苏州科技大学 Building energy consumption prediction method and monitoring prediction system based on WGAN algorithm
CN111223301A (en) * 2020-03-11 2020-06-02 北京理工大学 Traffic flow prediction method based on graph attention convolution network
CN111426344A (en) * 2020-03-20 2020-07-17 淮阴工学院 Building energy consumption intelligent detection system
CN111475546A (en) * 2020-04-09 2020-07-31 大连海事大学 Financial time sequence prediction method for generating confrontation network based on double-stage attention mechanism

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
孙秋野等: "基于GAN技术的自能源混合建模与参数辨识方法", 自动化学报, no. 05, pages 901 - 914 *
张建晋;王韫博;龙明盛;***;王海峰;: "面向季节性时空数据的预测式循环网络及其在城市计算中的应用", 计算机学报, no. 02, pages 286 - 302 *
林跃东;许巧玲;陈东;: "基于PCA―Elman神经网络的建筑能耗预测", 智能建筑电气技术, no. 04, pages 5 - 8 *
王悦;黄泽天;邹锋;: "基于生成对抗网络的Q学习能耗预测方法", 电脑知识与技术, no. 23 *
王鹏: "浅谈智能建筑新技术和建筑节能", 黑龙江科技信息, no. 06, pages 198 *
胡书山等: "Environmental and energy performance assessment of buildings using scenario modelling and fuzzy analytic network process", APPLIED ENERGY, vol. 255, pages 1 - 12 *
邹锋等: "基于生成对抗网络的深度学习能耗预测算法", 电脑知识与技术, vol. 15, no. 02, pages 198 - 200 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113343334A (en) * 2021-05-28 2021-09-03 同济大学 Cross-building air conditioner energy consumption prediction method and device based on air conditioner energy consumption sensitive variable
CN113469470A (en) * 2021-09-02 2021-10-01 国网浙江省电力有限公司杭州供电公司 Energy consumption data and carbon emission correlation analysis method based on electric brain center
CN113947186A (en) * 2021-10-13 2022-01-18 广东工业大学 Heat supply energy consumption circulation prediction method based on generation of countermeasure network
CN114021811A (en) * 2021-11-03 2022-02-08 重庆大学 Attention-based improved traffic prediction method and computer medium
CN116011088A (en) * 2023-03-20 2023-04-25 天津大学 Carbon emission calculation method and device for assembled bridge engineering based on IFC expansion
CN118114350A (en) * 2024-04-22 2024-05-31 华南理工大学建筑设计研究院有限公司 GAN and GA-based low-carbon building design decision method for summer heat and winter warm areas

Also Published As

Publication number Publication date
CN112183826B (en) 2023-08-01

Similar Documents

Publication Publication Date Title
CN112183826B (en) Building energy consumption prediction method based on deep cascade generation countermeasure network and related products
Luo et al. Development of an IoT-based big data platform for day-ahead prediction of building heating and cooling demands
Zhang et al. A systematic feature selection procedure for short-term data-driven building energy forecasting model development
Ding et al. Analysis of feature matrix in machine learning algorithms to predict energy consumption of public buildings
Xikai et al. Comparison of regression models for estimation of carbon emissions during building's lifecycle using designing factors: a case study of residential buildings in Tianjin, China
Cheng-wen et al. Application of ANN for the prediction of building energy consumption at different climate zones with HDD and CDD
Luo A novel clustering-enhanced adaptive artificial neural network model for predicting day-ahead building cooling demand
CN112149890A (en) Comprehensive energy load prediction method and system based on user energy label
Amasyali et al. Deep learning for building energy consumption prediction
CN114282730A (en) Data completeness inspection and feature learning method for building load prediction
Tien et al. Occupancy heat gain detection and prediction using deep learning approach for reducing building energy demand
CN116796141A (en) GBDT regression model-based office building energy consumption prediction method
Chen et al. A training pattern recognition algorithm based on weight clustering for improving cooling load prediction accuracy of HVAC system
Lu et al. A hybrid ensemble learning framework for zero-energy potential prediction of photovoltaic direct-driven air conditioners
Fang et al. Comprehensive clustering method to determine coincident design day for air-conditioning system design
Salami et al. Building energy loads prediction using bayesian-based metaheuristic optimized-explainable tree-based model
Xue et al. A framework for short-term energy consumption prediction based on room air conditioner group characteristics
Zhao et al. An artificial intelligence (AI)-driven method for forecasting cooling and heating loads in office buildings by integrating building thermal load characteristics
Zhang et al. Forecasting building plug load electricity consumption employing occupant-building interaction input features and bidirectional LSTM with improved swarm intelligent algorithms
Banda et al. Transfer learning for leisure centre energy consumption prediction
CN112232479B (en) Building energy time-consuming space factor characterization method based on deep cascade neural network and related products
Das et al. A study on the application of artificial intelligence techniques for predicting the heating and cooling loads of buildings
CN115081533A (en) Client side load prediction method and system based on two-stage clustering and MGRU-AT
CN114925929A (en) Building load prediction method and device based on degree days
Chu et al. Data-Driven Energy Prediction in Residential Buildings using LSTM and 1-D CNN.

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