CN115616163A - Gas accurate preparation and concentration measurement system - Google Patents
Gas accurate preparation and concentration measurement system Download PDFInfo
- Publication number
- CN115616163A CN115616163A CN202211306623.2A CN202211306623A CN115616163A CN 115616163 A CN115616163 A CN 115616163A CN 202211306623 A CN202211306623 A CN 202211306623A CN 115616163 A CN115616163 A CN 115616163A
- Authority
- CN
- China
- Prior art keywords
- neural network
- output
- gas concentration
- network model
- gas
- 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.)
- Withdrawn
Links
- 238000005259 measurement Methods 0.000 title claims abstract description 65
- 238000002360 preparation method Methods 0.000 title claims abstract description 41
- 238000001514 detection method Methods 0.000 claims abstract description 53
- 238000012544 monitoring process Methods 0.000 claims abstract description 53
- 238000003062 neural network model Methods 0.000 claims description 168
- 238000013528 artificial neural network Methods 0.000 claims description 150
- 230000001537 neural effect Effects 0.000 claims description 74
- 230000003044 adaptive effect Effects 0.000 claims description 25
- 230000004060 metabolic process Effects 0.000 claims description 18
- 239000003085 diluting agent Substances 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 8
- 230000006854 communication Effects 0.000 claims description 8
- 239000002994 raw material Substances 0.000 claims description 8
- 238000004891 communication Methods 0.000 claims description 7
- 238000002156 mixing Methods 0.000 claims description 6
- 230000000306 recurrent effect Effects 0.000 claims description 6
- 238000009472 formulation Methods 0.000 claims description 5
- 239000000203 mixture Substances 0.000 claims description 5
- 230000005540 biological transmission Effects 0.000 claims description 4
- 230000001105 regulatory effect Effects 0.000 claims description 4
- 230000003750 conditioning effect Effects 0.000 claims description 3
- 230000001276 controlling effect Effects 0.000 claims description 3
- 238000000034 method Methods 0.000 abstract description 54
- 230000008569 process Effects 0.000 abstract description 32
- 238000009826 distribution Methods 0.000 abstract description 16
- 230000004044 response Effects 0.000 abstract description 4
- 239000007789 gas Substances 0.000 description 179
- 230000006870 function Effects 0.000 description 60
- 230000009467 reduction Effects 0.000 description 22
- 239000002245 particle Substances 0.000 description 21
- 239000013598 vector Substances 0.000 description 20
- 238000013461 design Methods 0.000 description 19
- 238000004422 calculation algorithm Methods 0.000 description 18
- 238000005457 optimization Methods 0.000 description 14
- 210000002569 neuron Anatomy 0.000 description 12
- 239000011159 matrix material Substances 0.000 description 9
- 238000013507 mapping Methods 0.000 description 8
- 238000012546 transfer Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 7
- 230000006872 improvement Effects 0.000 description 7
- 230000008901 benefit Effects 0.000 description 6
- 230000001934 delay Effects 0.000 description 6
- 230000002457 bidirectional effect Effects 0.000 description 5
- 238000003825 pressing Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 230000003111 delayed effect Effects 0.000 description 4
- 230000005284 excitation Effects 0.000 description 4
- 238000011478 gradient descent method Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000010355 oscillation Effects 0.000 description 3
- 230000003213 activating effect Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000013016 damping Methods 0.000 description 2
- 230000006837 decompression Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000012886 linear function Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000007175 bidirectional communication Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000002503 metabolic effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 239000003921 oil Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004043 responsiveness Effects 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 231100000331 toxic Toxicity 0.000 description 1
- 230000002588 toxic effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N1/00—Sampling; Preparing specimens for investigation
- G01N1/28—Preparing specimens for investigation including physical details of (bio-)chemical methods covered elsewhere, e.g. G01N33/50, C12Q
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N1/00—Sampling; Preparing specimens for investigation
- G01N1/28—Preparing specimens for investigation including physical details of (bio-)chemical methods covered elsewhere, e.g. G01N33/50, C12Q
- G01N1/38—Diluting, dispersing or mixing samples
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0006—Calibrating gas analysers
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0011—Sample conditioning
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0011—Sample conditioning
- G01N33/0018—Sample conditioning by diluting a gas
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0027—General constructional details of gas analysers, e.g. portable test equipment concerning the detector
- G01N33/0031—General constructional details of gas analysers, e.g. portable test equipment concerning the detector comprising two or more sensors, e.g. a sensor array
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0027—General constructional details of gas analysers, e.g. portable test equipment concerning the detector
- G01N33/0031—General constructional details of gas analysers, e.g. portable test equipment concerning the detector comprising two or more sensors, e.g. a sensor array
- G01N33/0034—General constructional details of gas analysers, e.g. portable test equipment concerning the detector comprising two or more sensors, e.g. a sensor array comprising neural networks or related mathematical techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0062—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0062—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display
- G01N33/0068—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display using a computer specifically programmed
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0073—Control unit therefor
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/04—Programme control other than numerical control, i.e. in sequence controllers or logic controllers
- G05B19/042—Programme control other than numerical control, i.e. in sequence controllers or logic controllers using digital processors
- G05B19/0423—Input/output
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N1/00—Sampling; Preparing specimens for investigation
- G01N1/28—Preparing specimens for investigation including physical details of (bio-)chemical methods covered elsewhere, e.g. G01N33/50, C12Q
- G01N2001/2893—Preparing calibration standards
Landscapes
- Chemical & Material Sciences (AREA)
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Medicinal Chemistry (AREA)
- Food Science & Technology (AREA)
- Combustion & Propulsion (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a gas accurate preparation and concentration measurement system, which comprises a gas concentration detection and preparation subsystem and a gas concentration measurement and preparation monitoring platform, and is used for monitoring the gas concentration measurement and preparation process; the invention aims to improve the accuracy and quick response of gas concentration in gas detection and dynamic gas distribution, reduce gas concentration errors caused by lag of a gas measurement and configuration system, improve the response speed of the gas measurement and configuration system, accurately detect and adjust the gas concentration in the gas measurement and dynamic gas distribution process, and apply a gas concentration measurement and configuration monitoring platform to carry out information management and remote monitoring on the gas concentration detection and configuration process, thereby meeting the requirements of users on gas detection and gas distribution.
Description
Technical Field
The invention relates to the technical field of gas concentration detection and treatment, in particular to a gas accurate preparation and concentration measurement system.
Background
In the production process of coal, oil, natural gas and the like, the rapid and accurate monitoring and automatic control of inflammable, explosive, toxic and harmful gases become one of the important problems to be solved urgently at present. Effective monitoring and control of environmental pollution requires a series of novel sensing and measuring techniques, particularly a detection instrument and system capable of continuously monitoring the content of harmful pollution gases in the atmosphere. In order to effectively reduce the occurrence of accidents and the pollution to the environment, instruments and equipment capable of rapidly detecting gas on line in real time must be arranged. In the gas concentration preparation, the gas with standard concentration can be obtained by adopting a static gas distribution method and a dynamic gas distribution method. The static gas distribution method calculates the amount of substances according to the molecular weight and the concentration to be prepared, utilizes an ultrahigh precision balance to weigh the mass of each component, mixes and pressurizes a plurality of gases according to the calculated volume, and then fills the gases into a high-pressure gas steel cylinder. In the dynamic gas distribution method, a flow controller is mostly adopted to control the gas flow, and the set value of the mass flow controller is calculated according to the concentration value of the raw material gas and the gas distribution concentration value. When the diluent gas enters the gas mixing chamber and is mixed with the raw material gas, the diluent gas can enter the gas mixing chamber after a period of time, and the gas concentration lags for a certain time in the gas mixing process to generate fluctuation of the gas concentration, so that the gas concentration is inaccurate in the gas distribution process, and the final gas concentration cannot accurately and quickly reach the required value. The gas intelligent detection and intelligent configuration system is designed according to the actual needs of the industrial and agricultural production process, and can realize measurement, adjustment and configuration of gas concentration.
Disclosure of Invention
The invention discloses a gas accurate preparation and concentration measurement system, aiming at improving the accuracy and quick response of gas concentration in gas detection and dynamic gas distribution, reducing gas concentration errors caused by lag of the gas measurement and configuration system, improving the response speed of the gas measurement and configuration system, accurately detecting and adjusting the gas concentration in the gas measurement and dynamic gas distribution process, and performing informatization management and remote monitoring on the detection and configuration process of the gas concentration by using a gas concentration measurement and preparation monitoring platform to meet the requirements of users on the gas detection and gas distribution.
In order to solve the problems, the invention adopts the following technical scheme:
the system is characterized by comprising a gas concentration detection and preparation subsystem and a gas concentration measurement and preparation monitoring platform, and the system is used for monitoring the measurement and preparation process of the gas concentration.
The invention further adopts the technical improvement scheme that:
the parameter detection module consists of a plurality of noise reduction self-coding neural networks-NARX neural network models, an adaptive AP (access point) clustering device, wavelet recurrent neural network models of a plurality of PSOs (particle swarm optimization) and an ESN neural network model, wherein measurement parameters output by a plurality of parameter sensors for a period of time are respectively used as the input of the corresponding noise reduction self-coding neural networks-NARX neural network models, the output of the plurality of noise reduction self-coding neural networks-NARX neural network models is used as the input of the adaptive AP clustering device, the output values of the different types of noise reduction self-coding neural networks-NARX neural network models output by the adaptive AP clustering device are respectively used as the input of the corresponding PSOs wavelet recurrent neural network models, the output of the wavelet recurrent neural network models of the plurality of PSOs is used as the corresponding input of the ESN neural network models, and the output of the ESN neural network models is used as the output of the parameter detection module; the parameter detection module is shown in fig. 1. .
The invention further adopts the technical improvement scheme that:
the parameter prediction module comprises a parameter detection module, a TDL beat-to-beat delay line A, a metabolism GM (1, 1) trend model, a NARX neural network model A, a NARX neural network model B, a TDL beat-to-beat delay line C, a TDL beat-to-beat delay line D, a BAM neural network-AANN auto-associative neural network model with interval hesitation fuzzy numbers and a BAM neural network-NARX neural network model, wherein the output of the parameter detection module is used as the input of the TDL beat-to-beat delay line A, the output of the TDL beat-to-beat delay line A is used as the input of the metabolism GM (1, 1) trend model, the difference between the output of the TDL beat-to-beat delay line A and the output of the metabolism GM (1, 1) trend model and the output of the metabolism GM (1, 1) trend model are respectively used as the input of the NARX neural network model A and the NARX neural network model B, the output of NARX neural network model A and NARX neural network model B are used as the input of TDL beat-to-beat delay line B and TDL beat-to-beat delay line C, the output of TDL beat-to-beat delay line B, TDL beat-to-delay line C and BAM neural network-AANN auto-associative neural network model of interval hesitation fuzzy number are used as the corresponding input of BAM neural network-AANN auto-associative neural network model of interval hesitation fuzzy number, the 4 parameters output by BAM neural network-AANN auto-associative neural network model of interval hesitation fuzzy number are a, B, C and D, the interval number [ a, B ] of a and B is used as the minimum value of the detected parameter, the interval number [ C, D ] of C and D is used as the maximum value of the detected parameter, the interval number [ a, B ] and the interval number [ C, D ] are used as the minimum value of the detected parameter, the components ([ a, B ], [ C, D ]) and the interval fuzzy number [ C, D ]) are used as the interval number of the detected parameter, and the output of the BAM neural network-AANN auto-associative neural network model of the interval hesitation fuzzy number is used as the input of the BAM neural network-NARX neural network model, and the output of the BAM neural network-NARX neural network model is used as the output of the parameter prediction module. The parameter prediction module is shown in fig. 1.
The invention further adopts the technical improvement scheme that:
the gas concentration detection and preparation subsystem comprises an ANFIS fuzzy neural network-NARX neural network model, a parameter prediction module, an ESN neural network-NARX neural network controller, a TDL beat-to-beat delay line and a PI controller-NARX neural network controller, wherein a gas concentration set value, BAM neural network-NARX neural network model output and ESN neural network-NARX neural network controller output are respectively used as corresponding inputs of the ANFIS fuzzy neural network-NARX neural network model, the difference between the ANFIS fuzzy neural network-NARX neural network model output and the BAM neural network-NARX neural network model output is used as a gas concentration error, and the gas concentration error change rate are used as corresponding inputs of the ESN neural network-NARX neural network controller, the sum of the output of an ESN neural network-NARX neural network controller and the output of a PI controller-NARX neural network controller is used as the regulating quantity of the gas regulating device, the outputs of a plurality of gas concentration sensors are used as the input of a parameter detection module, the output of the parameter detection module is used as the input of a parameter prediction module, the output of the parameter prediction module is used as the input of a TDL (time domain delay line), the output of the TDL is used as the input of a BAM neural network-NARX neural network model, the output of the BAM neural network-NARX neural network model is used as the corresponding input of the BAM neural network-NARX neural network model, the gas concentration difference and the gas concentration error change rate output by a gas concentration set value and the parameter detection module are used as the input of the PI controller-NARX neural network controller, and a gas concentration control subsystem realizes the detection and regulation of the gas concentration. The gas concentration detection and dispensing subsystem is shown in figure 2.
The invention further adopts the technical improvement scheme that:
the noise reduction self-coding neural network-NARX neural network model, the BAM neural network-AANN self-association neural network model, the BAM neural network-NARX neural network model, the ANFIS fuzzy neural network-NARX neural network model, the ESN neural network-NARX neural network controller and the PI controller-NARX neural network controller are characterized in that the noise reduction self-coding neural network is connected with the NARX neural network model in series, the BAM neural network is connected with the AANN self-association neural network model in series, the BAM neural network is connected with the NARX neural network model in series, the ANFIS fuzzy neural network is connected with the NARX neural network model in series, the ESN neural network is connected with the NARX neural network controller in series, and the PI controller is connected with the NARX neural network controller in series.
The invention further adopts the technical improvement scheme that:
the gas concentration measurement and preparation monitoring platform is characterized in that: the gas concentration parameter measuring end is responsible for collecting gas concentration information, a gas concentration detection and preparation subsystem is arranged in the field monitoring end, bidirectional communication of the gas concentration parameter measuring end, the field monitoring end, the gas concentration parameter cloud platform and the gas concentration monitoring mobile phone APP is achieved through a gateway node, gas concentration parameter collection and gas concentration control are achieved, and the gas concentration detection and preparation subsystem is achieved in the right 1. The gas concentration measurement and dispensing monitoring platform is shown in fig. 3.
The invention further adopts the technical improvement scheme that:
the gas concentration parameter measuring terminal comprises a gas concentration sensor group, a temperature sensor group, a humidity sensor group, a pressure sensor group and corresponding signal conditioning circuits, an STM32 microprocessor and a CC2530 wireless transmission module, wherein the gas concentration sensor group, the temperature sensor group, the humidity sensor group and the pressure sensor group are used for acquiring gas concentration parameters; the gas concentration parameter measurement terminal is shown in fig. 4.
The invention further adopts the technical improvement scheme that:
the gas concentration preparation end comprises an STM32 single chip microcomputer, a CC2530 module and a gas adjusting device, wherein the gas adjusting device comprises a raw material gas, a diluent gas, a flow controller, a mixer and a mixing gas chamber, and the flow of the diluent gas is adjusted by controlling the flow controller through the STM32 single chip microcomputer at the control end, so that the diluent gas and the raw material gas are mixed in the mixer and flow to the mixer chamber. The gas concentration dispense end is shown in FIG. 5.
Compared with the prior art, the invention has the following obvious advantages:
1. aiming at the uncertainty and randomness of the problems of accuracy error, interference, measurement abnormity and the like of a parameter measurement sensor in the parameter measurement process, the output value of the gas concentration parameter measurement sensor is converted into a BAM neural network-AANN auto-associative neural network form of interval hesitation fuzzy numbers through a parameter prediction module, the ambiguity, the dynamics and the uncertainty of the gas concentration parameter measurement are effectively processed, and the objectivity and the reliability of the detection parameter of the gas concentration parameter sensor are improved.
2. The BAM neural network of the BAM neural network-AANN auto-associative neural network model is an associative memory neural network model, the output of the BAM neural network is used as the input of the AANN auto-associative neural network model, the BAM neural network can realize a bidirectional different associative model, input sample data pairs of the BAM neural network which is summarized in advance are stored through a bidirectional associative memory matrix, when new information is input into the BAM neural network, the BAM neural network parallelly recalls and associates out corresponding output results and uses the output results as the input of the AANN auto-associative neural network model, the bidirectional associative memory network of the BAM neural network is a two-layer nonlinear feedback neural network and has the functions of associative memory, distributed storage, self learning and the like, and when an input signal is added into one layer of the BAM neural network, the other layer of the BAM neural network can obtain an output signal.
3. The transfer function of the hidden layer in the PSO adaptive wavelet neural network model uses the wavelet function, and the parameters of the wavelet function are adaptively adjusted, so that the time-frequency characteristics of the signal can be more effectively input and extracted. The blindness of BP network in structural design, linear distribution of network weight coefficient and the convexity of learning target function are avoided, the problems of local optimization and the like are fundamentally avoided in the training process of the network, the algorithm concept is simple, the convergence speed is high, the function learning capability is strong, and any nonlinear function can be approached with high precision.
4. The invention adopts the PSO adaptive wavelet neural network, avoids the requirement of activating the function to be microminiaturized in the gradient descent method and the calculation of the derivation process of the function, and the iterative formula is simple when each particle is searched, thus the calculation speed is much faster than that of the gradient descent method. By adjusting the parameters in the iterative formula, local extreme values can be well jumped out, global optimization is carried out, and the training speed of the network is simply and effectively improved. The self-adaptive wavelet neural network model of the PSO algorithm has smaller error, faster convergence rate and stronger generalization capability. The method has the advantages of simple algorithm, stable structure, high calculation convergence speed, strong global optimization capability, high identification precision and strong generalization capability.
5. The invention adopts a metabolism GM (1, 1) trend model to predict the time span of the input parameters. The metabolism GM (1, 1) trend model can predict the future time value according to the input parameter values, the future time values of each input parameter predicted by the method are added into the original sequence of the metabolism GM (1, 1) trend model, the data at the beginning of the input sequence is correspondingly removed for modeling, and then the future value of the input parameter is predicted. And so on, predicting the future value of the output parameter of the metabolism GM (1, 1) trend model. The method is called as an equal-dimensional gray number successive compensation model, can realize the prediction for a long time, and can more accurately master the change trend of the input parameter value.
Drawings
FIG. 1 is a diagram of a parameter detection module and a parameter prediction module according to the present invention;
FIG. 2 is a gas concentration detection and dispensing subsystem of the present invention;
FIG. 3 is a gas concentration measurement and formulation monitoring platform of the present invention;
FIG. 4 is a diagram of a gas concentration parameter measuring terminal according to the present invention;
FIG. 5 is a schematic diagram of the gas concentration distribution end of the present invention;
FIG. 6 is a gas measurement and control gateway of the present invention;
FIG. 7 illustrates the function of the site monitoring end structure of the present invention.
Detailed Description
The technical solution of the present invention is further described with reference to the accompanying drawings 1-7:
1. constructing a parameter detection module, wherein the parameter detection module consists of a plurality of noise reduction self-coding neural network-NARX neural network models, a self-adaptive AP (access point) clustering device, a plurality of PSO (particle swarm optimization) wavelet self-adaptive neural network models and an ESN (electronic stability network) neural network model; measuring parameters output by a plurality of groups of parameter sensors for a period of time are respectively used as the input of corresponding noise reduction self-coding neural network-NARX neural network models, the output of a plurality of noise reduction self-coding neural network-NARX neural network models is used as the input of an adaptive AP (access point) clustering device, the output values of different types of noise reduction self-coding neural network-NARX neural network models output by the adaptive AP clustering device are respectively used as the input of corresponding PSO wavelet self-adaptive neural network models, the output of a plurality of PS0 wavelet self-adaptive neural network models is used as the corresponding input of an ESN (enterprise service network) neural network model, and the output of the ESN neural network model is used as the output of a parameter detection module;
(1) Design of noise reduction self-coding neural network-NARX neural network model
The noise reduction self-coding neural network-NARX neural network model is that the output of the noise reduction self-coding neural network is used as the input of the NARX neural network model, and the noise reduction self-coding neural network is a dimension reduction method, and high-dimensional data is converted into low-dimensional data by training a multilayer neural network with a small central layer. A noise-reducing self-coding neural network (DAE) is a typical three-layer neural network with an encoding process between a hidden layer and an input layer and a decoding process between an output layer and the hidden layer. The noise-reducing self-coding neural network obtains a coding representation (coder) through a coding operation on input data, and obtains reconstructed input data (decoder) through an output decoding operation on a hidden layer, wherein the data of the hidden layer is dimension reduction data. A reconstruction error function is then defined to measure the learning effect of the noise-reduced self-coding neural network. Constraints can be added based on the error function to generate various types of noise-reducing self-coding neural networks, and the encoder and decoder and the loss function are as follows:
an encoder: h = δ (Wx + b) (1)
the training process of AE is similar to that of BP neural network, W and W 'are weight matrix, b and b' are offset, h is output value of hidden layer, x is input vector,for the output vector, δ is the excitation function, typically Sigmoid is usedA function or a tanh function. The noise reduction self-coding neural network is divided into a coding process and a decoding process, wherein the coding process is from an input layer to a hidden layer, and the decoding process is from the hidden layer to an output layer. The noise reduction self-coding neural network aims to make input and output as close as possible by using an error function, and obtains the optimal weight and bias of the self-coding network by minimizing the error function through back propagation, so as to prepare for establishing a deep self-coding network model. According to the self-coding network coding and decoding principle, coded data and decoded data are obtained by using data containing noise measurement, an error function is constructed through the decoded data and original data, and an optimal network weight and an optimal bias are obtained through a back propagation minimized error function. The measurement data is corrupted by adding noise and the corrupted data is then input into the neural network as an input layer. The reconstruction result of the noise reduction self-coding neural network is similar to the original data of the measurement parameters, and by the method, the disturbance of the measurement parameters can be eliminated and a stable structure can be obtained. The original measurement parameters are input into an encoder to obtain a characteristic expression after being input into interference input obtained by adding noise, and then are mapped to an output layer through a decoder.
The current output of the NARX neural network model not only depends on the past NARX neural network model output y (t-n), but also depends on the current denoised self-coding neural network output vector X (t) of the NARX neural network model, the delay order of the denoised self-coding neural network output vector, and the like. The output of the noise reduction self-coding neural network of the NARX neural network model is transmitted to the hidden layer through the time delay layer, the hidden layer processes signals output by the noise reduction self-coding neural network and then transmits the processed signals to the output layer, the output layer linearly weights the output signals of the hidden layer to obtain final output signals of the neural network, and the time extension layer delays signals fed back by the NARX neural network model and signals output by the noise reduction self-coding neural network of the input layer and then transmits the delayed signals to the hidden layer. The NARX neural network model has the characteristics of nonlinear mapping capability, good robustness, adaptability and the like, and is suitable for further processing a plurality of high-frequency fluctuation parts of the output predicted values of the parameter sensors. The NARX neural network model (Nonlinear Auto-Regression with External input neural network) is a dynamic feedforward neural network, the NARX neural network is a Nonlinear autoregressive network with External input, has a dynamic characteristic of multistep time delay and is connected with a plurality of layers of closed networks through feedback, the NARX neural network model is a Regression dynamic neural network which is most widely applied in a Nonlinear dynamic system, and the performance of the model is generally superior to that of a full Regression neural network. The NARX neural network model mainly comprises an input layer, a hidden layer, an output layer and an input and output delay layer, wherein before application, the delay order and the number of hidden layer neurons of the input and the output are generally determined in advance, and the output of the NARX neural network model at the time not only depends on the output y (t-n) of the NARX neural network model in the past, but also depends on the output vector X (t) of the current particle swarm optimization adaptive wavelet neural network of the NARX neural network model, the delay order of the output vector of the particle swarm optimization adaptive wavelet neural network and the like. The output of the particle swarm optimization adaptive wavelet neural network of the NARX neural network model is transmitted to the hidden layer through the time delay layer, the hidden layer processes signals output by the particle swarm optimization adaptive wavelet neural network and then transmits the processed signals to the output layer, the output layer linearly weights the output signals of the hidden layer to obtain final output signals of the neural network, and the time extension layer delays signals fed back by the NARX neural network model and the output signals of the particle swarm optimization adaptive wavelet neural network of the input layer and then transmits the delayed signals to the hidden layer. The NARX neural network model has the characteristics of non-linear mapping capability, good robustness, adaptability and the like. x (t) represents the input of the NARX neural network model, i.e., the denoised self-encoding neural network output; m represents the delay order of the external input; y (t) is the output of the neural network, i.e., the predicted value of the noise-reduced self-coding neural network output for the next time period; n is the output delay order; s is the number of hidden layer neurons; the output of the ith implicit element can thus be found to be:
in the above formula, w ji As a connection weight between the ith input and the jth implicit neuron, b j Is the bias value for the jth implicit neuron.
(2) Adaptive AP clustering device design
The self-adaptive AP clustering device completes clustering through message transmission among data objects, and mainly uses similarity (adopting negative Euclidean distance as a standard) among data points as a basis, and uses two messages of attraction responsiveness and attribution to carry out cyclic updating iteration, and finally finds out an optimal clustering result. Data set X = { X) composed of N data points 1 ,x 2 ,…,x n H, wherein the similarity between any two data points is:
simi(i,k)=-||x i -x k || 2 (5)
wherein the value on the main diagonal of the semi (i, k) is replaced by a biased parameter value p, with a larger p indicating a larger probability that the point is selected as the representative point. Therefore, the final cluster number will change as p changes, and p is generally set to the median of sami (i, k) without a priori knowledge. Defining R (i, k) as the attraction degree of the candidate representative point k to each data point i, and A (i, k) as the degree that the data point i supports k as the representative point. The larger R (i, k) + A (i, k), the greater the likelihood that the representative point k will be the data center (exemplar). The specific adaptive AP clustering device algorithm flow is as follows:
A. the initial attraction degree R (i, k) and the attribution degree R (i, k) are both isomorphic zero matrices with the similarity matrix sima (i, k).
B. Let p = -50, lamda = -0.5, continuously and circularly update R (i, K) and a (i, K) until the constraint condition is reached, and the cluster number is recorded as K1.
C. Let p = p-10, continuously and cyclically update R (i, K) and a (i, K) until the constraint condition is reached to obtain a series of cluster numbers K2,3, \ 8230l (according to experience lmax = 10).
D. In steps B and C, if it is detected that the algorithm oscillates and cannot converge, lamda (value range 0.5-0.9) eliminates the oscillation with a step size of 0.1 until the algorithm converges.
E. And (4) evaluating the clustering quality and the clustering number in the steps B and C by using the contour coefficient index, wherein the larger the index is, the better the clustering quality is, and the corresponding clustering number K is the optimal clustering number.
F. The self-adaptive AP clustering device improves the accuracy and the rapidity of an algorithm by self-adaptively adjusting the deviation parameters and the damping factors of the original AP clustering device. The algorithm utilizes the contour coefficient as a judgment index of clustering effectiveness and clustering quality, utilizes the oscillation degree as an index for judging whether the algorithm is converged after oscillation occurs, adaptively adjusts and obtains the combination of the optimal deviation parameter and the damping factor, and finally obtains the optimal clustering result.
(3) Design of wavelet adaptive neural network model of PSO
The self-adaptive wavelet neural network of the PSO self-adaptive wavelet neural network is a feed-forward type network which is provided by adopting a nonlinear wavelet base to replace a common nonlinear Sigmoid function and combining an artificial neural network on the basis of a wavelet theory, wherein a transfer function of a hidden layer uses a wavelet function, and time-frequency characteristics of input parameters can be more effectively extracted by adaptively adjusting parameters of the wavelet function. The method takes a wavelet function as an excitation function of a neuron, and the expansion and translation factors and the connection weight of the wavelet are adaptively adjusted in the optimization process of an error energy function. Let the input signal of the wavelet neural network be expressed as a one-dimensional vector x of an input parameter i (i =1,2, \ 8230;, n), the output signal being denoted y k (k =1,2, \8230;, m), the calculation formula of the wavelet neural network output layer value is:
omega in the formula ij Inputting the connection weight between the i node of the layer and the j node of the hidden layer,as wavelet basis functions, b j Is a shift factor of the wavelet basis function, a j Scale factor, omega, of wavelet basis functions jk And the connection weight between the j node of the hidden layer and the k node of the output layer is shown. Weight and threshold of adaptive wavelet neural network in this patentThe correction algorithm adopts a gradient correction method to update the network weight and the wavelet basis function parameters, so that the output of the wavelet neural network continuously approaches to the expected output. The PSO adaptive wavelet neural network is adopted, the requirements of activating the function to be microminiaturized and calculating the derivation process of the function in the gradient descent method are avoided, and the iterative formula is simple during searching of each particle, so that the calculation speed is much higher than that of the gradient descent method. And local extreme values can be well jumped out by adjusting parameters in the iterative formula. A population of random particles is initialized and then an optimal solution is found through iteration. In each iteration, the particle updates itself by tracking two "extrema". The first is the optimal solution pbest found by the particle itself, this solution is called the individual extremum; the other is the best solution currently found for the entire population, this solution is called the global extremum gbest. The PSO adaptive wavelet neural network is characterized in that various parameters of the PSO adaptive wavelet neural network are taken as position vectors X of particles, a mean square error energy function formula is set as an objective function for optimization, iteration is carried out through a basic formula of a particle swarm optimization algorithm, and an optimal solution is sought. The PSO adaptive wavelet neural network training algorithm is as follows:
A. initializing a network structure, determining the number of neurons in a hidden layer of the network, and determining the dimension D of a target search space.
B. Determining the number m of the particles, initializing position vectors and velocity vectors of the particles, substituting the position vectors and the velocity vectors of the particles into an algorithm iterative formula for updating, performing optimization calculation by taking an error energy function as a target function, and recording the optimal position pbest searched by each particle so far and the optimal position gbest searched by the whole particle swarm so far.
C. And searching the whole particle swarm to the optimal position gbest so far, mapping the optimal position gbest to a network weight and a threshold value for carrying out the learning, and carrying out the chemical calculation by taking an error energy function as the fitness of the particle.
D. If the error energy function value is within the error range allowed by the actual problem, finishing the iteration; otherwise, the algorithm is switched back to continue the iteration.
(4) ESN neural network model design
An ESN (Echo state network) model is a novel dynamic neural network, has all the advantages of the dynamic neural network, and can better adapt to nonlinear system identification compared with a common dynamic neural network because the Echo state network introduces a reserve pool concept. The reserve pool is a randomly connected reserve pool which is formed by converting a part connected among traditional dynamic neural networks, and the whole learning process is a process of learning how to connect the reserve pool. The "pool" is actually a randomly generated large-scale recursive structure in which the interconnection of neurons is sparse, usually denoted SD as the percentage of interconnected neurons in the total number of neurons N. The state equation of the ESN neural network model is as follows:
wherein W is a weight matrix of the reserve pool, W in Is an input weight matrix; w back Is a feedback weight matrix; x (n) represents the internal state of the neural network; w out A connection weight matrix among a core reserve pool of the ESN neural network model, an input of the neural network and an output of the neural network;is the output deviation of the neural network or may represent noise; f = f [ ] 1 ,f 2 ,…,f n ]N activation functions for neurons within the "pool"; f. of i Is a hyperbolic tangent function; f. of out Is the epsilon output functions of the ESN neural network model.
2. Building parameter prediction module design
The parameter prediction module comprises a parameter detection module, a TDL beat-to-beat delay line A, a metabolism GM (1, 1) trend model, a NARX neural network model A, a NARX neural network model B, a TDL beat-to-beat delay line C, a TDL beat-to-beat delay line D, a BAM neural network-AANN self-association neural network model with interval hesitation fuzzy numbers and a BAM neural network-NARX neural network model; the output of the parameter detection module is used as the input of a TDL beat-pressing delay line A, the output of the TDL beat-pressing delay line A is used as the input of a metabolism GM (1, 1) trend model, the difference between the output of the TDL beat-pressing delay line A and the output of the metabolism GM (1, 1) trend model are respectively used as the input of a NARX neural network model A and a NARX neural network model B, the output of the NARX neural network model A and the NARX neural network model B are respectively used as the input of a TDL beat-pressing delay line B and a TDL beat-pressing delay line C, the output of a BAM neural network-AANN auto-associative neural network model of interval hesitation fuzzy numbers is respectively used as the corresponding input of the BAM neural network-AANN auto-associative neural network model of interval hesitation fuzzy numbers, the BAM neural network-AANN auto-associative neural network model of the interval hesitation fuzzy numbers outputs 4 parameters of a, B, C and d, the interval numbers [ a and B ] of a and B are used as the minimum values of the detected parameters, the interval numbers [ C and d ] of C and d are used as the maximum values of the detected parameters, the interval numbers [ a and B ] and the interval numbers [ C and d ] are used as the interval hesitation fuzzy numbers of the detected parameters, and the BAM neural network-AANN auto-associative neural network model of the interval hesitation fuzzy numbers outputs the interval hesitation fuzzy numbers of the detected parameters.
(1) Metabolic GM (1, 1) trend model design
The GM (1, 1) gray prediction method has more advantages than the traditional statistical prediction method, whether the prediction variable obeys normal distribution or not is not required to be determined, large sample statistics is not required, the prediction model is not required to be changed at any time according to the change of the output of the ESN neural network model, a uniform differential equation model is established through an accumulation generation technology, the accumulation ESN neural network model outputs original values to be restored to obtain a prediction result, and the differential equation model has higher prediction precision. The essence of establishing the GM (1, 1) trend model is that input original data are subjected to once accumulation generation, a generated sequence presents a certain rule, and a fitted curve is obtained by establishing a differential equation model so as to predict the trend output by the ESN neural network model. The invention adopts a metabolism GM (1, 1) trend model to predict the time span of the output trend of the ESN neural network model, the metabolism GM (1, 1) trend model can predict the output trend parameter value at the future time according to the ESN neural network model value, after each output trend parameter value predicted by the method is used, the predicted output parameter trend value at the future time is added into the original number series of the output parameters of the ESN neural network model respectively, one data at the beginning of the number series is correspondingly removed for modeling, and then a plurality of future output parameter trend values are predicted. And by analogy, predicting the trend value of the output parameter at the future moment. The method is called a metabolism model, can realize the trend prediction of the output parameters for a long time, and can more accurately master the change trend of the output parameter values.
(2) BAM neural network-AANN self-association neural network model design of interval hesitation fuzzy number
The BAM neural network-AANN auto-associative neural network model of interval hesitation fuzzy numbers is that BAM neural network output is used as AANN auto-associative neural network model input, in a BAM neural network model topological structure, the initial mode of the network input end is x (t), and the initial mode is obtained by passing through a weight matrix W 1 Weighted and then reaches the y end of the output end, and passes through the transfer characteristic f of the output node y Non-linear transformation of (1) and (W) 2 The matrix is weighted and returns to the input end x, and then the transfer characteristic f of the output node at the x end is passed x The nonlinear transformation of the BAM neural network model is changed into the output of the input terminal x, and the operation process is repeated, so that the state transition equation of the BAM neural network model is shown in an equation (8).
The BAM neural network output is an AANN Auto-associative neural network model input, the AANN Auto-associative neural network model is a feedforward Auto-associative neural network (AANN) with a special structure, and the AANN Auto-associative neural network model structure comprises an input layer, a certain number of hidden layers and an output layer. The method comprises the steps of firstly realizing compression of BAM neural network output data information through an input layer, a mapping layer and a bottleneck layer of parameters, extracting a most representative low-dimensional subspace reflecting BAM neural network output from a high-dimensional parameter space output by the BAM neural network, effectively filtering noise and measurement errors in BAM neural network output data, realizing decompression of the BAM neural network output through the bottleneck layer, the demapping layer and the output layer, and restoring the compressed information to each input value, thereby realizing reconstruction of the BAM neural network output data. In order to achieve the purpose of compressing output information of the BAM neural network, the number of nodes of a bottleneck layer of an AANN self-association neural network model is obviously smaller than that of an input layer, and in order to prevent the formation of simple single mapping between output of the BAM neural network and output layers of the AANN self-association neural network, except that excitation functions of the output layers of the AANN self-association neural network adopt linear functions, other layers adopt nonlinear excitation functions. In essence, the first layer of the hidden layer of the AANN auto-associative neural network model is called as a mapping layer, and the node transfer function of the mapping layer can be an S-shaped function or other similar nonlinear functions; the second layer of the hidden layer is called a bottleneck layer, the dimension of the bottleneck layer is the minimum in the network, the transfer function of the second layer can be linear or nonlinear, the bottleneck layer avoids the mapping relation that the output and the input are equal and can be easily realized in a one-to-one mode, the bottleneck layer enables the network to encode and compress the output signal of the BAM neural network, and the decoding and the decompression of the output data of the BAM neural network are carried out after the bottleneck layer to generate the estimation value of the output signal of the BAM neural network; the third layer or the last layer of the hidden layer is called a demapping layer, the node transfer function of the demapping layer is a generally nonlinear S-shaped function, and the AANN self-associative neural network is trained by an error back propagation algorithm.
(3) BAM neural network-NARX neural network model design
The output of the BAM neural network of the NARX neural network model is used as the input of the NARX neural network model, the NARX neural network model mainly comprises an input layer, a hidden layer, an output layer and an input and output delay layer, the delay order and the hidden layer neuron number of the input and the output are generally determined in advance before application, and the current output of the NARX neural network model not only depends on the past output y (t-n) of the NARX neural network model, but also depends on the current output vector X (t) of the BAM neural network of the NARX neural network model, the delay order of the output vector of the BAM neural network and the like. The output of the BAM neural network of the NARX neural network model is transmitted to the hidden layer through the time delay layer, the hidden layer processes signals output by the BAM neural network and then transmits the signals to the output layer, the output layer linearly weights the output signals of the hidden layer to obtain final neural network output signals, and the time delay layer delays signals fed back by the NARX neural network model and signals output by the BAM neural network of the input layer and then transmits the signals to the hidden layer.
3. Gas concentration detection and preparation subsystem design
The gas concentration detection and preparation subsystem comprises an ANFIS fuzzy neural network-NARX neural network model, a parameter prediction module, an ESN neural network-NARX neural network controller, a TDL beat-to-beat delay line and a PI controller-NARX neural network controller;
(1) ANFIS fuzzy neural network-NARX neural network model design
The ANFIS Fuzzy neural network output is used as the NARX neural network model input, the ANFIS Fuzzy neural network is an Adaptive neural-Fuzzy Inference System (Adaptive neural-Fuzzy Inference System), and the neural network and the Fuzzy Inference System are organically combined, so that the advantages of the neural network and the Fuzzy Inference System can be brought into play, and the respective defects can be made up. The fuzzy membership function and fuzzy rule in the adaptive neural network fuzzy system are obtained by learning a large amount of known data, and the maximum characteristic of ANFIS is a data-based modeling method instead of any given method based on experience or intuition. The main operation steps of the ANFIS fuzzy neural network are as follows:
layer 1: fuzzifying the data of the input, and representing the corresponding output of each node as:
in the formula, n is the number of each input membership function, and the membership function adopts a Gaussian membership function.
Layer 2: and realizing rule operation, outputting the applicability of the rule, and multiplying the rule operation of the ANFIS fuzzy neural network by adopting multiplication.
Layer 3: normalizing the applicability of each rule:
layer 4: the transfer function of each node is a linear function and represents a local linear model, and the output of each self-adaptive node i is as follows:
layer 5: the single node of the layer is a fixed node, and the total output of the compensation estimation value of the ANFIS fuzzy neural network is calculated as follows:
the condition parameters determining the shape of the membership function and the conclusion parameters of the inference rule in the ANFIS fuzzy neural network can be trained through a learning process, the parameters are adjusted by adopting an algorithm combining a linear least square estimation algorithm and gradient descent, in each iteration of the ANFIS fuzzy neural network, firstly, input signals are transmitted to the layer 4 along the forward direction of the network, the condition parameters are fixed, and the conclusion parameters are adjusted by adopting a least square estimation algorithm; the signal continues to propagate forward along the network to the output layer. The ANFIS fuzzy neural network reversely propagates the obtained error signal along the network, the condition parameters are updated by a gradient method, the given condition parameters in the ANFIS fuzzy neural network are adjusted in the mode, and the global optimum point of the conclusion parameters can be obtained, so that the dimension of a search space in the gradient method can be reduced, and the convergence speed of the ANFIS fuzzy neural network parameters can be improved. Before application, the delay order and the number of hidden layer neurons of the input and output are generally determined in advance, and the current output of the NARX neural network model not only depends on the past output y (t-n) of the NARX neural network model, but also depends on the current ANFIS fuzzy neural network output vector X (t) of the NARX neural network model, the delay order of the ANFIS fuzzy neural network output vector and the like. The ANFIS fuzzy neural network output of the NARX neural network model is transmitted to the hidden layer through the time delay layer, the hidden layer processes signals output by the ANFIS fuzzy neural network and then transmits the processed signals to the output layer, the output layer linearly weights the output signals of the hidden layer to obtain final neural network output signals, and the time extension layer delays signals fed back by the NARX neural network model and the signals output by the ANFIS fuzzy neural network of the input layer and then transmits the delayed signals to the hidden layer.
(2) ESN neural network-NARX neural network controller design
The ESN neural network-NARX neural network controller is used for outputting an ESN neural network as an input of the NARX neural network controller, an ESN neural network model (Echo state network, ESN) is a novel dynamic neural network, has all advantages of the dynamic neural network, and meanwhile, as the concept of 'reserve pool' is introduced into an Echo state network, the method can better adapt to nonlinear system identification compared with a common dynamic neural network. The reserve pool is a randomly connected reserve pool which is formed by converting a part connected among traditional dynamic neural networks, and the whole learning process is a process of learning how to connect the reserve pool. The current output of the NARX neural network model not only depends on the past NARX neural network model output y (t-n), but also depends on the current ESN neural network output vector X (t) of the NARX neural network model, the delay order of the ESN neural network output vector and the like. The output of an ESN neural network of the NARX neural network model is transmitted to the hidden layer through the time delay layer, the hidden layer processes signals output by the ESN neural network and then transmits the processed signals to the output layer, the output layer linearly weights the output signals of the hidden layer to obtain final neural network output signals, and the time extension layer delays signals fed back by the NARX neural network model and signals output by the ESN neural network of the input layer and then transmits the delayed signals to the hidden layer.
(3) PI controller-NARX neural network controller design
The PI controller output of the PI controller-NARX neural network controller is used as the input of the NARX neural network controller, and the output of the NARX neural network model at the moment is not only dependent on the output y (t-n) of the NARX neural network model in the past, but also dependent on the output vector X (t) of the current PI controller of the NARX neural network model, the delay order of the output vector of the PI controller, and the like. The output of a PI controller of the NARX neural network model is transmitted to the hidden layer through the time delay layer, the hidden layer processes signals output by the PI controller and then transmits the signals to the output layer, the output layer linearly weights the output signals of the hidden layer to obtain final neural network output signals, and the time extension layer delays signals fed back by the NARX neural network model and signals output by the PI controller of the input layer and then transmits the signals to the hidden layer.
2. Gas concentration measurement and preparation monitoring platform design
The gas concentration measurement and preparation monitoring platform comprises a gas concentration parameter measurement end, an on-site monitoring end, a gas concentration preparation end, a gas measurement and control gateway, a gas concentration parameter cloud platform and a gas concentration monitoring mobile phone APP, wherein the gas concentration parameter measurement end is responsible for collecting gas concentration parameter information to be detected, a gas concentration detection and preparation subsystem is arranged in the on-site monitoring end, the gas concentration parameter measurement end, the gas concentration preparation end, the on-site monitoring end, the two-way communication of the gas concentration parameter cloud platform and the gas concentration monitoring mobile phone APP is realized through the gas concentration preparation end, and the intelligent adjustment of gas concentration parameters is realized. The gas concentration monitoring mobile phone APP accesses a gas concentration parameter cloud platform through a 5G network to realize remote monitoring on gas concentration parameters, wherein the gas concentration detection and preparation subsystem is realized in the claim 1.
1. Design of overall system function
The gas concentration measurement and preparation monitoring platform realizes the detection of gas concentration parameters and the adjustment of the gas concentration parameters, and a plurality of gas concentration parameter measurement terminals of the system construct a wireless monitoring network in a self-organizing way to realize the bidirectional wireless communication of the gas concentration parameter measurement terminal, the gas concentration preparation end, the field monitoring end, the gas concentration parameter cloud platform and the gas concentration monitoring mobile phone APP; the gas concentration parameter measurement terminal sends the detected gas concentration parameters to the field monitoring terminal and the gas concentration parameter cloud platform through the gas measurement and control gateway, and the gas concentration detection and preparation subsystem of the field monitoring terminal processes the gas concentration parameters and intelligently adjusts the gas concentration parameters; and the gas concentration parameter is monitored in real time at the APP end of the gas concentration monitoring mobile phone by accessing a gas concentration parameter cloud platform. The gas concentration measurement and formulation monitoring platform is shown in fig. 3.
2. Design of gas concentration parameter measuring terminal
A large number of gas concentration parameter measuring terminals based on a wireless sensor network are used as gas concentration parameter sensing terminals, and the gas concentration parameter measuring terminals and a gas concentration preparation terminal realize mutual information interaction among an on-site monitoring terminal, a gas measurement and control gateway and a gas concentration parameter cloud platform through a self-organizing wireless network. The gas concentration parameter measuring terminal comprises a gas concentration sensor group, a temperature sensor group, a humidity sensor group, a pressure sensor group and corresponding signal conditioning circuits, an STM32 microprocessor and a CC2530 wireless transmission module, wherein the gas concentration sensor group, the temperature sensor group, the humidity sensor group and the pressure sensor group are used for acquiring gas concentration parameters; the software of the gas concentration parameter measuring terminal mainly realizes wireless communication and acquisition and pretreatment of gas concentration parameters. The software is designed by adopting a C language program, so that the compatibility degree is high, the working efficiency of software design and development is greatly improved, and the reliability, readability and transportability of program codes are enhanced. The structure of the gas concentration parameter measuring terminal is shown in figure 4.
3. Gas concentration dispensing tip design
The gas concentration preparation end comprises an STM32 single chip microcomputer, a CC2530 module and a gas adjusting device, wherein the gas adjusting device comprises a raw material gas, a diluent gas, a flow controller, a mixer and a mixing gas chamber, and the flow of the diluent gas is adjusted by controlling the flow controller through the STM32 single chip microcomputer at the control end, so that the diluent gas and the raw material gas are mixed in the mixer and flow to the mixer chamber. A gas concentration detection and preparation subsystem is designed in a field monitoring end to realize detection and adjustment of gas concentration parameters, and software is designed by adopting a C language program, so that the compatibility degree is high, the working efficiency of software design and development is greatly improved, and the reliability, readability and transportability of program codes are enhanced. The gas concentration distribution end structure is shown in FIG. 5.
3. Gas measurement and control gateway design
The gas measurement and control gateway comprises a CC2530 module, an NB-IoT module, an STM32 single chip microcomputer and an RS232 interface, the gas measurement and control gateway comprises a self-organizing communication network which is communicated with a gas concentration parameter measurement terminal and a gas concentration control terminal through the CC2530 module, the NB-IoT module is used for realizing data bidirectional interaction between the gas measurement and control gateway and a gas concentration parameter cloud platform, a gas concentration preparation end, a field control end and a gas concentration monitoring mobile phone APP, and the RS232 interface is connected with the field monitoring end to realize information interaction between the gas measurement and control gateway and the field monitoring end. The gas measurement and control gateway is shown in figure 6.
4. Site monitoring terminal software
The field monitoring terminal is an industrial control computer, the field monitoring terminal mainly achieves collection of gas concentration parameters and intelligent adjustment of the gas concentration parameters, information interaction between the field monitoring terminal and the gas concentration parameter preparation terminal and between the gas concentration parameter cloud platform and the gas concentration monitoring mobile phone APP is achieved, and the field monitoring terminal mainly has the functions of communication parameter setting, data analysis and data management, and gas concentration detection and preparation subsystems. The management software selects Microsoft Visua1+ +6.0 as a development tool, calls an Mscomm communication control of a system to design a communication program, and the functions of the field monitoring end software are shown in figure 7..
The technical means disclosed in the invention scheme are not limited to the technical means disclosed in the above embodiments, but also include the technical scheme formed by any combination of the above technical features. It should be noted that modifications and adaptations can be made by those skilled in the art without departing from the principles of the present invention and are intended to be within the scope of the present invention.
Claims (8)
1. The system is characterized by comprising a gas concentration detection and preparation subsystem and a gas concentration measurement and preparation monitoring platform, and is used for monitoring the measurement and preparation process of gas concentration;
the gas concentration detection and preparation subsystem comprises an ANFIS fuzzy neural network-NARX neural network model, a parameter prediction module, an ESN neural network-NARX neural network controller, a parameter detection module, a TDL beat-to-beat delay line and a PI controller-NARX neural network controller.
2. The system of claim 1, wherein: the gas concentration set value, BAM neural network-NARX neural network model output and ESN neural network-NARX neural network controller output are respectively used as corresponding inputs of the ANFIS fuzzy neural network-NARX neural network model, the difference between the ANFIS fuzzy neural network-NARX neural network model output and the BAM neural network-NARX neural network model output is used as a gas concentration error, the gas concentration error and the gas concentration error change rate are used as corresponding inputs of the ESN neural network-NARX neural network controller, the sum of the ESN neural network-NARX neural network controller output and the PI controller-NARX neural network controller output is used as a regulating variable of the gas regulating device, the output of the plurality of gas concentration sensors is used as the input of the parameter detection module, the output of the parameter detection module is used as the input of the parameter prediction module, the output of the parameter prediction module is used as the input of a TDL beat-by-beat delay line, the output of the TDL beat-by-beat delay line and the output of the BAM neural network-NARX neural network model are used as the corresponding input of the BAM neural network-NARX neural network model, the gas concentration difference and the gas concentration error change rate of the gas concentration set value and the output of the parameter detection module are used as the input of a PI controller-NARX neural network controller, and the gas concentration control subsystem realizes the detection and adjustment of the gas concentration.
3. The system for accurate gas formulation and concentration measurement according to claim 1, wherein: the parameter detection module is composed of a denoising self-coding neural network-NARX neural network model, an adaptive AP (access point) clustering device, a PSO wavelet recurrent neural network model and an ESN neural network model, measurement parameters output by a parameter sensor for a period of time are respectively used as the input of the corresponding denoising self-coding neural network-NARX neural network model, the output of the denoising self-coding neural network-NARX neural network model is used as the input of the adaptive AP clustering device, the output values of different types of denoising self-coding neural network-NARX neural network models are respectively used as the input of the corresponding PSO wavelet recurrent neural network model, the output of the PSO wavelet recurrent neural network model is used as the corresponding input of the ESN neural network model, and the output of the ESN neural network model is used as the output of the parameter detection module.
4. The system for accurate gas formulation and concentration measurement according to claim 1, wherein: the parameter prediction module comprises a parameter detection module, a TDL beat-to-beat delay line A, a metabolism GM (1, 1) trend model, a NARX neural network model A, a NARX neural network model B, a TDL beat-to-beat delay line C, a TDL beat-to-beat delay line D, a BAM neural network-AANN auto-associative neural network model with interval hesitation fuzzy numbers and a BAM neural network-NARX neural network model, wherein the output of the parameter detection module is used as the input of the TDL beat-to-beat delay line A, the output of the TDL beat-to-beat delay line A is used as the input of the metabolism GM (1, 1) trend model, the difference between the output of the TDL beat-to-beat delay line A and the output of the metabolism GM (1, 1) trend model and the output of the metabolism GM (1, 1) trend model are respectively used as the input of the NARX neural network model A and the NARX neural network model B, the output of the NARX neural network model A and the output of the NARX neural network model B are respectively used as the input of a TDL beat-to-beat delay line B and a TDL beat-to-beat delay line C, the output of the TDL beat-to-beat delay line B, the TDL beat-to-beat delay line C and the output of the BAM neural network-AANN auto-association neural network model of interval hesitation fuzzy numbers are respectively used as the corresponding input of the BAM neural network-AANN auto-association neural network model of interval hesitation fuzzy numbers, 4 parameters output by the BAM neural network-AANN auto-association neural network model of interval hesitation fuzzy numbers are respectively a, B, C and D, the output of the BAM neural network-AANN auto-association neural network model of interval hesitation fuzzy numbers is used as the input of the BAM neural network-NARX neural network model, and the output of the BAM neural network-NARX neural network model is used as the output of a parameter prediction module.
5. The system of claim 4, wherein: the a and b form interval number [ a, b ] as the minimum value of the detected parameter, the c and d form interval number [ c, d ] as the maximum value of the detected parameter, and the interval number [ a, b ] and the interval number [ c, d ] form ([ a, b ], [ c, d ]) as the interval hesitation fuzzy number of the detected parameter.
6. The system of claim 1, wherein: gas concentration measurement and configuration monitoring platform includes that gas concentration parameter measurement holds, gas concentration configures end, on-the-spot monitoring end, and gas concentration parameter measurement holds is responsible for gathering gas concentration information, has gas concentration detection and configuration subsystem in the on-the-spot monitoring end, realizes gas concentration parameter measurement end, on-the-spot monitoring end, gas concentration parameter cloud platform and gas concentration control cell-phone APP's both-way communication through the gateway node, realizes gas concentration parameter acquisition and gas concentration control.
7. The system for accurate gas formulation and concentration measurement according to claim 6, wherein: the gas concentration parameter measuring end comprises a gas concentration sensor group, a temperature sensor group, a humidity sensor group, a pressure sensor group and a corresponding signal conditioning circuit, an STM32 microprocessor and a CC2530 wireless transmission module, wherein the gas concentration sensor group, the temperature sensor group, the humidity sensor group and the pressure sensor group are used for collecting gas concentration parameters.
8. The system of claim 6, wherein: the gas concentration preparation end comprises an STM32 single chip microcomputer, a CC2530 module and a gas adjusting device, wherein the gas adjusting device comprises a raw material gas, a diluent gas, a flow controller, a mixer and a mixing gas chamber, the flow of the diluent gas is adjusted by controlling the flow controller through the STM32 single chip microcomputer at a control end, so that the diluent gas and the raw material gas are mixed in the mixer and flow to the mixer chamber.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211306623.2A CN115616163A (en) | 2022-10-24 | 2022-10-24 | Gas accurate preparation and concentration measurement system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211306623.2A CN115616163A (en) | 2022-10-24 | 2022-10-24 | Gas accurate preparation and concentration measurement system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115616163A true CN115616163A (en) | 2023-01-17 |
Family
ID=84863764
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211306623.2A Withdrawn CN115616163A (en) | 2022-10-24 | 2022-10-24 | Gas accurate preparation and concentration measurement system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115616163A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116577176A (en) * | 2023-04-17 | 2023-08-11 | 防灾科技学院 | Component gas constant proportion conveying device capable of correcting error automatically |
-
2022
- 2022-10-24 CN CN202211306623.2A patent/CN115616163A/en not_active Withdrawn
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116577176A (en) * | 2023-04-17 | 2023-08-11 | 防灾科技学院 | Component gas constant proportion conveying device capable of correcting error automatically |
CN116577176B (en) * | 2023-04-17 | 2024-01-23 | 防灾科技学院 | Component gas constant proportion conveying device capable of correcting error automatically |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106453293B (en) | A kind of network security situation prediction method based on improved BPNN | |
CN115016276B (en) | Intelligent water content adjustment and environment parameter Internet of things big data system | |
CN115630101B (en) | Hydrologic parameter intelligent monitoring and water resource big data management system | |
CN112308288A (en) | Particle swarm optimization LSSVM-based default user probability prediction method | |
CN111832825B (en) | Wind power prediction method and system integrating long-term memory network and extreme learning machine | |
CN109190834A (en) | Stock price trend prediction method and system neural network based | |
CN110428082A (en) | Water quality prediction method based on attention neural network | |
CN115687995A (en) | Big data environmental pollution monitoring method and system | |
CN115128978A (en) | Internet of things environment big data detection and intelligent monitoring system | |
CN115905938B (en) | Storage tank safety monitoring method and system based on Internet of things | |
CN115343784A (en) | Local air temperature prediction method based on seq2seq-attention model | |
CN115616163A (en) | Gas accurate preparation and concentration measurement system | |
CN112765902A (en) | RBF neural network soft measurement modeling method based on TentFWA-GD and application thereof | |
CN112903929A (en) | Food quality detection system | |
CN115482877A (en) | Fermentation process soft measurement modeling method based on time sequence diagram network | |
CN115659201A (en) | Gas concentration detection method and monitoring system for Internet of things | |
Li et al. | Predicting Software Quality by Optimized BP Network Based on PSO. | |
CN117221352A (en) | Internet of things data acquisition and intelligent big data processing method and cloud platform system | |
CN112257893A (en) | Complex electromechanical system health state prediction method considering monitoring error | |
CN114358244B (en) | Big data intelligent detection system of pressure based on thing networking | |
Wang et al. | LSTM wastewater quality prediction based on attention mechanism | |
CN115330082A (en) | PM2.5 concentration prediction method of LSTM network based on attention mechanism | |
CN114995248A (en) | Intelligent maintenance and environmental parameter big data internet of things system | |
CN116108361B (en) | Intelligent oil quality detection method and monitoring system | |
CN108960406B (en) | MEMS gyroscope random error prediction method based on BFO wavelet neural network |
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 | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20230117 |
|
WW01 | Invention patent application withdrawn after publication |