A limited sensor deployment strategy and intelligent ventilation system for industrial buildings
By adopting dimensionality reduction linear databases and unsupervised machine learning fuzzy clustering algorithms in industrial buildings, the rapidity and accuracy of pollutant prediction in ventilation systems are solved, and more efficient pollutant removal and energy conservation are achieved.
Patent Information
- Application Number
- CN202011279412.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-11-16
- Publication Date
- 2024-12-10
- Estimated Expiration
- 2040-11-16
AI Technical Summary
The ventilation systems of existing industrial buildings consume more energy and are difficult to achieve rapid and accurate pollutant prediction, which affects the operating efficiency of the intelligent ventilation system, and the unreasonable arrangement of sensors affects the pollutant removal effect.
Using the processing method based on dimensionality reduction linear database and unsupervised machine learning fuzzy clustering algorithm, a basic database is built and reasonable monitoring point deployment scheme is determined through clustering analysis, and using machine learning to predict the optimal monitoring point location and number to achieve pollution Fast and accurate prediction and online control of substance concentration.
It effectively reduces the storage and time costs of numerical calculations, realizes rapid and accurate prediction and online control of industrial building pollutants, improves the operating efficiency of intelligent ventilation systems, and reduces energy consumption.
Smart Images
Figure CN112417553B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to building pollutant concentration monitoring and ventilation system control, in particular to a limited sensor arrangement strategy and an intelligent ventilation system for industrial buildings. Background Art
[0002] Currently, due to the special nature of the processes in the metallurgical, steel and other industrial fields, the concentration of pollutants in industrial buildings is relatively high, and operators exposed to poor indoor air quality for a long time will increase the risk of respiratory and cardiovascular diseases. Therefore, how to efficiently remove pollutants in industrial plants has become a top priority in solving indoor air hygiene problems.
[0003] Conventional ventilation systems in industrial buildings often operate under specified ventilation conditions, and such specified conditions are often greater than the removal requirements of pollutants in industrial buildings, consuming more energy at the expense of creating safe, healthy, and energy-saving industrial buildings. With the development of computer technology and the advent of the machine learning era, it is of great significance to develop more advanced intelligent ventilation systems that can better meet real-time needs. In intelligent ventilation systems, the rapid prediction of pollutants in industrial buildings is an important part of the system, and it is still facing huge challenges. Reasonable arrangement of sensor monitoring points in industrial buildings and rapid and accurate prediction of pollutants directly affect the operating efficiency of intelligent ventilation systems. Therefore, the "rapidity" and "accuracy" of intelligent ventilation systems for indoor pollutants need to be solved urgently. Summary of the invention
[0004] The purpose of the present invention is to provide a limited sensor layout strategy and an intelligent ventilation system for industrial buildings. Based on the processing method of the dimensionality reduction linear database, the storage cost and time cost of numerical calculations are effectively reduced; based on the unsupervised machine learning fuzzy clustering algorithm, pollutants are clustered and a reasonable monitoring point deployment plan is determined, which is universal; according to information such as the location of pollutants, the optimal monitoring point location and number of deployment plans are quickly predicted, providing faster and more accurate pollution information distribution for further online control of pollutants in the intelligent ventilation system.
[0005] The purpose of the present invention can be achieved through the following technical solutions:
[0006] A limited sensor deployment strategy for an industrial building, the sensor deployment strategy comprising the following steps:
[0007] S1: Basic database construction
[0008] The basic database includes all known air change times and all combinations of individual pollution source locations under this ventilation mode;
[0009] S2: Database expansion and dimensionality reduction discretization preprocessing
[0010] The numerical simulation open source software was used to obtain the pollutant distribution simulation results of all ventilation cases and used for the construction of the basic database;
[0011] The linear superposition principle is used to expand the database. When there are multiple pollution sources, the pollutant distribution results generated are equal to the superposition of the pollutant distribution results generated when these pollution sources act alone; the dimensionality reduction discretization method is used for data preprocessing.
[0012] S3: Database training
[0013] Based on the new database obtained by dimensionality reduction, deep machine learning is performed according to specific ventilation case variables to explore the small cube Ω i The functional relationship between each variable or data is used for the next step of prediction or restoration.
[0014] S4: The limited sensor monitoring strategy will be obtained through cluster analysis, and the monitoring data will be used as the input of the artificial neural network
[0015] Given a specific ventilation case variable, the functional relationship between pollutant concentrations at all locations in the trained space is obtained; using this functional relationship, taking into account a specific number of ventilation changes, the point data is used to predict or restore the data at any location in the space; the point data is represented by the data from the actual monitoring point.
[0016] S5: Rapid prediction of pollutant concentration in industrial buildings
[0017] The monitoring point deployment scheme obtained by cluster analysis is used as the input layer of the artificial neural network, and the dimensionality-reduced discrete prediction value at any position in space is obtained.
[0018] S6: Building an intelligent ventilation system
[0019] Furthermore, the dimensionality reduction discretization method in S2 is: compressing multiple data nodes into one data node through volume averaging calculation.
[0020] Furthermore, the number of monitoring points in S4 is 3-5.
[0021] Furthermore, in S5, if the error between the predicted value of the pollutant concentration obtained by dimensionality reduction and discreteness at any position in space and the value of the pollutant concentration obtained by direct CFD simulation is less than 10%, it is considered that the deployment of the monitoring point is appropriate.
[0022] An intelligent ventilation system for an industrial building, the ventilation system is located on the building, an operation area and a sensor are arranged in the building, the ventilation system comprises an air conditioning unit, an air supply unit, a return air unit and a fresh air unit are connected to the air conditioning unit, the air supply unit comprises an air supply duct, one end of the air supply duct is connected to the air conditioning unit, the other end extends to the building where a slit-type air supply outlet is provided, a fan assembly is arranged between the air supply duct and the air conditioning unit, and the return air unit comprises a return air duct, one end of the return air duct is connected to the air conditioning unit, the other end extends to the building where a slit-type return air outlet is provided.
[0023] Furthermore, the air supply unit and the air return unit are respectively located on two sides of the air conditioning unit.
[0024] Furthermore, the slit-type air supply outlet is located at the top of the building.
[0025] Furthermore, the slit-type return air outlet is located on one side of the building, and the slit-type supply air outlet and the slit-type return air outlet form a top-supply and side-return air supply form.
[0026] Furthermore, the air conditioning unit forms a circulation loop with the air supply unit, the return air unit and the fresh air unit to realize air circulation in the operation area.
[0027] Beneficial effects of the present invention:
[0028] 1. The processing method of the present invention is based on the dimension-reduced linear database, which effectively reduces the storage cost and time cost of numerical calculation;
[0029] 2. The present invention is based on unsupervised machine learning fuzzy clustering algorithm to achieve clustering of pollutants and determine a reasonable monitoring point deployment plan, and it has universal applicability;
[0030] 3. The present invention can quickly predict the optimal monitoring point locations and number deployment plans based on information such as pollutant locations, and provide faster and more accurate pollution information distribution for further online control of pollutants in the intelligent ventilation system. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] The present invention will be further described below in conjunction with the accompanying drawings.
[0032] Figure 1 It is a flow chart of the method for rapid prediction of pollutant concentration of the present invention;
[0033] Figure 2 It is a diagram showing the principle of database training of the present invention;
[0034] Figure 3 It is a block diagram of the pollution source location and the dimensionality reduction discretization method of the present invention;
[0035] Figure 4 It is a schematic diagram of the dimension reduction model of the present invention. DETAILED DESCRIPTION
[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0037] A limited sensor deployment strategy for industrial buildings, the sensor deployment strategy includes the following steps:
[0038] S1: Basic database construction
[0039] The basic database includes all known air changes (air volume) under the ventilation mode and all combinations of the locations of all individual pollution sources.
[0040] S2: Database expansion and dimensionality reduction discretization preprocessing
[0041] The pollutant distribution simulation results of all ventilation cases are obtained by using open source numerical simulation software (such as OpenFOAM, etc.) and used for basic database construction. Then the linear superposition principle is used to expand the database, that is, when there are multiple pollution sources (locations), the pollutant distribution results generated by them are equal to the superposition of the pollutant distribution results generated by these pollution sources when they act alone; then a dimensionality reduction discretization method is used for data preprocessing, and many data nodes are compressed into one data node through volume average calculation to reduce database costs, such as Figure 4 shown.
[0042] S3: Database training
[0043] Based on the new database obtained by dimensionality reduction, deep machine learning can be performed according to specific ventilation case variables, that is, exploring the small cube Ω i The functional relationship between various variables or data within a space can be used. For example, based on a specific pollution source location and ventilation frequency, machine learning methods can be used to train to obtain a function curve or equation between the pollutant concentration data at all locations in the space for the next step of prediction or restoration.
[0044] S4: The limited sensor monitoring strategy will be obtained through cluster analysis, and the monitoring data will be used as the input of the artificial neural network
[0045] Given specific ventilation case variables (such as the location of the pollution source and the number of ventilation changes), the functional relationship between the pollutant concentrations at all locations in the trained space can be obtained. Then, using this functional relationship, considering the specific number of ventilation changes, the data of a very small number of points (3-5) are used to predict or restore the data at any location in the space (assuming that all data have been discretized). The data at this point can be represented by the data of the actual monitoring point.
[0046] S5: Rapid prediction of pollutant concentration in industrial buildings
[0047] The monitoring point deployment scheme obtained by cluster analysis is used as the input layer of the artificial neural network, and the predicted value of the dimensionality reduction discrete at any position in the space is obtained. If the error between the predicted value of the dimensionality reduction discrete pollutant concentration at any position in the space and the dimensionality reduction discrete pollutant concentration value obtained by direct CFD simulation is less than 10%, then the deployment of the monitoring point can be considered appropriate.
[0048] S6: Building an intelligent ventilation system
[0049] An intelligent ventilation system for an industrial building, the ventilation system is located on the building, an operation area and a sensor are arranged in the building, the ventilation system comprises an air conditioning unit, an air supply unit, a return air unit and a fresh air unit are connected to the air conditioning unit, the air supply unit and the return air unit are respectively located on both sides of the air conditioning unit, the air supply unit comprises an air supply duct, one end of the air supply duct is connected to the air conditioning unit, the other end extends to the building and is provided with a slit-type air supply outlet, the slit-type air supply outlet is located at the top of the building, a fan assembly is arranged inside the air conditioning unit, the return air unit comprises a return air duct, one end of the return air duct is connected to the air conditioning unit, the other end extends to the building and is provided with a slit-type return air outlet, the slit-type return air outlet is located on one side of the building, the slit-type supply air outlet and the slit-type return air outlet form a top-supply and side-return air supply form, the air conditioning unit and the air supply unit, the return air unit and the fresh air unit form a circulation loop to realize air circulation in the operation area.
[0050] Example
[0051] The building geometry is 20m (length) × 15m (width) × 10m (height). Considering 6 ventilation rates, 3 pollution sources in different operation areas are set up indoors, such as Figure 3 As shown, the pollution sources in the three different operating areas are A, B and C, so the basic database includes 6×3=18 ventilation cases in total.
[0052] The numerical simulation open source software FLUENT was used to obtain the pollutant distribution simulation results of all ventilation cases and used for the construction of the basic database;
[0053] The database is expanded using the linear superposition principle, that is, when there are multiple pollution sources (locations), the pollutant distribution results they produce are equal to the superposition of the pollutant distribution results produced when these pollution sources act alone.
[0054] The data was preprocessed using a dimensionality reduction discretization method; the grid of the ventilation structure was evenly divided into 3×3×3=27 small cubic blocks to achieve database reconstruction, and it was ensured to be within an acceptable engineering error range (≤10%).
[0055] Based on the new database obtained by dimensionality reduction, deep machine learning is performed according to specific ventilation case variables to explore the small cube Ω i The functional relationship between each variable or data in the space is based on a specific pollution source location and ventilation frequency. The function curve or equation between the pollutant concentration data at all locations in the space is obtained by training using machine learning methods for the next step of prediction or restoration.
[0056] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0057] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.
Claims
1. A limited sensor deployment strategy for industrial buildings, characterized in that: The sensor deployment strategy includes the following steps: S1: Basic database construction The basic database includes all known air change times and all combinations of individual pollution source locations under this ventilation mode; S2: Database expansion and dimensionality reduction discretization preprocessing The numerical simulation open source software was used to obtain the pollutant distribution simulation results of all ventilation cases and used for the construction of the basic database; The linear superposition principle is used to expand the database. When there are multiple pollution sources, the pollutant distribution results generated are equal to the superposition of the pollutant distribution results generated when these pollution sources act alone. The dimensionality reduction discretization method is used for data preprocessing. S3: Database training Based on the new database obtained by dimensionality reduction, deep machine learning is performed according to specific ventilation case variables to explore the small cube Ω i The functional relationship between each variable or data is used for the next step of prediction or restoration; S4: The limited sensor monitoring strategy will be obtained through cluster analysis, and the monitoring data will be used as the input of the artificial neural network Given a specific ventilation case variable, the functional relationship between the pollutant concentrations at all locations in the space after training is obtained; using this functional relationship, taking into account a specific ventilation rate, the point data is used to predict or restore the data at any location in the space; the point data is represented by the data of the actual monitoring point; S5: Rapid prediction of pollutant concentration in industrial buildings The monitoring point deployment scheme obtained by cluster analysis is used as the input layer of the artificial neural network, and the predicted value of dimension reduction and discreteness at any position in space is obtained; S6: Build intelligent ventilation system; The dimensionality reduction discretization method in S2 is: compressing multiple data nodes into one data node through volume average calculation; The number of monitoring points in S4 is 3-5; In S5, if the error between the predicted value of the pollutant concentration obtained by dimensionality reduction and discreteness at any position in space and the value of the pollutant concentration obtained by direct CFD simulation is less than 10%, it is considered that the deployment of the monitoring point is appropriate.
2. The intelligent ventilation system in the limited sensor deployment strategy according to claim 1, characterized in that: The ventilation system is located at the upper part of the building, and an operation area and a sensor are arranged in the building. The ventilation system includes an air-conditioning unit, and an air supply unit, a return air unit and a fresh air unit are connected to the air-conditioning unit. The air-supply unit includes an air supply duct, one end of the air supply duct is connected to the air-conditioning unit, and the other end extends to the building where a slit-type air supply outlet is provided. The air-conditioning unit is provided with a fan assembly, and the return air unit includes a return air duct, one end of the return air duct is connected to the air-conditioning unit, and the other end extends to the building where a slit-type return air outlet is provided.
3. The intelligent ventilation system according to claim 2, characterized in that: The air supply unit and the air return unit are respectively located on both sides of the air conditioning unit.
4. The intelligent ventilation system according to claim 2, characterized in that: The slit-type air supply outlet is located at the top of the building.
5. The intelligent ventilation system according to claim 4, characterized in that: The slit-type return air outlet is located at one side of the building, and the slit-type supply air outlet and the slit-type return air outlet form a top-supply and side-return air supply form.
6. The intelligent ventilation system according to claim 2, characterized in that: The air conditioning unit, the air supply unit, the return air unit and the fresh air unit form a circulation loop to realize air circulation in the operation area.