CN109000733A - Visibility sensor and its detection method based on simulated annealing optimization neural network - Google Patents
Visibility sensor and its detection method based on simulated annealing optimization neural network Download PDFInfo
- Publication number
- CN109000733A CN109000733A CN201810931051.4A CN201810931051A CN109000733A CN 109000733 A CN109000733 A CN 109000733A CN 201810931051 A CN201810931051 A CN 201810931051A CN 109000733 A CN109000733 A CN 109000733A
- Authority
- CN
- China
- Prior art keywords
- visibility
- neural network
- temperature
- value
- particle concentration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- 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
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Automation & Control Theory (AREA)
- Investigating Or Analyzing Materials By The Use Of Electric Means (AREA)
- Investigating Or Analyzing Materials By The Use Of Fluid Adsorption Or Reactions (AREA)
Abstract
The invention discloses a kind of visibility sensor and its detection method based on simulated annealing optimization neural network, its visibility sensor includes microprocessor module and the power circuit for electricity consumption module for power supply each in the visibility sensor, the input of the microprocessor module is terminated with visibility detection circuit, the visibility detection circuit includes particle concentration sensor and Temperature Humidity Sensor, and the microprocessor module, particle concentration sensor and Temperature Humidity Sensor are connect with the output end of power circuit.Its detection method is comprising steps of the acquisition of one, data and transmission;Step 2: data prediction;Step 3: data processing obtains visibility detected value.Of the invention novel in design reasonable, it is convenient and at low cost to realize, visibility detection accuracy is high, can be advantageously applied to visibility detection, and practical, using effect is good, is convenient for promoting the use of.
Description
Technical field
The invention belongs to visibility detection technique fields, and in particular to a kind of energy based on simulated annealing optimization neural network
Degree of opinion sensor and its detection method.
Background technique
In recent years, as China's expanding economy, the discharge of pollutant gradually increase, air visibility is caused generally to drop
It is low, affect the normal operation of communications and transportation.Meanwhile the reduction of visibility will affect the illuminating effect of street lamp, to the night of people
Between trip cause very big influence.For example, a large number of studies show that yellow light penetration capacity is stronger than the penetration capacity of white light, so working as
When visibility is lower, the bright orange light of street lamp;When visibility is too low, white light and one piece of yellow light it is bright, be achieved according to visibility
The different bright different colours of street lamp, realize the intelligence of street lamp, are beneficial to the trip in people's evening, can generate to people's lives
Preferable effect.Therefore, real-time detection visibility seems most important to people's lives.
Visibility sensor in the prior art is to be measured in air by laser light scattering principle by sampling type mostly
The sum of discrete light particle measures visibility.The measurement accuracy of such sensor is high, but expensive, for visibility requirement
For not high place, the economic benefit is not high.
Based on the above issues the considerations of, a large amount of scholars have researched and proposed different methods to visibility.These methods can
To be divided into two classes, one kind is curve matching, and another kind of is machine learning.Curve matching is fitted according to a large amount of weather history data
One using three concentration of PM2.5, temperature and humidity variables as independent variable out, and visibility is the formula of dependent variable, but such methods
It is larger to detect visibility error, it is undesirable.Machine learning is the parameter using a large amount of weather history data training patterns, so
Trained model inspection visibility is utilized afterwards, and a large amount of scholars detect energy using models such as support vector machines, BP neural networks
Degree of opinion.But this method does not carry out partitioning model according to the four seasons, detection effect may influence energy according to the difference in the four seasons
The detection of degree of opinion.
Summary of the invention
In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is that providing a kind of circuit structure
Simply, novel design is rationally, realization is convenient and at low cost, visibility detection accuracy is high, practical excellent based on simulated annealing
Change the visibility sensor of neural network.
In order to solve the above technical problems, the technical solution adopted by the present invention is that: one kind being based on simulated annealing optimization nerve net
The visibility sensor of network, it is characterised in that: including microprocessor module and be each electricity consumption module in the visibility sensor
The power circuit of power supply, the input of the microprocessor module are terminated with visibility detection circuit, the visibility detection circuit
Including particle concentration sensor and Temperature Humidity Sensor, the microprocessor module, particle concentration sensor and temperature and humidity
Sensor is connect with the output end of power circuit.
The above-mentioned visibility sensor based on simulated annealing optimization neural network, it is characterised in that: the power circuit
Including 5V battery, voltage stabilizing chip AMS1117, nonpolar capacitor C8, nonpolarity capacitor C9 and polar capacitor C10, the pressure stabilizing core
The 3rd pin of piece AMS1117 is connect with the 5V voltage output end of 5V battery, and is grounded by nonpolarity capacitor C8, the 5V electricity
The 5V voltage output end in pond is the 5V voltage output end of power circuit, and the 2nd pin of the voltage stabilizing chip AMS1117 is power supply electricity
The 3.3V voltage output end on road, and pass through nonpolar capacitor C9 and polar capacitor C10 ground connection in parallel, the voltage stabilizing chip
The 1st pin of AMS1117 is grounded.
The above-mentioned visibility sensor based on simulated annealing optimization neural network, it is characterised in that: the microprocessor
Module includes single-chip microcontroller MSP430F169, the first crystal oscillating circuit, the second crystal oscillating circuit and reset circuit, first crystal oscillating circuit
One end of one end and nonpolar capacitor C1 including crystal oscillator Y0, nonpolarity capacitor C1 and nonpolar capacitor C2, the crystal oscillator Y0 is equal
Connect with the 53rd pin of single-chip microcontroller MSP430F169, one end of the other end of the crystal oscillator Y0 and nonpolar capacitor C2 with list
The 52nd pin of piece machine MSP430F169 connects, and the other end of the other end of the nonpolarity capacitor C1 and nonpolar capacitor C2 are equal
Ground connection;Second crystal oscillating circuit includes crystal oscillator Y1, nonpolar capacitor C6, nonpolarity capacitor C7 and nonpolarity capacitor C8, described
One end of crystal oscillator Y1 and one end of nonpolar capacitor C6 are connect with the 8th pin of single-chip microcontroller MSP430F169, the crystal oscillator Y1
The other end and one end of nonpolar capacitor C7 connect with the 9th pin of single-chip microcontroller MSP430F169, the nonpolar capacitor
One end of C8 is connect with the 7th pin of single-chip microcontroller MSP430F169, the other end of the nonpolarity capacitor C6, nonpolar capacitor C7
The other end and the other end of nonpolar capacitor C8 be grounded;The reset circuit includes reset key S0, resistance R0 and non-pole
Property capacitor C0, one end of one end of the reset key S0, one end of resistance R0 and polar capacitor C0 is and single-chip microcontroller
The 58th pin of MSP430F169 connects, and the other end of the reset key S0 and the other end of polar capacitor C0 are grounded, institute
The 3.3V voltage output end of the other end and power circuit of stating resistance R0 connects.
The above-mentioned visibility sensor based on simulated annealing optimization neural network, it is characterised in that: the particulate matter is dense
Spending sensor includes particle concentration sensor PMS3003, and the Temperature Humidity Sensor includes Temperature Humidity Sensor DHT11, institute
The power end pin of the power end pin and Temperature Humidity Sensor DHT11 of stating particle concentration sensor PMS3003 is and power supply
The 5V voltage output end of circuit connects, the ground terminal pin and Temperature Humidity Sensor of the particle concentration sensor PMS3003
The ground terminal pin of DHT11 is grounded, the signal output end pin and single-chip microcontroller of the particle concentration sensor PMS3003
The 33rd pin of MSP430F169 connects, the signal output end pin and single-chip microcontroller of the Temperature Humidity Sensor DHT11
The 36th pin of MSP430F169 connects, and is connected by the 5V voltage output end of resistance R1 and power circuit.
The above-mentioned visibility sensor based on simulated annealing optimization neural network, it is characterised in that: the microprocessor
Spring and the autumn visibility detection model, summer energy of the BP neural network building based on simulated annealing optimization are stored in module
Degree of opinion detection model and winter visibility detection model.
That the invention also discloses a kind of method and steps is simple, detection accuracy is high, detection effect is good, it is practical, use effect
The visibility detecting method of good, convenient for popularization and use the visibility sensor based on simulated annealing optimization neural network of fruit,
It is characterized in that, method includes the following steps:
Step 1: data acquisition and transmission: particle concentration sensor adopts the particle concentration in environment in real time
Collection, and by the collected particle concentration real-time data transmission of institute to microprocessor module;Temperature Humidity Sensor is in environment
Temperature and humidity is acquired in real time, and gives the collected temperature data of institute and humidity data real-time Transmission to microprocessor mould
Block;
Step 2: data prediction: the microprocessor module is according to formula to received particle concentration data
It is normalized, the particle concentration value x after being normalized*;According to formula z*=(ymax-ymin)*(z-zmin)/
(zmax-zmin)+yminReceived temperature data is normalized, the temperature value z after being normalized*;And according to
Formula r*=(ymax-ymin)*(r-rmin)/(rmax-rmin)+yminReceived humidity data is normalized, is obtained
Humidity value r after to normalization*;Wherein, ymax=1, ymin=-1, x is that the particulate matter that microprocessor module is currently received is dense
Angle value, xminFor the minimum value for the particle concentration value that microprocessor module receives, xmaxIt is received for microprocessor module
The maximum value of particle concentration value;Z is the temperature value that microprocessor module is currently received, zminFor microprocessor module reception
The minimum value of the temperature value arrived, zmaxFor the maximum value for the temperature value that microprocessor module receives;R works as microprocessor module
Before the humidity value that receives, rminFor the minimum value for the humidity value that microprocessor module receives, rmaxIt is connect for microprocessor module
The maximum value of the humidity value received;
Step 3: data processing obtains visibility detected value: the microprocessor module judges that microprocessor module is being worked as
Temperature range locating for the daily mean of the preceding M days temperature values received of preceding detection, the day of the current M days temperature value received
When average value was within the temperature range of winter, the microprocessor module is by the particle concentration value x after normalization*, temperature
Value z*With humidity value r*Input the winter visibility that visibility is detected according to particle concentration, temperature and humidity being stored therein
In detection model, show that the output of the winter visibility detection model, the output of the winter visibility detection model are
Visibility detected value;It is described when the daily mean of the current M days temperature values received was within the temperature range of spring and autumn
Microprocessor module is by the particle concentration value x after normalization*, temperature value z*With humidity value r*Input the basis being stored therein
In particle concentration, the spring of temperature and humidity detection visibility and autumn visibility detection model, the spring and autumn are obtained
The output of its visibility detection model, the output of the spring and autumn visibility detection model are visibility detected value;When
When the daily mean of the preceding M days temperature value received was within the temperature range of summer, the microprocessor module will be normalized
Particle concentration value x afterwards*, temperature value z*With humidity value r*Input be stored therein according to particle concentration, temperature and humidity
In the summer visibility detection model for detecting visibility, the output of the summer visibility detection model, the summer energy are obtained
The output of degree of opinion detection model is visibility detected value.
Above-mentioned method, it is characterised in that: the temperature range in winter described in step 3 be less than 10 DEG C, the spring and
The temperature range in autumn is more than or equal to 10 DEG C and less than or equal to 22 DEG C, and the temperature range in the summer is 22 DEG C big.
Above-mentioned method, it is characterised in that: spring described in step 3 and autumn visibility detection model, summer visibility
Detection model and winter visibility detection model are the model of the BP neural network building based on simulated annealing optimization, the spring
The construction method of it and autumn visibility detection model, summer visibility detection model or winter visibility detection model are as follows:
Step 301, data classification and storage: by the historical data of particle concentration, temperature, humidity and visibility according to season
Section is divided into particle concentration, temperature, humidity and the visibility historical data in spring and autumn, the particle concentration in summer, temperature,
Humidity and visibility historical data and the particle concentration in winter, temperature, humidity and visibility historical data, and store and arrive
In computer;
Step 302, data normalization processing: computer is in MATLAB software according to formula α*=(ymax-ymin)*(α-
αmin)/(αmax-αmin)+yminRespectively to the particle concentration in spring and autumn, temperature, humidity and visibility historical data, summer
Particle concentration, temperature, humidity and visibility historical data and the particle concentration in winter, temperature, humidity and visibility
Historical data is normalized, particle concentration, temperature, humidity and the visibility in spring and autumn after being normalized
Historical data, particle concentration, temperature, humidity and the visibility historical data and the particle concentration in winter, temperature in summer
Degree, humidity and visibility historical data;Wherein, ymax=1, ymin=-1, α is the variable for needing to be normalized, αminTo need
The minimum value of the corresponding historical data of the variable being normalized, αmaxTo need the corresponding history number of variable being normalized
According to maximum value, α*Value after the variable normalization being normalized for needs;
Step 303 establishes the variable number of three layers of BP neural network of hidden layer neuron: computer is in MATLAB software
The input of particle concentration value, temperature value and humidity value as BP neural network after normalizing, input layer number n1It is 3
It is a, using visibility value as the output of BP neural network, output layer number of nodes n3It is 1, according to formulaDetermine the node in hidden layer n of three layers of BP network2, establish three layers of BP neural network;Wherein,
Taking a is 1~10 natural number;
Three layers of BP neural network of step 304, each different node in hidden layer of training, detailed process are as follows:
Step 3041, firstly, computer in MATLAB software by the particle concentration in spring and autumn after normalization,
Input of the temperature and humidity historical data as three layers of BP neural network, and with the particle concentration with spring and autumn, temperature
Output of the visibility historical data corresponding with humidity history as BP neural network constructs training sample;Then, it calculates
Machine is trained three layers of BP neural network of corresponding different node in hidden layer when to take a be 1~10 natural number, and
Called during being trained simulated annealing parameter optimization module to the weight W and threshold value B of three layers of BP neural network into
Row optimization, obtains trained three layers of BP neural network weight W and threshold value B optimal when each different node in hidden layer;
Step 3042, firstly, computer in MATLAB software by the particle concentration in the summer after normalization, temperature and
Input of the humidity history as three layers of BP neural network, and with particle concentration, the temperature and humidity history number with summer
Output according to corresponding visibility historical data as BP neural network constructs training sample;Then, computer to take a be 1~
Three layers of BP neural network of corresponding different node in hidden layer are trained when 10 natural number, and in the mistake being trained
It calls simulated annealing parameter optimization module to optimize the weight W and threshold value B of three layers of BP neural network in journey, obtains each
Weight W and threshold value B optimal trained three layers of BP neural network when a difference node in hidden layer;
Step 3043, firstly, computer in MATLAB software by the particle concentration in the winter after normalization, temperature and
Input of the humidity history as three layers of BP neural network, and with particle concentration, the temperature and humidity history number with winter
Output according to corresponding visibility historical data as BP neural network constructs training sample;Then, computer to take a be 1~
Three layers of BP neural network of corresponding different node in hidden layer are trained when 10 natural number, and in the mistake being trained
It calls simulated annealing parameter optimization module to optimize the weight W and threshold value B of three layers of BP neural network in journey, obtains each
Weight W and threshold value B optimal trained three layers of BP neural network when a difference node in hidden layer;
Step 305 determines spring and autumn visibility detection model, summer visibility detection model or winter visibility
Detection model, detailed process are as follows:
Step 3051, computer call network error computing module to calculate each different node in hidden layer in step 3041
When the weight W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and select network error minimum
Node in hidden layer the optimal three layers of BP neural network of weight W and threshold value B, be determined as trained three layers of BP nerve net
Network, and it is defined as spring and autumn visibility detection model;
Step 3052, computer call network error computing module to calculate each different node in hidden layer in step 3042
When the weight W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and select network error minimum
Node in hidden layer the optimal three layers of BP neural network of weight W and threshold value B, be determined as trained three layers of BP nerve net
Network, and it is defined as summer visibility detection model;
Step 3053, computer call network error computing module to calculate each different node in hidden layer in step 3043
When the weight W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and select network error minimum
Node in hidden layer the optimal three layers of BP neural network of weight W and threshold value B, be determined as trained three layers of BP nerve net
Network, and it is defined as winter visibility detection model.
Above-mentioned method, it is characterised in that: step 3041, step 3042 and step 3043 are carrying out three layers of BP neural network
In trained process, computer calls simulated annealing parameter optimization module to the weight W and threshold value B of three layers of BP neural network
The detailed process optimized are as follows:
Step A, the solution for defining optimization problem is the weight W and threshold value B of three layers of BP neural network, and selects three layers of BP nerve
The weight W and threshold value B of network are combined into function, are indicated with s;
Step B, three layers of BP neural network are calculated according to formula s (i+1)=s (i)+(0.2rand-0.1) errorBP
Weight W and threshold value B new solution s (i+1);Wherein, s (i) is the new solution of the weight W and threshold value B of three layers of BP neural network
Previous solution, rand be 0~1 random number, errorBP be BP neural network network error;
Step C, compare s (i+1) and s (i), as s (i+1) < s (i), receiving s (i+1) is three layers of BP neural network
The new solution of weight W and threshold value B;Otherwise, as s (i+1) >=s (i), s (i is received with probability exp [- (s (i+1)-s (i))/T]
It+1) is the new solution of the weight W and threshold value B of three layers of BP neural network;Wherein, T is the Current Temperatures of simulated annealing, and exp is certainly
Right index;
Step D, step B and step C is repeated, until the temperature of simulated annealing has reached preset simulated annealing and terminated temperature
Degree, has obtained the globally optimal solution of the weight W and threshold value B of three layers of BP neural network.
Above-mentioned method, it is characterised in that: the n2Value be 10.
Compared with the prior art, the present invention has the following advantages:
1, the present invention a kind of novel can be shown in integrated formed of particle concentration sensor and Temperature Humidity Sensor
Sensor to be spent, using particle concentration and temperature and humidity as input, obtains visibility, the circuit structure of visibility sensor is simple,
Novel in design reasonable, it is convenient and at low cost to realize.
2, microprocessor module of the invention uses single-chip microcontroller MSP430F169, and low in energy consumption, data-handling capacity is strong, can
Data prediction is realized well and carries out data processing obtains visibility detected value.
3, the historical data of particle concentration, temperature, humidity and visibility is divided into spring and autumn according to season by the present invention
It historical data, the historical data in summer and the historical data in winter, are respectively trained with the data of these three parts and are based on
Simulated annealing optimization neural network, establish with particle concentration, temperature and humidity for input, visibility be export spring and
Autumn visibility detection model, summer visibility detection model and winter visibility detection model, then pass particle concentration
Sensor and Temperature Humidity Sensor are integrated, detect visibility throughout the year using established visibility detection model,
Method and step is simple, and detection accuracy is high, and detection effect is good.
4, the present invention introduces simulated annealing in the weight and threshold adjustment of BP neural network, keeps BP refreshing
Weight and threshold value after terminating through network training is optimal, so that output can preferably approach actual value, overcomes BP nerve
The shortcomings that network is easily trapped into local optimum when finding parameter, can obtain the higher visibility detected value of precision.
5, the present invention is establishing spring and autumn visibility detection model, summer visibility detection model and winter visibility
When detection model, the variable number of three layers of BP neural network of hidden layer neuron is established, and it is the smallest to select network error
The three layers of weight W and threshold value B of node in hidden layer optimal BP neural network are determined as trained three layers of BP neural network,
The visibility detection accuracy of visibility sensor can be further increased.
6, the present invention can be advantageously applied to visibility detection, and the visibility detected according to the present invention carries out street lamp control
System, can be realized the intelligence of street lamp, is conducive to the trip in people's evening, ensure that the normal operation of communications and transportation, the present invention
It is practical, using effect is good, convenient for promote the use of.
In conclusion novel design of the invention is rationally, it is convenient and at low cost to realize, visibility detection accuracy is high, can
It is advantageously applied to visibility detection, practical, using effect is good, convenient for promoting the use of.
Below by drawings and examples, technical scheme of the present invention will be described in further detail.
Detailed description of the invention
Fig. 1 is that the present invention is based on the schematic block circuit diagrams of the visibility sensor of simulated annealing optimization neural network.
Fig. 2 is the circuit diagram of power circuit of the present invention.
Fig. 3 is the circuit diagram of microprocessor module of the present invention.
Fig. 4 is the circuit diagram of visibility detection circuit of the present invention.
Fig. 5 is that the present invention is based on the method flows of the detection method of the visibility sensor of simulated annealing optimization neural network
Block diagram.
Description of symbols:
1-microprocessor module;2-power circuits;3-visibility detection circuits;
3-1-particle concentration sensor;3-2-Temperature Humidity Sensor.
Specific embodiment
As shown in Figure 1, the visibility sensor of the invention based on simulated annealing optimization neural network, including microprocessor
Module 1 and power circuit 2 for electricity consumption module for power supply each in the visibility sensor, the input of the microprocessor module 1
It is terminated with visibility detection circuit 3, the visibility detection circuit 3 includes particle concentration sensor 3-1 and temperature and humidity sensing
Device 3-2, the microprocessor module 1, particle concentration sensor 3-1 and Temperature Humidity Sensor 3-2 is defeated with power circuit 2
Outlet connection.
In the present embodiment, as shown in Fig. 2, the power circuit 2 includes 5V battery, voltage stabilizing chip AMS1117, nonpolarity electricity
It is defeated to hold C8, nonpolarity capacitor C9 and polar capacitor C10, the 5V voltage of the 3rd pin and 5V battery of the voltage stabilizing chip AMS1117
Outlet connection, and be grounded by nonpolarity capacitor C8, the 5V voltage output end of the 5V battery is that the 5V voltage of power circuit 2 is defeated
Outlet, the 2nd pin of the voltage stabilizing chip AMS1117 is the 3.3V voltage output end of power circuit 2, and passes through non-pole in parallel
Property capacitor C9 and polar capacitor C10 ground connection, the voltage stabilizing chip AMS1117 the 1st pin ground connection.
In the present embodiment, as shown in figure 3, the microprocessor module 1 includes single-chip microcontroller MSP430F169, the first crystal oscillator electricity
Road, the second crystal oscillating circuit and reset circuit, first crystal oscillating circuit include crystal oscillator Y0, nonpolarity capacitor C1 and nonpolar capacitor
One end of C2, the crystal oscillator Y0 and one end of nonpolar capacitor C1 are connect with the 53rd pin of single-chip microcontroller MSP430F169, institute
One end of the other end and nonpolar capacitor C2 of stating crystal oscillator Y0 is connect with the 52nd pin of single-chip microcontroller MSP430F169, described non-
The other end of the other end of polar capacitor C1 and nonpolar capacitor C2 are grounded;Second crystal oscillating circuit includes crystal oscillator Y1, non-
One end of one end and nonpolar capacitor C6 of polar capacitor C6, nonpolarity capacitor C7 and nonpolar capacitor C8, the crystal oscillator Y1 is equal
Connect with the 8th pin of single-chip microcontroller MSP430F169, one end of the other end of the crystal oscillator Y1 and nonpolar capacitor C7 with list
The 9th pin of piece machine MSP430F169 connects, one end of the nonpolarity capacitor C8 and the 7th pin of single-chip microcontroller MSP430F169
The other end of connection, the other end of the nonpolarity capacitor C6, the other end of nonpolar capacitor C7 and nonpolar capacitor C8 connects
Ground;The reset circuit includes reset key S0, resistance R0 and nonpolar capacitor C0, one end of the reset key S0, resistance
One end of R0 and one end of polar capacitor C0 are connect with the 58th pin of single-chip microcontroller MSP430F169, the reset key S0's
The other end of the other end and polar capacitor C0 are grounded, the other end of the resistance R0 and the 3.3V voltage output of power circuit 2
End connection.
In the present embodiment, as shown in figure 4, the particle concentration sensor 3-1 includes particle concentration sensor
PMS3003, the Temperature Humidity Sensor 3-2 include Temperature Humidity Sensor DHT11, the particle concentration sensor PMS3003
Power end pin and the power end pin of Temperature Humidity Sensor DHT11 connect with the 5V voltage output end of power circuit 2, institute
The ground terminal pin of the ground terminal pin and Temperature Humidity Sensor DHT11 of stating particle concentration sensor PMS3003 is grounded, institute
The signal output end pin for stating particle concentration sensor PMS3003 is connect with the 33rd pin of single-chip microcontroller MSP430F169, institute
The signal output end pin for stating Temperature Humidity Sensor DHT11 is connect with the 36th pin of single-chip microcontroller MSP430F169, and passes through electricity
Resistance R1 is connect with the 5V voltage output end of power circuit 2.
Wherein, for resistance R1 for dividing, the temperature degree signal stabilization for being easy to implement Temperature Humidity Sensor DHT11 detection is reliable
Ground is transferred to single-chip microcontroller MSP430F169.
In the present embodiment, the BP neural network building based on simulated annealing optimization is stored in the microprocessor module 1
Spring and autumn visibility detection model, summer visibility detection model and winter visibility detection model.
As shown in figure 5, the visibility of the visibility sensor of the invention based on simulated annealing optimization neural network detects
Method, comprising the following steps:
Step 1: data acquisition and transmission: particle concentration sensor 3-1 carries out the particle concentration in environment real-time
Acquisition, and by the collected particle concentration real-time data transmission of institute to microprocessor module 1;Temperature Humidity Sensor 3-2 is to ring
Temperature and humidity in border is acquired in real time, and by the collected temperature data of institute and humidity data real-time Transmission to micro process
Device module 1;
Step 2: data prediction: the microprocessor module 1 is according to formula x*=(ymax-ymin)*(x-xmin)/
(xmax-xmin)+yminReceived particle concentration data are normalized, the particulate matter after being normalized is dense
Angle value x*;According to formula z*=(ymax-ymin)*(z-zmin)/(zmax-zmin)+yminNormalizing is carried out to received temperature data
Change processing, the temperature value z after being normalized*;And according to formula r*=(ymax-ymin)*(r-rmin)/(rmax-rmin)+yminIt is right
Received humidity data is normalized, the humidity value r after being normalized*;Wherein, ymax=1, ymin=-1, x
For the particle concentration value that microprocessor module 1 is currently received, xminThe particle concentration received for microprocessor module 1
The minimum value of value, xmaxFor the maximum value for the particle concentration value that microprocessor module 1 receives;Z is that microprocessor module 1 is worked as
Before the temperature value that receives, zminFor the minimum value for the temperature value that microprocessor module 1 receives, zmaxFor microprocessor module 1
The maximum value of the temperature value received;R is the humidity value that microprocessor module 1 is currently received, rminFor microprocessor module 1
The minimum value of the humidity value received, rmaxFor the maximum value for the humidity value that microprocessor module 1 receives;
Step 3: data processing obtains visibility detected value: the microprocessor module 1 judges that microprocessor module 1 exists
Temperature range locating for the daily mean of the preceding M days temperature values received of current detection, the current M days temperature value received
When daily mean was within the temperature range of winter, the microprocessor module 1 is by the particle concentration value x after normalization*, temperature
Angle value z*With humidity value r*Inputting the winter according to particle concentration, temperature and humidity detection visibility being stored therein can see
It spends in detection model, show that the output of the winter visibility detection model, the output of the winter visibility detection model are
For visibility detected value;When the daily mean of the current M days temperature values received was within the temperature range of spring and autumn, institute
Microprocessor module 1 is stated by the particle concentration value x after normalization*, temperature value z*With humidity value r*Input the root being stored therein
According to particle concentration, temperature and humidity detection visibility spring and autumn visibility detection model in, obtain the spring and
The output of autumn visibility detection model, the output of the spring and autumn visibility detection model are visibility detected value;
When the daily mean of the current M days temperature values received was within the temperature range of summer, the microprocessor module 1 is by normalizing
Particle concentration value x after change*, temperature value z*With humidity value r*Input be stored therein according to particle concentration, temperature and wet
In the summer visibility detection model of degree detection visibility, the output of the summer visibility detection model, the summer are obtained
The output of visibility detection model is visibility detected value.Wherein, the positive integer that the value of M is 3~30.
In the present embodiment, the temperature range in winter described in step 3 is the temperature in the spring and autumn less than 10 DEG C
Range is more than or equal to 10 DEG C and less than or equal to 22 DEG C, and the temperature range in the summer is 22 DEG C big.
In the present embodiment, spring described in step 3 and autumn visibility detection model, summer visibility detection model and
Winter visibility detection model is the model of the BP neural network building based on simulated annealing optimization, the spring and autumn energy
The construction method of degree of opinion detection model, summer visibility detection model or winter visibility detection model are as follows:
Step 301, data classification and storage: by the historical data of particle concentration, temperature, humidity and visibility according to season
Section is divided into particle concentration, temperature, humidity and the visibility historical data in spring and autumn, the particle concentration in summer, temperature,
Humidity and visibility historical data and the particle concentration in winter, temperature, humidity and visibility historical data, and store and arrive
In computer;
Step 302, data normalization processing: computer is in MATLAB software according to formula α*=(ymax-ymin)*(α-
αmin)/(αmax-αmin)+yminRespectively to the particle concentration in spring and autumn, temperature, humidity and visibility historical data, summer
Particle concentration, temperature, humidity and visibility historical data and the particle concentration in winter, temperature, humidity and visibility
Historical data is normalized, particle concentration, temperature, humidity and the visibility in spring and autumn after being normalized
Historical data, particle concentration, temperature, humidity and the visibility historical data and the particle concentration in winter, temperature in summer
Degree, humidity and visibility historical data;Wherein, ymax=1, ymin=-1, α be need be normalized variable (i.e. spring and
Particle concentration, temperature, humidity or the visibility historical data in autumn, particle concentration, temperature, humidity or the visibility in summer
The particle concentration in historical data and winter, temperature, humidity or visibility historical data), αminTo need to be normalized
The corresponding historical data of variable minimum value, αmaxMaximum value for the corresponding historical data of variable for needing to be normalized,
α*Value after the variable normalization being normalized for needs;
Step 303 establishes the variable number of three layers of BP neural network of hidden layer neuron: computer is in MATLAB software
The input of particle concentration value, temperature value and humidity value as BP neural network after normalizing, input layer number n1It is 3
It is a, using visibility value as the output of BP neural network, output layer number of nodes n3It is 1, according to formulaDetermine the node in hidden layer n of three layers of BP network2, establish three layers of BP neural network;Wherein,
Taking a is 1~10 natural number;
In the present embodiment, the n2Value be 10.
Three layers of BP neural network of step 304, each different node in hidden layer of training, detailed process are as follows:
Step 3041, firstly, computer in MATLAB software by the particle concentration in spring and autumn after normalization,
Input of the temperature and humidity historical data as three layers of BP neural network, and with the particle concentration with spring and autumn, temperature
Output of the visibility historical data corresponding with humidity history as BP neural network constructs training sample;Then, it calculates
Machine is trained three layers of BP neural network of corresponding different node in hidden layer when to take a be 1~10 natural number, and
Called during being trained simulated annealing parameter optimization module to the weight W and threshold value B of three layers of BP neural network into
Row optimization, obtains trained three layers of BP neural network weight W and threshold value B optimal when each different node in hidden layer;
Step 3042, firstly, computer in MATLAB software by the particle concentration in the summer after normalization, temperature and
Input of the humidity history as three layers of BP neural network, and with particle concentration, the temperature and humidity history number with summer
Output according to corresponding visibility historical data as BP neural network constructs training sample;Then, computer to take a be 1~
Three layers of BP neural network of corresponding different node in hidden layer are trained when 10 natural number, and in the mistake being trained
It calls simulated annealing parameter optimization module to optimize the weight W and threshold value B of three layers of BP neural network in journey, obtains each
Weight W and threshold value B optimal trained three layers of BP neural network when a difference node in hidden layer;
Step 3043, firstly, computer in MATLAB software by the particle concentration in the winter after normalization, temperature and
Input of the humidity history as three layers of BP neural network, and with particle concentration, the temperature and humidity history number with winter
Output according to corresponding visibility historical data as BP neural network constructs training sample;Then, computer to take a be 1~
Three layers of BP neural network of corresponding different node in hidden layer are trained when 10 natural number, and in the mistake being trained
It calls simulated annealing parameter optimization module to optimize the weight W and threshold value B of three layers of BP neural network in journey, obtains each
Weight W and threshold value B optimal trained three layers of BP neural network when a difference node in hidden layer;
Step 305 determines spring and autumn visibility detection model, summer visibility detection model or winter visibility
Detection model, detailed process are as follows:
Step 3051, computer call network error computing module to calculate each different node in hidden layer in step 3041
When the weight W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and select network error minimum
Node in hidden layer the optimal three layers of BP neural network of weight W and threshold value B, be determined as trained three layers of BP nerve net
Network, and it is defined as spring and autumn visibility detection model;
Step 3052, computer call network error computing module to calculate each different node in hidden layer in step 3042
When the weight W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and select network error minimum
Node in hidden layer the optimal three layers of BP neural network of weight W and threshold value B, be determined as trained three layers of BP nerve net
Network, and it is defined as summer visibility detection model;
Step 3053, computer call network error computing module to calculate each different node in hidden layer in step 3043
When the weight W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and select network error minimum
Node in hidden layer the optimal three layers of BP neural network of weight W and threshold value B, be determined as trained three layers of BP nerve net
Network, and it is defined as winter visibility detection model.
In the present embodiment, step 3041, step 3042 and step 3043 are in the process for carrying out the training of three layers of BP neural network
In, computer calls simulated annealing parameter optimization module to optimize the weight W and threshold value B of three layers of BP neural network
Detailed process are as follows:
Step A, the solution for defining optimization problem is the weight W and threshold value B of three layers of BP neural network, and selects three layers of BP nerve
The weight W and threshold value B of network are combined into function, are indicated with s;
Step B, three layers of BP neural network are calculated according to formula s (i+1)=s (i)+(0.2rand-0.1) errorBP
Weight W and threshold value B new solution s (i+1), i.e. the solution of i+1 time;Wherein, s (i) be three layers of BP neural network weight W and
The previous solution of the new solution of threshold value B, the i.e. solution of i-th, the random number that rand is 0~1, errorBP is BP neural network
Network error;
Step C, compare s (i+1) and s (i), as s (i+1) < s (i), receiving s (i+1) is three layers of BP neural network
The new solution of weight W and threshold value B;Otherwise, as s (i+1) >=s (i), s (i is received with probability exp [- (s (i+1)-s (i))/T]
It+1) is the new solution of the weight W and threshold value B of three layers of BP neural network;Wherein, T is the Current Temperatures of simulated annealing, and exp is certainly
Right index;
Step D, step B and step C is repeated, until the temperature of simulated annealing has reached preset simulated annealing and terminated temperature
Degree, has obtained the globally optimal solution of the weight W and threshold value B of three layers of BP neural network.
When it is implemented, the input of hidden layer neuron is the sum of the weighting of all inputs in three layers of BP neural network, i.e.,Wherein, xi′For the i-th ' a input of three layers of BP neural network, wi′j′For three layers of BP neural network
The i-th ' a input to the weight of a hidden layer neuron of j ', bj′For hidden layer jth ' a neuron threshold value;Hidden layer
The output x ' of a neuron of j 'j′X is excited using S functionj′, obtainOutput layer neuron
Output isWherein, y ' is the output of output neuron, wj′oFor hidden layer jth ' a neuron pair
The weight of output layer neuron, b are the threshold value of output layer neuron.W is wi′j′With wj′oSet, B bj′With the set of b.
The above is only presently preferred embodiments of the present invention, is not intended to limit the invention in any way, it is all according to the present invention
Technical spirit any simple modification to the above embodiments, change and equivalent structural changes, still fall within skill of the present invention
In the protection scope of art scheme.
Claims (10)
1. a kind of visibility sensor based on simulated annealing optimization neural network, it is characterised in that: including microprocessor module
(1) and for electricity consumption module for power supply each in the visibility sensor power circuit (2), the microprocessor module (1) it is defeated
Enter to be terminated with visibility detection circuit (3), the visibility detection circuit (3) includes particle concentration sensor (3-1) and temperature
Humidity sensor (3-2), the microprocessor module (1), particle concentration sensor (3-1) and Temperature Humidity Sensor (3-2)
It is connect with the output end of power circuit (2).
2. the visibility sensor described in accordance with the claim 1 based on simulated annealing optimization neural network, it is characterised in that: institute
Stating power circuit (2) includes 5V battery, voltage stabilizing chip AMS1117, nonpolar capacitor C8, nonpolarity capacitor C9 and polar capacitor
The 3rd pin of C10, the voltage stabilizing chip AMS1117 are connect with the 5V voltage output end of 5V battery, and pass through nonpolar capacitor C8
Ground connection, the 5V voltage output end of the 5V battery are the 5V voltage output end of power circuit (2), the voltage stabilizing chip AMS1117
The 2nd pin be the 3.3V voltage output end of power circuit (2), and pass through in parallel nonpolar capacitor C9 and polar capacitor C10
Ground connection, the 1st pin ground connection of the voltage stabilizing chip AMS1117.
3. the visibility sensor described in accordance with the claim 1 based on simulated annealing optimization neural network, it is characterised in that: institute
Stating microprocessor module (1) includes single-chip microcontroller MSP430F169, the first crystal oscillating circuit, the second crystal oscillating circuit and reset circuit, institute
Stating the first crystal oscillating circuit includes crystal oscillator Y0, nonpolarity capacitor C1 and nonpolar capacitor C2, one end of the crystal oscillator Y0 and nonpolarity
One end of capacitor C1 is connect with the 53rd pin of single-chip microcontroller MSP430F169, the other end of the crystal oscillator Y0 and nonpolar capacitor
One end of C2 is connect with the 52nd pin of single-chip microcontroller MSP430F169, the other end and nonpolarity electricity of the nonpolarity capacitor C1
The other end for holding C2 is grounded;Second crystal oscillating circuit includes crystal oscillator Y1, nonpolar capacitor C6, nonpolarity capacitor C7 and non-pole
Property capacitor C8, one end of one end of the crystal oscillator Y1 and nonpolar capacitor C6 connect with the 8th pin of single-chip microcontroller MSP430F169
It connects, one end of the other end of the crystal oscillator Y1 and nonpolar capacitor C7 are connect with the 9th pin of single-chip microcontroller MSP430F169, institute
The one end for stating nonpolar capacitor C8 is connect with the 7th pin of single-chip microcontroller MSP430F169, the other end of the nonpolarity capacitor C6,
The other end of the other end of nonpolar capacitor C7 and nonpolar capacitor C8 are grounded;The reset circuit include reset key S0,
Resistance R0 and nonpolar capacitor C0, one end of one end of the reset key S0, one end of resistance R0 and polar capacitor C0 with
The 58th pin of single-chip microcontroller MSP430F169 connects, and the other end of the reset key S0 and the other end of polar capacitor C0 connect
The other end on ground, the resistance R0 is connect with the 3.3V voltage output end of power circuit (2).
4. the visibility sensor described in accordance with the claim 3 based on simulated annealing optimization neural network, it is characterised in that: institute
Stating particle concentration sensor (3-1) includes particle concentration sensor PMS3003, and the Temperature Humidity Sensor (3-2) includes
The power end pin and Temperature Humidity Sensor DHT11 of Temperature Humidity Sensor DHT11, the particle concentration sensor PMS3003
Power end pin connect with the 5V voltage output end of power circuit (2), the particle concentration sensor PMS3003's connects
The ground terminal pin of ground terminal pin and Temperature Humidity Sensor DHT11 are grounded, the letter of the particle concentration sensor PMS3003
Number output pin is connect with the 33rd pin of single-chip microcontroller MSP430F169, the signal output of the Temperature Humidity Sensor DHT11
End pin is connect with the 36th pin of single-chip microcontroller MSP430F169, and passes through the 5V voltage output of resistance R1 and power circuit (2)
End connection.
5. the visibility sensor described in accordance with the claim 1 based on simulated annealing optimization neural network, it is characterised in that: institute
State the spring that the building of the BP neural network based on simulated annealing optimization is stored in microprocessor module (1) and the inspection of autumn visibility
Survey model, summer visibility detection model and winter visibility detection model.
6. a kind of visibility detection side based on simulated annealing optimization neural network of visibility sensor as described in claim 1
Method, which is characterized in that method includes the following steps:
Step 1: data acquisition and transmission: particle concentration sensor (3-1) adopts the particle concentration in environment in real time
Collection, and give the collected particle concentration real-time data transmission of institute to microprocessor module (1);Temperature Humidity Sensor (3-2) is right
Temperature and humidity in environment is acquired in real time, and gives the collected temperature data of institute and humidity data real-time Transmission to micro- place
It manages device module (1);
Step 2: data prediction: the microprocessor module (1) is according to formula x*=(ymax-ymin)*(x-xmin)/(xmax-
xmin)+yminReceived particle concentration data are normalized, the particle concentration value after being normalized
x*;According to formula z*=(ymax-ymin)*(z-zmin)/(zmax-zmin)+yminPlace is normalized to received temperature data
Reason, the temperature value z after being normalized*;And according to formula r*=(ymax-ymin)*(r-rmin)/(rmax-rmin)+yminIt is connect
The humidity data received is normalized, the humidity value r after being normalized*;Wherein, ymax=1, ymin=-1, x is micro-
The particle concentration value that processor module (1) is currently received, xminThe particle concentration received for microprocessor module (1)
The minimum value of value, xmaxFor the maximum value for the particle concentration value that microprocessor module (1) receives;Z is microprocessor module
(1) temperature value being currently received, zminFor the minimum value for the temperature value that microprocessor module (1) receives, zmaxFor micro process
The maximum value for the temperature value that device module (1) receives;R is the humidity value that microprocessor module (1) is currently received, rminIt is micro-
The minimum value for the humidity value that processor module (1) receives, rmaxFor the maximum for the humidity value that microprocessor module (1) receives
Value;
Step 3: data processing obtains visibility detected value: the microprocessor module (1) judges that microprocessor module (1) exists
Temperature range locating for the daily mean of the preceding M days temperature values received of current detection, the current M days temperature value received
When daily mean was within the temperature range of winter, the microprocessor module (1) is by the particle concentration value x after normalization*、
Temperature value z*With humidity value r*Input the winter energy that visibility is detected according to particle concentration, temperature and humidity being stored therein
In degree of opinion detection model, the output of the winter visibility detection model, the output of the winter visibility detection model are obtained
As visibility detected value;When the daily mean of the current M days temperature values received was within the temperature range of spring and autumn,
The microprocessor module (1) is by the particle concentration value x after normalization*, temperature value z*With humidity value r*Input is stored therein
According to particle concentration, temperature and humidity detection visibility spring and autumn visibility detection model in, obtain the spring
The output of it and autumn visibility detection model, the output of the spring and autumn visibility detection model are visibility detection
Value;When the daily mean of the current M days temperature values received was within the temperature range of summer, the microprocessor module (1)
By the particle concentration value x after normalization*, temperature value z*With humidity value r*Input be stored therein according to particle concentration, temperature
In degree and the summer visibility detection model of Humidity Detection visibility, the output of the summer visibility detection model, institute are obtained
The output for stating summer visibility detection model is visibility detected value.
7. according to the method for claim 6, it is characterised in that: the temperature range in winter described in step 3 is less than 10
DEG C, the temperature range in the spring and autumn is more than or equal to 10 DEG C and to be less than or equal to 22 DEG C, and the temperature range in the summer is
It is 22 DEG C big.
8. according to the method for claim 6, it is characterised in that: spring described in step 3 and autumn visibility detect mould
Type, summer visibility detection model and winter visibility detection model are the BP neural network building based on simulated annealing optimization
Model, the spring and autumn visibility detection model, summer visibility detection model or winter visibility detection model
Construction method are as follows:
Step 301, data classification and storage: by the historical data of particle concentration, temperature, humidity and visibility according to season point
For the particle concentration in spring and autumn, temperature, humidity and visibility historical data, the particle concentration in summer, temperature, humidity
With the particle concentration in visibility historical data and winter, temperature, humidity and visibility historical data, and store to calculate
In machine;
Step 302, data normalization processing: computer is in MATLAB software according to formula α*=(ymax-ymin)*(α-αmin)/
(αmax-αmin)+yminRespectively to the particle concentration in spring and autumn, temperature, humidity and visibility historical data, in summer
Grain object concentration, temperature, humidity and visibility historical data and the particle concentration in winter, temperature, humidity and visibility history
Data are normalized, particle concentration, temperature, humidity and the visibility history in spring and autumn after being normalized
Data, it is particle concentration, temperature, humidity and the visibility historical data and the particle concentration in winter in summer, temperature, wet
Degree and visibility historical data;Wherein, ymax=1, ymin=-1, α is the variable for needing to be normalized, αminTo need to carry out
The minimum value of the corresponding historical data of normalized variable, αmaxTo need the variable corresponding historical data being normalized
Maximum value, α*Value after the variable normalization being normalized for needs;
Step 303 establishes the variable number of three layers of BP neural network of hidden layer neuron: computer is in MATLAB software to return
The input of particle concentration value, temperature value and humidity value as BP neural network after one change, input layer number n1It is 3,
Using visibility value as the output of BP neural network, output layer number of nodes n3It is 1, according to formula
Determine the node in hidden layer n of three layers of BP network2, establish three layers of BP neural network;Wherein, taking a is 1~10 nature
Number;
Three layers of BP neural network of step 304, each different node in hidden layer of training, detailed process are as follows:
Step 3041, firstly, computer in MATLAB software by the particle concentration in spring and autumn after normalization, temperature
Input with humidity history as three layers of BP neural network, and with the particle concentration with spring and autumn, temperature and wet
Output of the corresponding visibility historical data of historical data as BP neural network is spent, training sample is constructed;Then, computer pair
Three layers of BP neural network of corresponding different node in hidden layer are trained when to take a be 1~10 natural number, and are being carried out
Simulated annealing parameter optimization module is called to carry out the weight W and threshold value B of three layers of BP neural network in trained process excellent
Change, obtains trained three layers of BP neural network weight W and threshold value B optimal when each different node in hidden layer;
Step 3042, firstly, computer in MATLAB software by the particle concentration in the summer after normalization, temperature and humidity
Input of the historical data as three layers of BP neural network, and with the particle concentration with summer, temperature and humidity historical data pair
Output of the visibility historical data answered as BP neural network constructs training sample;Then, computer is 1~10 to a is taken
Three layers of BP neural network of corresponding different node in hidden layer are trained when natural number, and during being trained
Call simulated annealing parameter optimization module the weight W and threshold value B of three layers of BP neural network are optimized, obtain it is each not
With trained three layers of BP neural network that weight W when node in hidden layer and threshold value B are optimal;
Step 3043, firstly, computer in MATLAB software by the particle concentration in the winter after normalization, temperature and humidity
Input of the historical data as three layers of BP neural network, and with the particle concentration with winter, temperature and humidity historical data pair
Output of the visibility historical data answered as BP neural network constructs training sample;Then, computer is 1~10 to a is taken
Three layers of BP neural network of corresponding different node in hidden layer are trained when natural number, and during being trained
Call simulated annealing parameter optimization module the weight W and threshold value B of three layers of BP neural network are optimized, obtain it is each not
With trained three layers of BP neural network that weight W when node in hidden layer and threshold value B are optimal;
Step 305 determines spring and autumn visibility detection model, summer visibility detection model or the detection of winter visibility
Model, detailed process are as follows:
Step 3051, computer call network error computing module to weigh when calculating each different node in hidden layer in step 3041
The value W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and it is the smallest hidden to select network error
Three layers of the weight W and threshold value B of the number containing node layer optimal BP neural network are determined as trained three layers of BP neural network, and
It is defined as spring and autumn visibility detection model;
Step 3052, computer call network error computing module to weigh when calculating each different node in hidden layer in step 3042
The value W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and it is the smallest hidden to select network error
Three layers of the weight W and threshold value B of the number containing node layer optimal BP neural network are determined as trained three layers of BP neural network, and
It is defined as summer visibility detection model;
Step 3053, computer call network error computing module to weigh when calculating each different node in hidden layer in step 3043
The value W and threshold value B optimal corresponding network error of trained three layers of BP neural network, and it is the smallest hidden to select network error
Three layers of the weight W and threshold value B of the number containing node layer optimal BP neural network are determined as trained three layers of BP neural network, and
It is defined as winter visibility detection model.
9. according to the method for claim 8, it is characterised in that: step 3041, step 3042 and step 3043 are carrying out three
During layer BP neural network training, computer calls simulated annealing parameter optimization module to three layers of BP neural network
The detailed process that weight W and threshold value B are optimized are as follows:
Step A, the solution for defining optimization problem is the weight W and threshold value B of three layers of BP neural network, and selects three layers of BP neural network
Weight W and threshold value B be combined into function, indicated with s;
Step B, the power of three layers of BP neural network is calculated according to formula s (i+1)=s (i)+(0.2rand-0.1) errorBP
The new solution s (i+1) of value W and threshold value B;Wherein, before the new solution that s (i) is the weight W and threshold value B of three layers of BP neural network
Primary solution, the random number that rand is 0~1, errorBP are the network error of BP neural network;
Step C, compare s (i+1) and s (i), as s (i+1) < s (i), receive the weight W that s (i+1) is three layers of BP neural network
With the new solution of threshold value B;Otherwise, as s (i+1) >=s (i), receiving s (i+1) with probability exp [- (s (i+1)-s (i))/T] is
The new solution of the weight W and threshold value B of three layers of BP neural network;Wherein, T is the Current Temperatures of simulated annealing, and exp is refers to naturally
Number;
Step D, step B and step C is repeated, until the temperature of simulated annealing has reached preset simulated annealing and terminated temperature, is obtained
To the globally optimal solution of the weight W and threshold value B of three layers of BP neural network.
10. according to the method for claim 8, it is characterised in that: the n2Value be 10.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810931051.4A CN109000733B (en) | 2018-08-15 | 2018-08-15 | Visibility sensor detection method based on simulated annealing optimization neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810931051.4A CN109000733B (en) | 2018-08-15 | 2018-08-15 | Visibility sensor detection method based on simulated annealing optimization neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109000733A true CN109000733A (en) | 2018-12-14 |
CN109000733B CN109000733B (en) | 2021-11-26 |
Family
ID=64592790
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810931051.4A Active CN109000733B (en) | 2018-08-15 | 2018-08-15 | Visibility sensor detection method based on simulated annealing optimization neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109000733B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109668593A (en) * | 2018-12-26 | 2019-04-23 | 西安清宇网络科技有限公司 | A kind of integrated sensor, Illumination adjusting device and adjusting method for street lamp |
CN109738808A (en) * | 2019-01-03 | 2019-05-10 | 温州大学 | The method of SOC is estimated based on the BP neural network after Simulated Anneal Algorithm Optimize |
CN110441374A (en) * | 2019-08-08 | 2019-11-12 | 南京邮电大学 | A kind of dissolved oxygen sensor compensation method based on SA-BP neural network |
CN110717422A (en) * | 2019-09-25 | 2020-01-21 | 北京影谱科技股份有限公司 | Method and system for identifying interactive action based on convolutional neural network |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102262092A (en) * | 2011-04-29 | 2011-11-30 | 中国气象局北京城市气象研究所 | Visibility measurement system and method |
CN104700153A (en) * | 2014-12-05 | 2015-06-10 | 江南大学 | PH (potential of hydrogen) value predicting method of BP (back propagation) neutral network based on simulated annealing optimization |
US20160307096A1 (en) * | 2015-04-19 | 2016-10-20 | International Business Machines Corporation | Annealed dropout training of neural networks |
CN207197576U (en) * | 2017-09-11 | 2018-04-06 | 西安邮电大学 | Home environment quality monitoring device based on WIFI communications |
US20180100843A1 (en) * | 2016-10-11 | 2018-04-12 | International Business Machines Corporation | Estimation of concentration of particulate matter in atmosphere |
CN108171318A (en) * | 2017-11-30 | 2018-06-15 | 河南大学 | One kind is based on the convolutional neural networks integrated approach of simulated annealing-Gaussian function |
-
2018
- 2018-08-15 CN CN201810931051.4A patent/CN109000733B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102262092A (en) * | 2011-04-29 | 2011-11-30 | 中国气象局北京城市气象研究所 | Visibility measurement system and method |
CN104700153A (en) * | 2014-12-05 | 2015-06-10 | 江南大学 | PH (potential of hydrogen) value predicting method of BP (back propagation) neutral network based on simulated annealing optimization |
US20160307096A1 (en) * | 2015-04-19 | 2016-10-20 | International Business Machines Corporation | Annealed dropout training of neural networks |
US20180100843A1 (en) * | 2016-10-11 | 2018-04-12 | International Business Machines Corporation | Estimation of concentration of particulate matter in atmosphere |
CN207197576U (en) * | 2017-09-11 | 2018-04-06 | 西安邮电大学 | Home environment quality monitoring device based on WIFI communications |
CN108171318A (en) * | 2017-11-30 | 2018-06-15 | 河南大学 | One kind is based on the convolutional neural networks integrated approach of simulated annealing-Gaussian function |
Non-Patent Citations (2)
Title |
---|
HONGGUANG PAN等: "Air Visibility Prediction Based on Multiple Models", 《PROCEEDINGS OF 2018 IEEE 8TH ANNUAL INTERNATIONAL CONFERENCE ON CYBER TECHNOLOGY IN AUTOMATION, CONTROL, AND INTELLIGENT SYSTEMS》 * |
路婵: "大气污染物与气象条件对长沙城市能见度的影响分析", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109668593A (en) * | 2018-12-26 | 2019-04-23 | 西安清宇网络科技有限公司 | A kind of integrated sensor, Illumination adjusting device and adjusting method for street lamp |
CN109738808A (en) * | 2019-01-03 | 2019-05-10 | 温州大学 | The method of SOC is estimated based on the BP neural network after Simulated Anneal Algorithm Optimize |
CN110441374A (en) * | 2019-08-08 | 2019-11-12 | 南京邮电大学 | A kind of dissolved oxygen sensor compensation method based on SA-BP neural network |
CN110441374B (en) * | 2019-08-08 | 2021-12-07 | 南京邮电大学 | Dissolved oxygen sensor compensation method based on SA-BP neural network |
CN110717422A (en) * | 2019-09-25 | 2020-01-21 | 北京影谱科技股份有限公司 | Method and system for identifying interactive action based on convolutional neural network |
Also Published As
Publication number | Publication date |
---|---|
CN109000733B (en) | 2021-11-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109000733A (en) | Visibility sensor and its detection method based on simulated annealing optimization neural network | |
CN103267832A (en) | Online water quality monitoring system for surface water | |
CN103323749B (en) | The partial discharge diagnostic method of multi-classifier information fusion | |
CN108510006A (en) | A kind of analysis of business electrical amount and prediction technique based on data mining | |
CN107169145A (en) | A kind of method of user's stealing menace level detection based on clustering algorithm | |
CN109711609A (en) | Photovoltaic plant output power predicting method based on wavelet transformation and extreme learning machine | |
CN110045209A (en) | Detection method, device, equipment and the readable storage medium storing program for executing of electricity consumption data exception | |
CN104463168B (en) | A kind of useless house site information automation extraction method of sky based on remote sensing image | |
CN107203752A (en) | A kind of combined depth study and the face identification method of the norm constraint of feature two | |
WO2019214016A1 (en) | Lora technology-based multi-functional led smart street lamp system | |
CN105445283A (en) | Detection method for filthy conditions of insulator images | |
CN102902985A (en) | Coastal water quality evaluation method based on two-classification support vector machines and particle swarm algorithm | |
CN109002926A (en) | The photovoltaic power generation quantity prediction model and its construction method of a kind of high accuracy and application | |
CN108011367A (en) | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms | |
CN103886316A (en) | Combustion monitoring and diagnosis method based on feature extraction and fuzzy C-means cluster | |
CN107025468A (en) | Highway congestion recognition methods based on PCA GA SVM algorithms | |
CN109165770A (en) | Novel photovoltaic power forecasting method based on AFSA-Elman | |
CN113780402A (en) | User electricity stealing detection method based on improved generation countermeasure network | |
CN109085099A (en) | The application of visibility sensor, visibility detecting method and visibility sensor | |
CN102496062B (en) | Personnel information fusion method based on Spiking neural network | |
Lin et al. | On-line monitoring and fault diagnosis of PV array based on BP neural network optimized by genetic algorithm | |
CN106340874A (en) | Identification decision-making method and system for power load decomposition | |
CN103942574B (en) | 3D-handwritten-recognition SVM classifier nuclear-parameter selection method and purpose thereof | |
CN115273645A (en) | Map making method for automatically clustering indoor surface elements | |
Li et al. | Anormaly intrusion detection based on SOM |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |