CN112149364A - Intelligent human residential environment airflow organization optimization method - Google Patents

Intelligent human residential environment airflow organization optimization method Download PDF

Info

Publication number
CN112149364A
CN112149364A CN202010919008.3A CN202010919008A CN112149364A CN 112149364 A CN112149364 A CN 112149364A CN 202010919008 A CN202010919008 A CN 202010919008A CN 112149364 A CN112149364 A CN 112149364A
Authority
CN
China
Prior art keywords
data
parameters
self
ventilation
building
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
CN202010919008.3A
Other languages
Chinese (zh)
Other versions
CN112149364B (en
Inventor
曾令杰
高军
张承全
贺廉洁
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.)
Tongji University
Original Assignee
Tongji 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 Tongji University filed Critical Tongji University
Priority to CN202010919008.3A priority Critical patent/CN112149364B/en
Publication of CN112149364A publication Critical patent/CN112149364A/en
Application granted granted Critical
Publication of CN112149364B publication Critical patent/CN112149364B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
    • Y02B30/00—Energy efficient heating, ventilation or air conditioning [HVAC]
    • Y02B30/70—Efficient control or regulation technologies, e.g. for control of refrigerant flow, motor or heating

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Analysis (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Fluid Mechanics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Algebra (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Architecture (AREA)
  • Civil Engineering (AREA)
  • Structural Engineering (AREA)
  • Computational Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

An intelligent human-living environment airflow organization optimization method realizes rapid and global optimization of human-living environment airflow organization through introduction of artificial intelligence. Firstly, quickly organizing and summarizing geometric data and ventilation parameters in a building room and flow field information data corresponding to the geometric data and the ventilation parameters, and abstracting hidden association behind the data in a machine learning mode; secondly, customizing a building environment performance simulation tool based on machine learning, and applying the tool to a ventilation scheme and the design of a building indoor structure. The technical scheme of the application has the advantages that the simulation calculation efficiency of the human-living environment airflow organization can be greatly improved and the time required for optimizing the airflow organization is reduced due to the introduction of the artificial intelligence algorithm based on machine learning.

Description

Intelligent human residential environment airflow organization optimization method
Technical Field
The invention belongs to the field of human-living environment airflow organization optimization, and relates to an artificial intelligence-assisted intelligent human-living environment airflow organization optimization method.
Background
The air conditioning system brings comfortable indoor environment for people, also plays a role in ventilation and decontamination, and creates a relatively healthy space for the work and the residence of people. Facing to the general requirement of building energy conservation, people continuously improve the sealing property of buildings in order to prevent the cold and hot loads of the buildings caused by air permeation, thereby reducing the exchange of indoor and outdoor air, causing the concentration of indoor carbon dioxide and other airborne pollutants to be continuously increased, causing the quality reduction of the indoor air, and leading residents to have uncomfortable reactions and diseases in the environment. Relevant researches show that ventilation can greatly improve indoor air quality and reduce indoor air pollution, thereby reducing the probability of sickening people. Because the distribution of indoor air pollution after emission is mainly influenced by airflow organization which is mainly generated by the synergistic action of a building structure and a ventilation system, the optimization of the airflow organization of the human-living environment is an important ring of healthy and intelligent building design and an important means for improving the indoor air quality. At present, the main method for optimizing indoor airflow organization is to perform traversal simulation on indoor airflow organization adopting different ventilation schemes by means of Computational Fluid Dynamics (CFD), and the simulation iteration process needs to be repeated by a computer for each ventilation scheme or the fine adjustment of ventilation parameters, which wastes time and labor for building designers. Meanwhile, because a large amount of computing resources are consumed in the simulation process, the airflow organization optimization obviously can only screen a better scheme from a limited number of ventilation schemes, and the scheme is likely to be only local optimization under a specific condition, and the global optimization of the airflow organization is difficult to realize.
Disclosure of Invention
In view of the above disadvantages of the prior art, the present invention aims to provide an artificial intelligence-assisted intelligent human-living environment airflow organization optimization method, which realizes rapid and global optimization of human-living environment airflow organization by introducing artificial intelligence. The intelligent human-living environment is an indoor environment which is optimized and built by means of artificial intelligence and is suitable for people to live in.
In order to achieve the purpose, the invention adopts the technical scheme that:
firstly, quickly organizing and summarizing geometric data and ventilation parameters in a building room and flow field information data corresponding to the geometric data and the ventilation parameters, and abstracting hidden association behind the data in a machine learning mode; secondly, customizing a building environment performance simulation tool based on machine learning, and applying the tool to a ventilation scheme and preliminary promotion of building indoor construction. The method can exponentially reduce the simulation duration of the complex CFD and provide decision support for the indoor airflow organization global optimization.
The technical scheme of the invention mainly comprises the following steps: (1) establishing a database of matching pairs of the geometric parameters and the ventilation parameters of the building and the corresponding flow field data; (2) dividing a database of matching pairs into a training set and a testing set, training the database matching pairs by a machine learning method, and abstracting a hidden function relation behind data; (3) excavating mathematical mapping relations among the geometry, ventilation parameters and airflow organization of the building in the test set; (4) and customizing a machine learning-based performance simulation tool to realize real-time synchronous acquisition of the airflow organization under the change of the ventilation parameters, so that the global optimization of the airflow organization becomes possible.
Further, the method comprises the following steps:
(1) establishing a matching pair database consisting of the geometrical parameters and the ventilation parameters of the building and the flow field data corresponding to the geometrical parameters and the ventilation parameters:
(1.1) extracting the geometric parameters and the ventilation parameters of the buildings and the flow field data corresponding to the geometric parameters and the ventilation parameters from various indoor building airflow organizations obtained by the disclosed CFD simulation, and constructing a matching pair database consisting of three types of data, wherein the three types of data aiming at different working conditions are in one-to-one correspondence, namely, the corresponding flow field data are unique under a group of geometric parameters of the buildings and set ventilation parameters.
(1.2) the building geometric parameters are multidimensional vectors X consisting of building external dimensions and internal structures; the ventilation parameter is a three-dimensional vector Y consisting of a tuyere position, a wind speed and an angle; the air flow field data is a multidimensional matrix P formed by velocity vectors on CFD grid nodesv。
(1.3) matching the geometric parameters and the ventilation parameters of the buildings in the database as input data sets, which can be simplified into
Figure BDA0002666016570000021
For n-dimensional input variables (n ═x + y), airflow field data as output data set y(k)。
(1.4) splitting the database of matching pairs into training sets TtrainAnd test set TtestTwo parts.
Namely, the matching pair database T is:
T＝TTrain∪TTest (1)
(1.5) the new CFD simulation data disclosed can be added into the matching pair database at any time to expand the learning sample number.
(2) And learning the data set in the database in a matching way by using a machine learning model formed by stacking a plurality of self-coding networks, and abstracting a hidden function relation behind the data.
(2.1) the self-coding network is an unsupervised neural network comprising an input layer, a hidden layer and an output layer. The network obtains a limited number of characteristic representations through self-learning of the original characteristics, and achieves the purpose of reconstructing input by using the characteristic representations.
(2.2) parameter learning from the coding network is divided into two processes: an encoding process and a decoding process. In the encoding process, firstly, the hidden layer lambda is added1(x) Self-learning is carried out, wherein1(x) The calculation formula of (a) is as follows:
λ1(x)＝w(Y1x+c1) (2)
wherein, Y1To code the matrix, c1To encode the bias vector, w (-) is a tan h function.
(2.3) self-coding network architecture with M hidden layers as shown in FIG. 1. The decoding process realizes the representation of the hidden layer by determining a decoding matrix1(x) Decoding into reconstructed data lambda2(x) Process of reconstructing data λ2(x) Is output by the formula
λ2(x)＝f(Y2x+c2) (3)
Wherein, Y2To decode the matrix, c2To decode the offset vector, f (-) is a tan h function.
The self-coding network learning process implements the optimization process of the network parameters by minimizing the mean square error cost function as shown below, i.e.
Figure BDA0002666016570000031
Thus, the optimal set of parameters for a self-coding network can be translated into solving the following optimization problem
Figure BDA0002666016570000032
The optimization problem is typically solved by a BP neural network algorithm. And stacking a plurality of self-coding networks on the basis to obtain the machine learning model for mining the implicit functional relation of the data.
(2.4) the structure and training method of the stacked self-coding learning network with j hidden layers is shown in fig. 2.
In training set TTestThe final output of the model at the jth hidden layer can be expressed as:
Figure BDA0002666016570000033
wherein the content of the first and second substances,
Figure BDA0002666016570000034
and
Figure BDA0002666016570000035
(j ═ 1,2, …, m) are the coding matrix and coding offset vector of the jth self-coding network, respectively, and w (·) is the tan h function.
The above Machine Learning Model (MLM) learns network parameters through a layer-by-layer pre-training method, and how to mine the mathematical mapping between inputs and outputs will be discussed in the following section.
(3) And (3) excavating mathematical mapping relations among the building geometry, ventilation parameters and airflow organization in a training set and a testing set: the method comprises two parts of stack self-coding network pre-training and least square learning of output weight.
(3.1) stacked self-coding network pre-training first trains the first layer of MLM as a self-coding network, minimizes equation (4) with training data as input, and initializes χ ═ 2.
(3.2) when training the chi-layer, will
Figure BDA0002666016570000036
As an input to minimize equation (4).
(3.3) making χ ═ χ +1, and iterating (3.2) steps; and stopping iteration when x is larger than j, and turning to (3.4).
(3.4) the final output of the network is
Figure BDA0002666016570000037
It is input as a learning model.
And (3.5) in the part for mining the input and output mathematical mapping relation, optimizing the output weight theta by adopting a least square method. When all the self-coding network parameters
Figure BDA0002666016570000038
All determined, inputting the data set
Figure BDA0002666016570000039
The corresponding hidden layer represents lambdaj(x(k)) Are known. In the air flow texture prediction problem, it is always desirable to relate to x(k)Is estimated value of
Figure BDA00026660165700000310
Can accurately approximate to the actual output y(k)It is written as the following equation:
Figure BDA00026660165700000311
wherein the content of the first and second substances,
Figure BDA0002666016570000041
namely:
λj(x)θ＝y (7)
the above formula describes the mathematical mapping relationship between the input parameters (building geometry, ventilation parameters) and the output parameters (airflow organization node data), wherein the implicit function lambdaj(x) This can be learned from the following formula:
Figure BDA0002666016570000042
according to the matrix theory, the optimal output weight vector θ is a least-squares solution of the equation, i.e.:
θ＝λj(x)↑y (9)
equations (7) - (9) are used to directly learn the node airflow field data that matches the output given the input building geometry and ventilation parameters.
(4) Customizing a machine learning-based performance simulation tool to realize real-time synchronous acquisition of airflow organization under ventilation parameter change
Based on the MLM models of sections 2 and 3, a functional simulation tool for the air flow organization can be obtained. The condition of the ventilation airflow organization of a certain public building is simulated, as shown in fig. 3, the building has the external dimension (length, width and height) of 25.4m multiplied by 30m multiplied by 4.5m, and is provided with a set of full air system which comprises 1 set of air supply system (SA1) and two sets of air return systems (RA1, RA2), the tail end of the air supply system is provided with 16 air supply outlets, the air return systems are provided with 8 air return outlets, the sizes of the air supply outlets are 0.3m multiplied by 0.3m, the sizes of the air return outlets are 0.4m multiplied by 0.2m, and the size of the air outlet is 0.63m multiplied by 0.32 m. The SF6 was used as a tracer gas to simulate the indoor SF6 concentration field under different ventilation conditions to identify the flow pattern. If fig. 4A shows the concentration distribution of SF6 in the fully open scene of all the air ports simulated by the above machine learning model, and fig. 4B shows the concentration distribution of SF6 in the closed scene of a part of the air ports at the same time, with the aid of the machine learning model, the simulation of fig. 4A to 4B can be completed in real time by changing the ventilation parameters without performing repeated calculation like the conventional CFD simulation, which greatly saves the time required by the simulation of the air flow organization, and provides technical support for the real-time online optimization of the air flow organization, such as controlling the concentration of a certain area to be lower than a certain threshold value.
Due to the adoption of the technical scheme, the invention has the following technical effects: the technical scheme has the advantages that the artificial intelligence algorithm based on machine learning is introduced, the simulation calculation efficiency of the human-living environment air flow organization can be greatly improved, the time required by the air flow organization optimization is shortened, and a new technical means and a new method are provided for the overall optimization of the indoor environment of the intelligent building. The method is completely different from the conventional indoor airflow field simulation optimization based on CFD in the aspect of main technical paths, and the CFD is mainly based on a fluid dynamics basic equation and obtains an approximate solution of a fluid control equation by utilizing the quick computing capability of a computer; the method is based on the existing building airflow organization simulation database, the hidden association behind the data is abstracted in a machine learning mode, the mathematical mapping relation between the building geometry, the ventilation parameters (air port position, wind speed, area and the like) and the airflow organization is mined on the basis of the full training of the model, and then the airflow organization under the ventilation parameter change is synchronously acquired in real time by customizing a performance simulation tool based on the machine learning, so that the global optimization of the airflow organization becomes possible. In the present invention, the results of CFD and related experiments are only used as a database for machine learning training.
Drawings
Fig. 1 is a diagram of a self-coding network architecture with M hidden layers.
FIG. 2 is a diagram of a structure and training method of a stacked self-coding learning network with j hidden layers.
Figure 3 is a schematic view of a building having a set of all-air systems.
Fig. 4A is a SF6 concentration distribution diagram under the condition that all the tuyere is fully opened and simulated by the machine learning model of the embodiment shown in fig. 3.
FIG. 4B is a graph of SF6 concentration distribution at the same time with partial tuyere closure simulated by the machine learning model of the embodiment shown in FIG. 4A.
FIG. 5 is a flowchart of the method for optimizing airflow organization in smart human-occupiable environment according to the present invention.
Fig. 6 is a schematic view of the internal environment of a supermarket.
Fig. 7A is a CO2 concentration distribution diagram under the condition that all the air ports are fully opened, simulated by the supermarket machine learning model shown in fig. 6.
FIG. 7B is a CO2 concentration profile simulated by the machine learning model at the same time when a portion of the tuyere in FIG. 7A was closed.
Detailed Description
The invention is further described below with reference to the figures and examples of the invention.
The flow chart of the intelligent human-living environment airflow organization optimization method is shown in fig. 5. The specific air flow tissue learning optimization is carried out according to the steps included in the flow, and the process is as follows:
s101, establishing a matching pair database consisting of the building geometric parameters, the ventilation parameters and the flow field data corresponding to the building geometric parameters and the ventilation parameters, namely extracting the building geometric parameters, the ventilation parameters and the flow field data corresponding to the building geometric parameters and the ventilation parameters from various building indoor airflow organizations obtained by the disclosed CFD simulation, and constructing the matching pair database consisting of the three types of data.
Specifically, for the internal environment of the supermarket as shown in fig. 6, the building geometric parameters and the internal structure thereof can be extracted from the modeling data of the CFD simulation, which constitutes a multidimensional vector X; similarly, ventilation parameters such as the position, the wind speed and the angle of the wind gap can also be extracted from the modeling data to form a multi-dimensional vector Y; the air flow field data is extracted from the arithmetic calculation data of CFD simulation, and a multidimensional matrix P is formed by velocity vectors on nodes of CFD gridv. Further, taking multidimensional vectors X and Y as input data sets, it can be simplified to
Figure BDA0002666016570000061
Which is an n-dimensional input variable (n ═ x + y), and a gas multidimensional matrix PvAs an output data set y(k). Further, the above may be combined with the input data set x(k)And output data set y(k)Splitting the formed matching pair database T into training sets TtrainAnd test set TtestTwo parts, i.e.
T＝TTrain∪TTest
Furthermore, the simulation data of the supermarket is only needed for describing the embodiment, and the CFD simulation data of various public environments (home, office and conference room) can be added into the matching pair database at any time to expand the number of learning samples.
S102, learning a data set in a database in a matching mode by using a machine learning model formed by stacking a plurality of self-coding networks, and abstracting a hidden functional relation behind the data.
Specifically, for the supermarket as shown in fig. 6, the self-encoding network is shown in fig. 1, and the network specifically includes 1 input layer, 1 hidden layer, and 1 output layer, and the parameter learning thereof can be divided into an encoding process and a decoding process.
Wherein, in the encoding process, firstly, a single hidden layer lambda is subjected to1(x) Self-learning is carried out, wherein1(x) The calculation formula of (a) is as follows:
λ1(x)＝w(Y1x+c1)
in the formula, Y1To code the matrix, c1To encode the bias vector, w (-) is a tan h function.
The decoding process realizes the single hidden layer representation lambda by determining a decoding matrix1(x) Decoding into reconstructed data lambda2(x) Process of reconstructing data λ2(x) Is output by the formula
λ2(x)＝f(Y2x+c2)
Wherein, Y2To decode the matrix, c2To decode the offset vector, f (-) is a tan h function.
Further, the self-coding network learning process implements the optimization process of the network parameters by minimizing the mean square error cost function as shown below, i.e.
Figure BDA0002666016570000062
Thus, the optimal set of parameters for a self-coding network can be translated into solving the following optimization problem
Figure BDA0002666016570000063
The optimization problem is solved by the disclosed standard BP neural network algorithm.
Further, for the present embodiment, the learning module is composed of a stack self-coding learning network having 5 hidden layers, and the structure and training method of the stack self-coding learning network are shown in fig. 2.
In training set TTestThe final output of the model at the 5 th hidden layer can be expressed as:
Figure BDA0002666016570000071
wherein the content of the first and second substances,
Figure BDA0002666016570000072
and
Figure BDA0002666016570000073
the coding matrix and the coding bias vector of the 5 th self-coding network, respectively, w (·) is a tan h function. The above machine learning model learns the network parameters through a layer-by-layer pre-training algorithm, and how to mine the mathematical mapping relationship between the input and the output will be detailed in S103.
S103, excavating mathematical mapping relations among the building geometry, ventilation parameters and airflow organization in a training set and a testing set: the method comprises two parts of stack self-coding network pre-training and least square learning of output weight.
Specifically, for a supermarket as shown in fig. 6, the stacked self-coding network pre-training involved in the method first trains the first layer of the MLM as a self-coding network, minimizes the formula (4) with training data as input, and initializes χ ═ 2. The network trains 5 layers in total, wherein, when training the chi layer, the network trains
Figure BDA0002666016570000074
As an input to minimize equation (4). And let χ ═ χ +1, and iterate (3.2) step; when χ is more than 5, the iteration is stopped, and the process is shifted to (3.4). The final output of the network is
Figure BDA0002666016570000075
It is input as a learning model.
Further, in the part of mining the input and output mathematical mapping relation, a least square method is adopted to optimize the output weight theta. When all the self-coding network parameters
Figure BDA0002666016570000076
All determined, inputting the data set
Figure BDA0002666016570000077
The corresponding hidden layer represents lambdaj(x(k)) Are known. In the air flow texture prediction problem, it is always desirable to relate to x(k)Is estimated value of
Figure BDA0002666016570000078
Can accurately approximate to the actual output y(k)It is written as the following equation:
Figure BDA0002666016570000079
wherein the content of the first and second substances,
Figure BDA00026660165700000710
namely:
λj(x)θ＝y(j＝1,2…5)
the above formula describes the mathematical mapping relationship between the input parameters (building geometry, ventilation parameters) and the output parameters (airflow organization node data) of the supermarket as shown in fig. 6, wherein the implicit function λ isj(x) This can be learned from the following formula:
Figure BDA0002666016570000081
Figure BDA0002666016570000082
Figure BDA0002666016570000083
finally, according to the matrix theory, the optimal output weight vector θ is the minimum norm least squares solution of the formula, i.e.:
θ＝λj(x)↑y
the above formula is expressed in the self-learning network lambdaj(x)↑Giving an input building geometry and ventilation parameter data set x on the basis of full training and the determination of the optimal output weight vector theta(k)To learn and match the output node airflow field data set y(k)。
And S104, customizing a machine learning-based performance simulation tool to realize real-time synchronous acquisition of the airflow organization under the condition of ventilation parameter change.
In particular, the performance simulation tool is implemented by software packaging of the machine learning process described above. For a supermarket such as that shown in fig. 6, the geometric parameters and ventilation parameters of the building can be directly extracted from CFD model data, the data of the airflow field can also be extracted from the computational data of the model by CFD, and through the steps of the above embodiment, the indoor CO when the air opening is fully opened can be firstly realized2Concentration (shown in FIG. 7A), and indoor CO when part of the tuyere is closed2The concentration (shown in fig. 7B) can then be done in real time by changing the ventilation parameters without the need for recalculation as in conventional CFD simulations.
The embodiments described above are described to facilitate an understanding and use of the invention by those skilled in the art. It will be readily apparent to those skilled in the art that various modifications to these embodiments may be made, and the generic principles described herein may be applied to other embodiments without the use of the inventive faculty. Therefore, the present invention is not limited to the above embodiments, and those skilled in the art should make improvements and modifications within the scope of the present invention based on the disclosure of the present invention.

Claims (6)

1. An intelligent human-living environment airflow organization optimization method is characterized by comprising the following steps: the rapid and global optimization of the human-living environment airflow organization is realized by introducing artificial intelligence.
2. The intelligent human-occupiable environment airflow organization optimizing method of claim 1, characterized in that: firstly, quickly organizing and summarizing geometric data and ventilation parameters in a building room and flow field information data corresponding to the geometric data and the ventilation parameters, and abstracting hidden association behind the data in a machine learning mode; secondly, customizing a building environment performance simulation tool based on machine learning, and applying the tool to a ventilation scheme and the design of a building indoor structure.
3. The intelligent human-occupiable environment airflow organization optimizing method of claim 1, comprising the steps of:
(1) establishing a matching pair database consisting of the geometric parameters and the ventilation parameters of the building and the flow field data corresponding to the geometric parameters and the ventilation parameters;
(2) learning a data set in a database in a matching way by using a machine learning model formed by stacking a plurality of self-coding networks, and abstracting a hidden function relation behind the data;
(3) and (3) excavating mathematical mapping relations among the building geometry, ventilation parameters and airflow organization in a training set and a testing set: the method comprises two parts of stack self-coding network pre-training and least square learning of output weight;
(4) and customizing a machine learning-based performance simulation tool to realize real-time synchronous acquisition of airflow organization under the condition of ventilation parameter change.
4. The method for optimizing airflow organization in a smart human-occupiable environment according to claim 3, wherein the step (1) comprises:
(1.1) extracting geometric parameters and ventilation parameters of buildings and flow field data corresponding to the geometric parameters and the ventilation parameters from various indoor airflow organizations of the buildings obtained by the disclosed CFD simulation, and constructing a matching pair database consisting of three types of data, wherein the three types of data for different working conditions are in one-to-one correspondence;
(1.2) the building geometric parameters are multidimensional vectors X consisting of building external dimensions and internal structures; the ventilation parameter is a three-dimensional vector Y consisting of a tuyere position, a wind speed and an angle; the air flow field data is a multidimensional matrix P formed by velocity vectors on CFD grid nodesv；
(1.3) matching the geometric parameters and the ventilation parameters of the buildings in the database as input data sets, which can be simplified into
Figure FDA0002666016560000011
For n-dimensional input variables (n ═ x + y), the flow field data are used as the output data set y(k)；
(1.4) splitting the database of matching pairs into training sets TtrainAnd test set TtestTwo parts;
namely, the matching pair database T is:
T＝TTrain∪TTest (1)
(1.5) the new CFD simulation data disclosed can be added into the matching pair database at any time to expand the learning sample number.
5. The method of claim 3, wherein the step (2) comprises:
(2.1) the self-coding network is an unsupervised neural network comprising an input layer, a hidden layer and an output layer; the network obtains a limited number of characteristic representations through self-learning of the original characteristics, and achieves the purpose of input reconstruction by using the characteristic representations;
(2.2) parameter learning from the coding network is divided into two processes: an encoding process and a decoding process; in the encoding process, firstly, the hidden layer lambda is added1(x) Self-learning is carried out, wherein1(x) The calculation formula of (a) is as follows:
λ1(x)＝w(Y1x+c1) (2)
wherein, Y1To code the matrix, c1To encode the bias vector, w (-) is a tan h function;
(2.3) self-coding network structure with M hidden layers, wherein the decoding process realizes the representation of the hidden layers by lambda through determining a decoding matrix1(x) Decoding into reconstructed data lambda2(x) Process of reconstructing data λ2(x) Is output by the formula
λ2(x)＝f(Y2x+c2) (3)
Wherein, Y2To decode the matrix, c2To decode the offset vector, f (-) is a tan h function;
the self-coding network learning process implements the optimization process of the network parameters by minimizing the mean square error cost function as shown below, i.e.
Figure FDA0002666016560000021
Thus, the optimal set of parameters for a self-coding network can be translated into solving the following optimization problem
Figure FDA0002666016560000022
The optimization problem is generally solved by a BP neural network algorithm; stacking a plurality of self-coding networks on the basis to obtain a machine learning model for mining the implicit functional relation of the data;
(2.4) Structure and training method of Stack self-coding learning network with j hidden layers
In training set TTestThe final output of the model at the jth hidden layer can be expressed as:
Figure FDA0002666016560000023
wherein the content of the first and second substances,
Figure FDA0002666016560000024
and
Figure FDA0002666016560000025
the coding matrix and the coding bias vector of the jth self-coding network are respectively, and w (·) is a tan h function.
6. The method of claim 3, wherein the step (3) comprises:
(3.1) the stacked self-coding network pre-training trains the first layer of MLM as a self-coding network, minimizes equation (4) with training data as input, and initializes χ ═ 2;
(3.2) when training the chi-layer, will
Figure FDA0002666016560000026
Minimizing equation (4) as an input;
(3.3) making χ ═ χ +1, and iterating (3.2) steps; stopping iteration when x is larger than j, and turning to (3.4);
(3.4) the final output of the network is
Figure FDA0002666016560000027
Inputting the learning model;
(3.5) in the part for mining the input and output mathematical mapping relation, optimizing the output weight theta by adopting a least square method; when all the self-coding network parameters
Figure FDA0002666016560000028
All determined, inputting the data set
Figure FDA0002666016560000031
The corresponding hidden layer represents lambdaj(x(k)) Is known; in the air flow texture prediction problem, it is always desirable to relate to x(k)Is estimated value of
Figure FDA0002666016560000032
Can accurately approximate to the actual output y(k)It is written as the following equation:
Figure FDA0002666016560000033
wherein the content of the first and second substances,
Figure FDA0002666016560000034
namely:
λj(x)θ＝y (7)
the above formula describes the mathematical mapping relationship between the input parameters including building geometry, ventilation parameters and the output parameters including airflow organization node data, wherein the implicit function lambdaj(x) This can be learned from the following formula:
Figure FDA0002666016560000035
Figure FDA0002666016560000036
Figure FDA0002666016560000037
according to the matrix theory, the optimal output weight vector θ is a least-squares solution of the equation, i.e.:
θ＝λj(x)↑y (9)
equations (7) - (9) are used to directly learn the node airflow field data that matches the output given the input building geometry and ventilation parameters.
CN202010919008.3A 2020-09-04 2020-09-04 Intelligent human residential environment airflow organization optimization method Active CN112149364B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010919008.3A CN112149364B (en) 2020-09-04 2020-09-04 Intelligent human residential environment airflow organization optimization method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010919008.3A CN112149364B (en) 2020-09-04 2020-09-04 Intelligent human residential environment airflow organization optimization method

Publications (2)

Publication Number Publication Date
CN112149364A true CN112149364A (en) 2020-12-29
CN112149364B CN112149364B (en) 2022-11-29

Family

ID=73889727

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010919008.3A Active CN112149364B (en) 2020-09-04 2020-09-04 Intelligent human residential environment airflow organization optimization method

Country Status (1)

Country Link
CN (1) CN112149364B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113158593A (en) * 2021-04-13 2021-07-23 中南建筑设计院股份有限公司 Intelligent design method of hospital ward ventilation system based on artificial intelligence
CN113204827A (en) * 2021-05-31 2021-08-03 升维科技有限公司 Building energy-saving design method and system based on BIM
WO2022210516A1 (en) * 2021-03-31 2022-10-06 ダイキン工業株式会社 Visualization system, and program
US12025330B2 (en) 2021-03-31 2024-07-02 Daikin Industries, Ltd. Visualization system for target area of air conditioner

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103049612A (en) * 2012-12-24 2013-04-17 江苏大学 Building indoor environment optimization method based on model order reduction technology
CN104633856A (en) * 2015-01-27 2015-05-20 天津大学 Method for controlling artificial environment by combining CFD numerical simulation and BP neural network
US20180284752A1 (en) * 2016-05-09 2018-10-04 StrongForce IoT Portfolio 2016, LLC Methods and systems for industrial internet of things data collection in downstream oil and gas environment
CN111191316A (en) * 2020-01-06 2020-05-22 沈阳建筑大学 Response surface-based building natural ventilation performance optimization model and optimization method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103049612A (en) * 2012-12-24 2013-04-17 江苏大学 Building indoor environment optimization method based on model order reduction technology
CN104633856A (en) * 2015-01-27 2015-05-20 天津大学 Method for controlling artificial environment by combining CFD numerical simulation and BP neural network
US20180284752A1 (en) * 2016-05-09 2018-10-04 StrongForce IoT Portfolio 2016, LLC Methods and systems for industrial internet of things data collection in downstream oil and gas environment
CN111191316A (en) * 2020-01-06 2020-05-22 沈阳建筑大学 Response surface-based building natural ventilation performance optimization model and optimization method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
HUANG JIANCA等: "3D Topography Measurement and Completion Method of Blast Furnace Burden Surface Using High-Temperature Industrial Endoscope", 《IEEE》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022210516A1 (en) * 2021-03-31 2022-10-06 ダイキン工業株式会社 Visualization system, and program
JP2022159090A (en) * 2021-03-31 2022-10-17 ダイキン工業株式会社 Visualization system, and program
JP7161138B2 (en) 2021-03-31 2022-10-26 ダイキン工業株式会社 Visualization system and program
US12025330B2 (en) 2021-03-31 2024-07-02 Daikin Industries, Ltd. Visualization system for target area of air conditioner
CN113158593A (en) * 2021-04-13 2021-07-23 中南建筑设计院股份有限公司 Intelligent design method of hospital ward ventilation system based on artificial intelligence
CN113158593B (en) * 2021-04-13 2023-02-07 中南建筑设计院股份有限公司 Intelligent design method of hospital ward ventilation system based on artificial intelligence
CN113204827A (en) * 2021-05-31 2021-08-03 升维科技有限公司 Building energy-saving design method and system based on BIM
CN113204827B (en) * 2021-05-31 2023-09-05 山东华烨规划建筑设计有限公司 Building energy-saving design method and system based on BIM

Also Published As

Publication number Publication date
CN112149364B (en) 2022-11-29

Similar Documents

Publication Publication Date Title
CN112149364B (en) Intelligent human residential environment airflow organization optimization method
Xu et al. Prediction and optimization of heating and cooling loads in a residential building based on multi-layer perceptron neural network and different optimization algorithms
Amasyali et al. Machine learning for occupant-behavior-sensitive cooling energy consumption prediction in office buildings
Li et al. An ANN-based fast building energy consumption prediction method for complex architectural form at the early design stage
Hosamo et al. Multiobjective optimization of building energy consumption and thermal comfort based on integrated BIM framework with machine learning-NSGA II
Razmi et al. PCA-ANN integrated NSGA-III framework for dormitory building design optimization: Energy efficiency, daylight, and thermal comfort
Shen et al. Rapid multi-objective optimization with multi-year future weather condition and decision-making support for building retrofit
Wu et al. Intelligent optimization framework of near zero energy consumption building performance based on a hybrid machine learning algorithm
Gan et al. Physics-based, data-driven approach for predicting natural ventilation of residential high-rise buildings
Zhao et al. Construction cost prediction based on genetic algorithm and BIM
Srinivasan et al. Urban building energy CPS (UBE-CPS): real-time demand response using digital twin
Dong et al. Scenario prediction and decoupling analysis of carbon emission in Jiangsu Province, China
Wang et al. BIM-based automated design for HVAC system of office buildings—An experimental study
CN116187175A (en) Power transformation and distribution station neural network identification method based on electrical industry knowledge graph
CN114811713A (en) Secondary inter-user balanced heat supply regulation and control method based on hybrid deep learning
Fan Research on deep learning energy consumption prediction based on generating confrontation network
Liu et al. Building information modelling-enabled multi-objective optimization for energy consumption parametric analysis in green buildings design using hybrid machine learning algorithms
Peng et al. Energy consumption optimization for heating, ventilation and air conditioning systems based on deep reinforcement learning
CN116415713A (en) Building energy consumption prediction method based on E+ and artificial intelligence
Allam et al. Metaheuristic optimization of daylighting and energy performances in office spaces
May-Ostendorp Offline model predictive control of mixed mode buildings for near-optimal supervisory control strategy development
Gui et al. Visual comfort generative design framework based on parametric network in underground space
Li Innovation of Interior Design Automation Tools Based on CAD Algorithm
Pektas Parametric Design as a Tool/As a Goal: Shifting Focus from Form to Function
Qiao et al. Simulation and Application of Low Carbon Heating Buildings Based on Genetic Algorithm Optimization PID

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