Ventilation system performance prediction and optimization method based on machine learning

Through the performance prediction and optimization method of ventilation system based on machine learning, the problem of large-scale pollutant concentration prediction errors caused by complex factors in the existing technology is solved, and the rapid and accurate ventilation system optimization design is achieved, and the operation control is simplified.

CN118551512BActive Publication Date: 2024-12-03HUBEI CONSTR ENG TECH CO LTD +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410640771.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-22
Publication Date
2024-12-03
Estimated Expiration
2044-05-22

AI Technical Summary

Technical Problem

The ventilation control factors for existing pollutant gas emissions are too numerous and complex, resulting in large errors in predicting pollutant concentrations. The existing methods such as experiments and CFD methods have long cycles, high costs and difficult to implement.

Method used

Using machine learning-based ventilation system performance prediction and optimization methods, we will build a sample library by determining the characteristic parameters and design goals of the ventilation system, selecting a variety of machine learning models for training and testing, and selecting the model with the best comprehensive performance as the proxy model, combining it with The CFD method performs correction of the optimal design scheme and iterates repeatedly until the results are consistent.

Benefits of technology

It realizes the optimal solution for quickly and accurately searching the parameter space of the ventilation system, ensuring the accuracy of prediction and optimization results, and simplifying the operation control and optimization design of the ventilation system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118551512B_ABST
    Figure CN118551512B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of ventilation devices, and discloses a method for predicting and optimizing the performance of a ventilation system based on machine learning, comprising the following steps: determining characteristic parameters and design objectives of a ventilation system; extracting data of characteristic parameters and design objectives of a ventilation system; constructing a sample library for a mapping set of parameter features and outputs; standardizing the sample set; dividing the standardized sample set into a training set and a test set; selecting a plurality of machine learning models of different algorithms; training and testing the machine learning model based on the sample set; comparing and evaluating machine learning models of different algorithms, and selecting the machine learning model with the best comprehensive performance as a proxy model. The present invention, based on Taguchi orthogonal experimental design, extracts data from CFD calculation results and experimental results of ventilation systems with different characteristics and different ventilation modes, constructs a sample library, and can serve as a support for subsequent work to perform an iterative process of prediction and optimization.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of ventilation devices, and in particular to a ventilation system performance prediction and optimization method based on machine learning. Background Art

[0002] Ventilation is an important means of controlling air pollutants in buildings. However, ventilation systems are affected by many factors and produce different ventilation effects. Therefore, when indoor pollutant concentrations are at dangerous levels, appropriate air supply and exhaust patterns need to be formulated to keep them within appropriate limits.

[0003] After searching, the Chinese patent number CN103109137B discloses a control method for a ventilation system for ventilation of an enclosure. The ventilation system includes two or more ventilation units, each of which can be selectively and repeatedly controlled to establish an air flow to and / or from the enclosure. The method includes the following steps: determining information related to the operating conditions of one or more of the two or more ventilation units by measuring the air flow through the ventilation unit or the pressure difference across the ventilation unit, generating operation control data for one or more of the ventilation units from information related to the operating conditions determined for at least one other ventilation unit in the ventilation units, and controlling the operation of one or more of the two or more ventilation units by applying the generated operation control data. The present invention also relates to a ventilation system for ventilating an enclosure.

[0004] After searching, the Chinese patent number CN104315680A discloses a ventilation system, which includes a fresh air host and one or more intelligent detection and exhaust devices, the fresh air host and the intelligent detection and exhaust devices are both provided with wireless communication devices and data exchange is performed through the wireless communication devices, the fresh air host is provided with a host processor and a display controller connected to the host processor, and the wireless communication device on the fresh air host is connected to the host processor, and the fresh air host also includes a wireless transmission device connected to the host processor; the ventilation system also includes a wireless routing device and an intelligent mobile client, and the wireless routing device is configured to perform data exchange with the wireless transmission device and the intelligent mobile client through a wireless protocol. According to the ventilation system provided by the present invention, it is possible to realize intelligent detection of indoor air quality and realize automatic air exchange and optimization of indoor air quality according to the detection results, and it is possible to understand the indoor air quality in real time and remotely control air exchange and optimization of indoor air quality.

[0005] However, there are too many factors to consider in the existing ventilation control of pollutant gas emissions, and multiple factors will affect each other, so that the distribution of environmental control factors is uneven and complex. These complex environmental conditions may introduce large errors in the prediction of pollutant concentrations. Considering the prediction of pollutant distribution, it is usually implemented by experiments and CFD methods, but these two methods always have long cycles and high costs, and are even difficult to carry out because of their complex structures. Based on this, the present invention designs a ventilation system performance prediction and optimization method based on machine learning to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to provide a ventilation system performance prediction and optimization method based on machine learning, which solves the problem that experiments and methods in the background technology are difficult to implement.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] A ventilation system performance prediction and optimization method based on machine learning is required to provide a ventilation system based on machine learning, the ventilation system comprising a speed control mechanism, the front and back sides of the speed control mechanism are both equipped with adjustment mechanisms, and the side of the adjustment mechanism away from the speed control mechanism is equipped with a lifting mechanism;

[0009] The ventilation system performance prediction and optimization method based on machine learning includes the following steps:

[0010] Step S1, determining characteristic parameters and design objectives of the ventilation system;

[0011] Step S2, extracting data of ventilation system characteristic parameters and design objectives;

[0012] Step S3, constructing a sample library for the mapping set of parameter features and outputs;

[0013] Step S4, standardizing the sample set;

[0014] Step S5, dividing the standardized sample set into a training set and a test set;

[0015] Step S6: Select multiple machine learning models with different algorithms;

[0016] Step S7: training and testing the machine learning model based on the sample set;

[0017] Step S8: Compare and evaluate machine learning models of different algorithms, and select the machine learning model with the best comprehensive performance as the proxy model;

[0018] Step S9, determining the constraints and optimization targets of the characteristic parameters, and obtaining the optimal design solution;

[0019] Step S10: Calculate the optimal design solution based on the CFD method. If the calculation result is inconsistent with the calculation result of the proxy model, return to step 7 to re-perform the machine learning training-optimization process, and iterate repeatedly until the calculation results are consistent.

[0020] Preferably, in step S1, geometric characteristic parameters, operation characteristic parameters and pollutant transmission characteristic parameters of the ventilation system are used as design parameters, including air supply port size, exhaust port size, distance between supply and exhaust port centers, air supply port height, exhaust port height, air supply port temperature, air supply port speed, pollution source type and pollution source emission rate, a total of 9 design parameters, and the pollutant removal efficiency of the ventilation system is used as the design target.

[0021] Preferably, in step S2, the experiment design is performed by an orthogonal method, numerical simulation calculations are performed according to the experimental design scheme in combination with the CFD method, and experimental verification is performed on the numerical simulation calculations, data is extracted from the CFD calculation results and experiments, and a mapping between parameter characteristics and design objectives is extracted.

[0022] Preferably, in step S4, for unrecognizable features in the sample library, they are converted into sparse matrices for processing by a hot encoder method, and data normalization is achieved by a feature scaling method.

[0023] Preferably, in step S5, the ratio of the training set to the test set is 8:2.

[0024] Preferably, in step S6, the specific algorithms of the machine learning model include back propagation neural network algorithm (BPNN), support vector machine algorithm (SVM), random forest algorithm (RF), extreme gradient boosting algorithm (XGBoost), adaptive boosting algorithm (AdaBoost) and gradient boosting algorithm (GBoost).

[0025] Preferably, in step S8, the comprehensive evaluation index includes the correlation coefficient R2, the root mean square error (RMSE) and the mean absolute relative error (MAPE).

[0026] Preferably, the speed control mechanism includes a speed control frame, an electric hoist is installed on the top of the inner cavity of the speed control frame, a moving frame is installed on the bottom of the electric hoist, the inner cavity of the moving frame is provided with through holes, and the through holes are distributed in a plurality of matrix arrays, the inner wall of the speed control frame is provided with a first slide groove, both sides of the moving frame are fixedly connected with first sliders, the first slider is slidably connected in the inner cavity of the first slide groove, a first motor is installed on the top of the moving frame, the bottom of the output shaft of the first motor is fixedly connected with a first screw rod, and the outer ring of the first screw rod is threadedly connected with a moving plate.

[0027] Preferably, the adjustment mechanism includes an adjustment frame, which is a hollow structure. An electric push rod is fixedly connected to the inner wall of the adjustment frame, and an adjustment plate is fixedly connected to one side of the electric push rod. Grooves are provided at the top and bottom of the inner cavity of the adjustment frame, and the adjustment plate can move left and right in the inner cavity of the groove.

[0028] Preferably, the lifting mechanism includes a base plate, a second motor is installed on the top of the base plate, a connecting plate is installed on the top of the base plate, a second screw rod is rotatably connected between the second motor and the connecting plate, the outer ring of the second screw rod is threadedly connected to the first moving block, the top of the first moving block is hinged with a rotating plate, the top of the rotating plate is hinged with the second moving block, the top of the second moving block is installed with a lifting plate, the top of the lifting plate is installed with a fixed seat, the top of the fixed seat is installed with a transport tube, and one end of the transport tube is fixedly connected to the adjustment frame.

[0029] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0030] 1. The present invention, based on Taguchi orthogonal experimental design, extracts data from CFD calculation results and experimental results of ventilation systems with different ventilation modes with different characteristics, constructs a sample library, and can serve as a support for subsequent work. By comparing and evaluating machine learning models, predictors with better performance are screened out as proxy models to perform an iterative process of prediction-optimization.

[0031] 2. The present invention predicts the design target sample based on a smaller sample set space and establishes a machine learning model suitable for a small sample library. Combined with an intelligent algorithm, it can quickly and accurately search for the optimal solution in the parameter space for the working condition of the input feature parameter set, and correct the optimal solution according to the CFD method to ensure the accuracy of the prediction and optimization results, and can quickly realize the operation control and optimization design of the ventilation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flow chart of the ventilation system performance prediction and optimization method of the present invention;

[0033] Figure 2 It is a structural schematic diagram of the ventilation system of the present invention;

[0034] Figure 3 It is a structural schematic diagram of the speed control structure of the present invention;

[0035] Figure 4 For the present invention Figure 3 A magnified image of point A;

[0036] Figure 5 It is a rear view of the speed control structure of the present invention;

[0037] Figure 6 is a cross-sectional view of the adjustment mechanism of the present invention;

[0038] Figure 7 It is a structural schematic diagram of the lifting mechanism of the present invention;

[0039] Figure 8 It is a side view of the lifting mechanism of the present invention.

[0040] Wherein: 1. speed control mechanism; 2. adjustment mechanism; 3. lifting mechanism; 101. speed control frame; 102. electric hoist; 103. moving frame; 104. through hole; 105. first slide groove; 106. first slider; 107. first motor; 108. first screw rod; 109. moving plate; 201. adjustment frame; 202. electric push rod; 203. adjustment plate; 204. groove; 301. bottom plate; 302. second motor; 303. connecting plate; 304. second screw rod; 305. first moving block; 306. rotating plate; 307. second moving block; 308. lifting plate; 309. fixed seat; 310. transport tube. DETAILED DESCRIPTION

[0041] 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.

[0042] Embodiment 1;

[0043] See also Figure 1-Figure 8 In an embodiment of the present invention, a ventilation system performance prediction and optimization method based on machine learning needs to provide a ventilation system based on machine learning, the ventilation system includes a speed control mechanism 1, an adjustment mechanism 2 is installed on the front and back of the speed control mechanism 1, and a lifting mechanism 3 is installed on the side of the adjustment mechanism 2 away from the speed control mechanism 1;

[0044] The ventilation system performance prediction and optimization method based on machine learning includes the following steps:

[0045] Step S1, determining characteristic parameters and design objectives of the ventilation system;

[0046] Step S2, extracting data of ventilation system characteristic parameters and design objectives;

[0047] Step S3, constructing a sample library for the mapping set of parameter features and outputs;

[0048] Step S4, standardizing the sample set;

[0049] Step S5, dividing the standardized sample set into a training set and a test set;

[0050] Step S6: Select multiple machine learning models with different algorithms;

[0051] Step S7: training and testing the machine learning model based on the sample set;

[0052] Step S8: Compare and evaluate machine learning models of different algorithms, and select the machine learning model with the best comprehensive performance as the proxy model;

[0053] Step S9, determining the constraints and optimization targets of the characteristic parameters, and obtaining the optimal design solution;

[0054] Step S10: Calculate the optimal design solution based on the CFD method. If the calculation result is inconsistent with the calculation result of the proxy model, return to step 7 to re-perform the machine learning training-optimization process, and iterate repeatedly until the calculation results are consistent.

[0055] like Figure 1 As shown, in step S1, the geometric characteristic parameters, operation characteristic parameters and pollutant transmission characteristic parameters of the ventilation system are used as design parameters, including the air supply port size, the exhaust port size, the distance between the centers of the supply and exhaust ports, the air supply port height, the exhaust port height, the air supply port temperature, the air supply port speed, the type of pollution source and the pollution source emission rate, a total of 9 design parameters, and the pollutant removal efficiency of the ventilation system is used as the design target.

[0056] like Figure 1 As shown, in step S2, the orthogonal method is used to design the experiment, and the numerical simulation calculation is performed according to the experimental design scheme combined with the CFD method, and the experimental verification is performed, and data is extracted from the CFD calculation results and experiments, and the mapping between parameter characteristics and design goals is extracted.

[0057] like Figure 1 As shown, in step S4, for the unrecognizable features in the sample library, they are converted into sparse matrices for processing through the hot encoder method, and data normalization is achieved through the feature scaling method.

[0058] like Figure 1 As shown, in step S5, the ratio of the training set to the test set is 8:2.

[0059] like Figure 1 As shown, in step S6, the specific algorithms of the machine learning model include back propagation neural network algorithm BPNN, support vector machine algorithm SVM, random forest algorithm RF, extreme gradient boosting algorithm XGBoost, adaptive boosting algorithm AdaBoost and gradient boosting algorithm GBoost.

[0060] In step S8, the comprehensive evaluation index includes the correlation coefficient R 2 , root mean square error RMSE and mean absolute relative error MAPE.

[0061] The working principle of the embodiment of the present invention is: standardize the sample set, specifically: for unrecognizable features such as ventilation "mode", convert them into sparse matrices for processing through the hot encoder method; data normalization is achieved through feature scaling method.

[0062] Based on the Bayesian optimization method, hyperparameters are tuned using 10-fold cross-validation. The tuned machine learning model is trained and tested using the training set and test set in the sample set.

[0063] Embodiment 2;

[0064] In the embodiment of the present invention, the speed control mechanism 1 includes a speed control frame 101, an electric hoist 102 is installed on the top of the inner cavity of the speed control frame 101, a moving frame 103 is installed on the bottom of the electric hoist 102, the inner cavity of the moving frame 103 is provided with a through hole 104, and the through holes 104 are distributed in a plurality of matrix arrays. A first slide groove 105 is provided on the inner wall of the speed control frame 101, and first sliders 106 are fixedly connected to both sides of the moving frame 103, and the first slider 106 is slidably connected in the inner cavity of the first slide groove 105. A first motor 107 is installed on the top of the moving frame 103, and a first screw rod 108 is fixedly connected to the bottom of the output shaft of the first motor 107, and a moving plate 109 is threadedly connected to the outer ring of the first screw rod 108.

[0065] The working principle of the embodiment of the present invention is: when the flow rate of the ventilation system needs to be adjusted, the electric hoist 102 can be controlled to drive the moving frame 103 to move up and down in the inner cavity of the speed control frame 101, thereby changing the flow area at the bottom of the speed control frame 101. When the ventilation speed needs to be further changed, the first motor 107 is started to drive the first screw rod 108 to rotate, thereby realizing the movement of the moving plate 109, thereby changing the number of through holes 104 used to meet the needs of speed adjustment.

[0066] Embodiment 3;

[0067] See also Figure 1-Figure 8 ,

[0068] The adjustment mechanism 2 includes an adjustment frame 201, which is a hollow structure. An electric push rod 202 is fixedly connected to the inner wall of the adjustment frame 201, and an adjustment plate 203 is fixedly connected to one side of the electric push rod 202. Grooves 204 are provided at the top and bottom of the inner cavity of the adjustment frame 201, and the adjustment plate 203 can move left and right in the inner cavity of the groove 204.

[0069] The lifting mechanism 3 includes a base plate 301, a second motor 302 is installed on the top of the base plate 301, a connecting plate 303 is installed on the top of the base plate 301, a second screw rod 304 is rotatably connected between the second motor 302 and the connecting plate 303, the outer ring of the second screw rod 304 is threadedly connected to the first moving block 305, the top of the first moving block 305 is hinged with a rotating plate 306, the top of the rotating plate 306 is hinged with a second moving block 307, a lifting plate 308 is installed on the top of the second moving block 307, a fixed seat 309 is installed on the top of the lifting plate 308, a transport tube 310 is installed on the top of the fixed seat 309, and one end of the transport tube 310 is fixedly connected to the adjustment frame 201.

[0070] The working principle of the embodiment of the present invention is: the electric push rod 202 is started to drive the adjustment plate 203 to move in the inner cavity of the groove 204, thereby changing the internal usable area of ​​the adjustment frame 201 and changing the ventilation area.

[0071] The second motor 302 is started to drive the second screw rod 304 to rotate, thereby changing the moving distance of the first moving block 305. During the movement of the first moving block 305, the rotating plate 306 is driven to rotate, thereby driving the second moving block 307 to move and complete the lifting and lowering of the top lifting plate 308. The lifting and lowering of the lifting plate 308 will change the height of the transport tube 310 to meet the use requirements.

[0072] Working principle: The geometric characteristic parameters, operation characteristic parameters and pollutant transmission characteristic parameters of the ventilation system are used as design parameters, including the air supply port size, exhaust port size, distance between the centers of the air supply and exhaust ports, air supply port height, exhaust port height, air supply port temperature, air supply port speed, pollution source type and pollution source emission rate, a total of 9 design parameters, and the pollutant removal efficiency of the ventilation system is used as the design target. Extract the data of the characteristic parameters and design targets of the ventilation system, use the orthogonal method to design the experiment, perform numerical simulation calculations based on the experimental design scheme combined with the CFD method, and conduct experimental verification, extract data from the CFD calculation results and experiments, and extract the mapping of parameter characteristics and design targets.

[0073] A sample library is constructed for the mapping set of parameter features and outputs, and the sample set is standardized. For features that cannot be identified in the sample library, such as ventilation "mode", they are converted into sparse matrices for processing through the hot encoder method; data normalization is achieved through feature scaling methods. The standardized sample set is divided into training sets and test sets in a random manner with a division ratio of 8:2.

[0074] A variety of machine learning models with different algorithms were selected, including back propagation neural network algorithm (BPNN), support vector machine algorithm (SVM), random forest algorithm (RF), extreme gradient boosting algorithm (XGBoost), adaptive boosting algorithm (AdaBoost) and gradient boosting algorithm (GBoost).

[0075] The machine learning model is trained and tested based on the sample set, and the hyperparameters are tuned using 10-fold cross-validation based on the Bayesian optimization method; the tuned machine learning model is trained and tested using the sample set.

[0076] Compare and evaluate machine learning models of different algorithms, and select the machine learning model with the best comprehensive performance as the proxy model. The comprehensive evaluation indicators include the correlation coefficient R2, root mean square error (RMSE) and mean absolute relative error (MAPE). R2 is the square of the correlation coefficient reflecting the correlation between variables, and its value range is [0,1]. When it exceeds 0.8, it is considered to have a good fit, and the larger the value, the better. RMSE represents the dispersion of the sample, and the smaller its value, the better. MAPE characterizes the accuracy of the sample, and it is more sensitive to relative errors. It can be expressed as follows:

[0077]

[0078] Among them, p i , a i (i=0,1,…,n) represent the predicted output and actual value respectively.

[0079] Determine the constraints and optimization objectives of characteristic parameters to obtain the optimal design solution;

[0080] The influence of characteristic parameters is quantified by the surrogate model, and the constraints and optimization objectives of the characteristic parameters are determined according to the verification model. The surrogate model is used as the fitness function, and the genetic algorithm optimization model is used to search for the optimal characteristic parameter design scheme.

[0081] The optimal design solution is calculated based on the CFD method. If the calculation result is inconsistent with the calculation result of the proxy model, return to step 7 and repeat the machine learning training-optimization process, and iterate repeatedly until the calculation results are consistent.

[0082] For the ventilation system performance prediction and optimization method based on machine learning and intelligent algorithms disclosed in the above-mentioned embodiments, based on Taguchi orthogonal experimental design, data is extracted from the CFD calculation results and experimental results of ventilation systems with different characteristics and different ventilation modes, and a sample library is constructed, which can serve as a support for subsequent work. By comparing and evaluating machine learning models, predictors with better performance are screened out as proxy models, and an iterative process of prediction-optimization is performed. The design target samples can be predicted based on a smaller sample set space, and a machine learning model suitable for a small sample library is established. Combined with the intelligent algorithm, it can quickly and accurately search for the optimal solution in the parameter space for the working conditions of the input feature parameter set, and correct the optimal solution according to the CFD method to ensure the accuracy of the prediction and optimization results, and quickly realize the operation control and optimal design of the ventilation system.

[0083] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit thereof, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A ventilation system performance prediction and optimization method based on machine learning, characterized by: A ventilation system based on machine learning is required, the ventilation system comprising a speed control mechanism (1), an adjustment mechanism (2) being installed on the front and back sides of the speed control mechanism (1), and a lifting mechanism (3) being installed on the side of the adjustment mechanism (2) away from the speed control mechanism (1); The ventilation system performance prediction and optimization method based on machine learning includes the following steps: Step S1, determining characteristic parameters and design objectives of the ventilation system; Step S2, extracting data of ventilation system characteristic parameters and design objectives; Step S3, constructing a sample library for the mapping set of parameter features and outputs; Step S4, standardizing the sample set; Step S5, dividing the standardized sample set into a training set and a test set; Step S6: Select multiple machine learning models with different algorithms; Step S7: training and testing the machine learning model based on the sample set; Step S8: Compare and evaluate machine learning models of different algorithms, and select the machine learning model with the best comprehensive performance as the proxy model; Step S9, determining the constraints and optimization targets of the characteristic parameters, and obtaining the optimal design solution; Step S10: Calculate the optimal design solution based on the CFD method. If the calculation result is inconsistent with the calculation result of the proxy model, return to step 7 and re-perform the machine learning training-optimization process, and iterate repeatedly until the calculation results are consistent; In step S1, the geometric characteristic parameters, operation characteristic parameters and pollutant transmission characteristic parameters of the ventilation system are used as design parameters, including the air supply port size, the exhaust port size, the distance between the centers of the air supply and exhaust ports, the air supply port height, the exhaust port height, the air supply port temperature, the air supply port speed, the type of pollution source and the pollution source emission rate, a total of 9 design parameters, and the pollutant removal efficiency of the ventilation system is used as the design target; In step S2, an orthogonal method is used to design an experiment, a numerical simulation calculation is performed according to the experimental design scheme in combination with a CFD method, and experimental verification is performed on the numerical simulation calculation, data is extracted from the CFD calculation results and experiments, and a mapping between parameter characteristics and design objectives is extracted; The speed control mechanism (1) comprises a speed control frame (101), an electric hoist (102) is installed on the top of the inner cavity of the speed control frame (101), a moving frame (103) is installed on the bottom of the electric hoist (102), the electric hoist (102) is controlled to drive the moving frame (103) to move up and down in the inner cavity of the speed control frame (101), the inner cavity of the moving frame (103) is provided with through holes (104), the through holes (104) are distributed in a matrix array, the inner wall of the speed control frame (101) is provided with a first sliding groove (105), the moving frame (103) is provided with a plurality of through holes (104) arranged in a plurality of matrix arrays, and the inner wall of the speed control frame (101) is provided with a first sliding groove (105). A first slider (106) is fixedly connected on both sides, and the first slider (106) is slidably connected in the inner cavity of the first slide groove (105). A first motor (107) is installed on the top of the moving frame (103). A first screw rod (108) is fixedly connected to the bottom of the output shaft of the first motor (107). The outer ring of the first screw rod (108) is threadedly connected to a moving plate (109). When the first motor (107) is started, the first screw rod (108) is driven to rotate, thereby realizing the movement of the moving plate (109), thereby changing the number of through holes (104) used; The adjustment mechanism (2) comprises an adjustment frame (201), the adjustment frame (201) is a hollow structure, an inner wall of the adjustment frame (201) is fixedly connected to an electric push rod (202), one side of the electric push rod (202) is fixedly connected to an adjustment plate (203), the top and bottom of the inner cavity of the adjustment frame (201) are both provided with grooves (204), the adjustment plate (203) can move left and right in the inner cavity of the groove (204), and the electric push rod (202) is started to drive the adjustment plate (203) to move in the inner cavity of the groove (204), thereby changing the internal use area of ​​the adjustment frame (201) and changing the ventilation area; The lifting mechanism (3) comprises a bottom plate (301), a second motor (302) is mounted on the top of the bottom plate (301), a connecting plate (303) is mounted on the top of the bottom plate (301), a second screw rod (304) is rotatably connected between the second motor (302) and the connecting plate (303), an outer ring of the second screw rod (304) is threadedly connected to a first moving block (305), a rotating plate (306) is hingedly connected to the top of the first moving block (305), a second moving block (307) is hingedly connected to the top of the rotating plate (306), and the second moving block (307) A lifting plate (308) is installed on the top of the lifting plate (308), a fixing seat (309) is installed on the top of the fixing seat (309), a transport tube (310) is installed on the top of the fixing seat (309), one end of the transport tube (310) is fixedly connected to the adjustment frame (201), and the second motor (302) is started to drive the second screw rod (304) to rotate, thereby changing the moving distance of the first moving block (305), thereby driving the rotating plate (306) to rotate and the second moving block (307) to move, so as to complete the change of the height of the top lifting plate (308) and the transport tube (310).

2. The ventilation system performance prediction and optimization method based on machine learning according to claim 1, characterized in that: In step S4, for the unrecognizable features in the sample library, they are converted into sparse matrices for processing by a hot encoder method, and data normalization is achieved by a feature scaling method.

3. The ventilation system performance prediction and optimization method based on machine learning according to claim 1, characterized in that: In step S5, the ratio of the training set to the test set is 8:

2.

4. The ventilation system performance prediction and optimization method based on machine learning according to claim 1, characterized in that: In step S6, the specific algorithms of the machine learning model include back propagation neural network algorithm (BPNN), support vector machine algorithm (SVM), random forest algorithm (RF), extreme gradient boosting algorithm (XGBoost), adaptive boosting algorithm (AdaBoost) and gradient boosting algorithm (GBoost).

5. The ventilation system performance prediction and optimization method based on machine learning according to claim 1, characterized in that: In step S8, the comprehensive evaluation index includes the correlation coefficient R2, the root mean square error (RMSE) and the mean absolute relative error (MAPE).

Citation Information

Patent Citations

  • Control methods and ventilation systems for ventilation of enclosed spaces

    CN103109137B

  • Ventilating system

    CN104315680A

  • System for controlling heating, ventilation and air conditioning devices

    WO2023075631A1

  • KR20230130831A