CN111783940A - Method for reducing fire false alarm rate based on GA-BP neural network algorithm - Google Patents

Method for reducing fire false alarm rate based on GA-BP neural network algorithm Download PDF

Info

Publication number
CN111783940A
CN111783940A CN202010505476.6A CN202010505476A CN111783940A CN 111783940 A CN111783940 A CN 111783940A CN 202010505476 A CN202010505476 A CN 202010505476A CN 111783940 A CN111783940 A CN 111783940A
Authority
CN
China
Prior art keywords
neural network
weight
output
hidden layer
bias
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.)
Pending
Application number
CN202010505476.6A
Other languages
Chinese (zh)
Inventor
文春明
李科畅
廖义奎
黄天星
李大庆
罗丽平
陈昌毅
曾璐
杨林
陈博文
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangxi University for Nationalities
Original Assignee
Guangxi University for Nationalities
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangxi University for Nationalities filed Critical Guangxi University for Nationalities
Priority to CN202010505476.6A priority Critical patent/CN111783940A/en
Publication of CN111783940A publication Critical patent/CN111783940A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • 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
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B17/00—Fire alarms; Alarms responsive to explosion

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Business, Economics & Management (AREA)
  • Emergency Management (AREA)
  • Fire Alarms (AREA)

Abstract

The invention discloses a method for reducing fire false alarm rate based on GA-BP neural network algorithm, relating to the technical field of fire alarm, which comprises the steps of obtaining fire-related sensor data, using the sensor data as the input of BP neural network, determining the structure of the BP neural network, adopting genetic algorithm to determine the weight w and bias b of the BP neural network, training the BP neural network according to the weight w and the bias b to obtain GA-BP neural network model, and judging whether to early warn fire according to real-time sensor data by the GA-BP neural network model. The invention reduces the error probability through the common judgment of multiple sensors; the BP neural network is used as a classifier, fire sample data training is carried out, and a model is trained by known sample data, so that the robustness of the invention is improved; and the weight w and the bias b of the BP neural network are initialized by adopting a GA algorithm, so that the BP neural network is prevented from falling into a local optimal solution.

Description

Method for reducing fire false alarm rate based on GA-BP neural network algorithm
Technical Field
The invention belongs to the technical field of fire alarm, and particularly relates to a method for reducing fire false alarm rate based on a GA-BP neural network algorithm.
Background
With the increase of the existing alternative fire and electricity consumption, the fire disaster is more and more frequently generated. Once a fire disaster occurs, adverse factors such as untimely fire fighting, lack of fire extinguishing equipment, scare and misbehavior of bystanders, slow escape and the like easily occur, and finally, serious life and property loss is caused. The characteristics and fire prevention strategies of the family fire are discussed, and the method has practical significance for preventing the family fire and reducing the fire loss.
At present, various fire alarm systems have the problem of high false alarm rate, the false alarm of fire easily causes people to be nervous and panic, and frequent false alarm can cause users to lose trust in the fire alarm system or even close the fire alarm system, which is very dangerous. Therefore, the problem that the false alarm rate of the fire alarm system is reduced and the alarm accuracy of the fire alarm system is improved is the problem which needs to be solved urgently at present.
There is currently no technology that can reduce the false alarm rate to a very low level. At present, no effective method exists for the high false alarm rate of the fire alarm system. If false alarm occurs, the alarm can be turned off and turned on again or the alarm is called, people generally manage the alarm system, and the people need to check if false alarm is found. The problem that the false alarm rate of the current fire alarm system is high causes manpower waste and panic, and is not favorable for market popularization and large-scale use. Moreover, the problem of high false alarm rate of the fire alarm exists all the time. The false alarm rate of a single sensor is extremely high and can hardly be used. The multi-sensor simple circuit connection is easily interfered by various kinds of interference, and the accuracy also needs to be improved. And the contribution degree of each sensor in fire identification is determined by using multi-sensor fusion and a genetic algorithm and a neural network, so that the effect of greatly reducing the false alarm rate is achieved, the accuracy is greatly improved, and the robustness is also enhanced.
Genetic Algorithm (GA) is a computational model of the biological evolution process that simulates the natural selection and Genetic mechanism of darwinian biological evolution theory, and it searches for the optimal solution by simulating the natural evolution process, and the main contents include, encoding, selecting, crossing, mutating, and decoding.
The BP (Back Propagation) neural network is a feedforward network trained according to the error Back Propagation, and is mainly applied to the aspects of data fitting, function approximation and the like. The BP neural network model establishment is to construct an input layer, a hidden layer and an output layer of the neural network, and then determine the weight w and the bias b according to the data samples to obtain a complete neural network model. However, the BP neural network has the defects of slow learning rate, easy falling into local minimum value and the like.
FPGA (field Programmable Gate array) is a Programmable logic device with low power consumption, embedded with abundant hardmac multiplier and memory resources, has the characteristics of parallel computing capability and repeatable configuration, and becomes an ideal device for researching neural network hardware implementation.
Disclosure of Invention
The invention aims to provide a method for reducing the fire false alarm rate based on a GA-BP neural network algorithm, thereby solving the defect that the existing fire alarm system has high false alarm rate.
In order to achieve the aim, the invention provides a method for reducing fire false alarm rate based on a GA-BP neural network algorithm, which comprises the following steps:
acquiring fire related sensor data;
taking the sensor data as the input of a BP neural network, and determining the structure of the BP neural network;
determining the weight w and the bias b of the BP neural network by adopting a genetic algorithm, and training the BP neural network according to the weight w and the bias b to obtain a GA-BP neural network model;
and the GA-BP neural network model judges whether to early warn fire or not according to the real-time sensor data.
Further, the method also comprises the step of transplanting the GA-BP neural network model to a terminal, and carrying out fire early warning through the terminal.
Further, determining the weight w and the bias b of the BP neural network by adopting a genetic algorithm, and training the BP neural network according to the weight w and the bias b to obtain a GA-BP neural network model, wherein the GA-BP neural network model comprises the following steps:
training the BP neural network according to the weight w and the bias b to obtain an error, and taking the error as a fitness value;
iteratively selecting, crossing and mutating the fitness value through a genetic algorithm until the fitness value meets a set end condition to obtain a first optimized weight w1And bias b1；
Weighting w of the first optimization1And bias b1Putting the weight into a BP neural network, performing iterative calculation on errors, updating the weight and the offset until an ending condition is met, and obtaining a second optimized weight w2And bias b2According to the second optimized weight w2And bias b2And establishing a GA-BP neural network model.
Furthermore, the S-shaped activation function in the GA-BP neural network model adopts approximate processing to convert the S-shaped activation function into a piecewise function, and each piecewise function is a linear function.
Further, the GA-BP neural network model is applied to a terminal in a mode of combining a parallel structure and a serial structure.
Further, the terminal adopts an FPGA, when the output of the GA-BP neural network model on the FPGA is 1, the alarm is given, and when the output of the GA-BP neural network model on the FPGA is 0, the alarm is in a normal state, and the alarm is not given.
Further, the GA-BP neural network model performs modular operation on the FPGA, and the BP neural network model is divided into modules, including: the device comprises a data acquisition module, a hidden layer calculation module and an output layer calculation module;
the data acquisition module is used for storing the acquired real-time sensor data into a register in real time;
the hidden layer calculation module is used for calculating the real-time sensor data xiAnd w between the input layer and the hidden layer(1) i,jAnd bjZ is obtained by calculation of the formula (1)jThen, z is addedjObtaining the output y of the hidden layer through an S-type functionjAnd update xi(ii) a The formula (1) is:
zj＝xi*w(1) i,j+bj(1)
in the formula (1), xiI represents the ith information of the input layer, i is 1,2,3, 4; w is a(1) i,jRepresenting the weight from the jth node of the hidden layer to the ith node of the input layer, bjThreshold, z, representing the jth node of the hidden layerjIs an intermediate variable of the jth node of the hidden layer, yjIs the output of the jth hidden layer, j is 1, 2.. and j is the number of neurons in the hidden layer;
the output layer calculation module is used for calculating the yjAnd w between the hidden layer and the output layer(2) j，kAnd bkZ is obtained by calculation of the formula (2)kThen, z is addedkObtaining the output out of the BP neural network through the S function, wherein the out is 0 and represents no alarm; out is 1, representing an alarm, outputs a voltage signal, and updates xiAnd yj(ii) a The formula (2) is:
zk＝yj*w(2) j，k+bk(2)
in the formula (2), yjIs the output of the jth hidden layer, j is 1, 2.. and j is the number of neurons in the hidden layer; w is a(2) j,kRepresenting the weight from the jth node of the hidden layer to the kth node of the output layer, bkThreshold, z, representing the kth node of the output layerkIs an intermediate variable at the kth node of the output layer and out is the output of the output layer, i.e. the output of the whole GA-BP neural network.
Further, the sensor data includes: temperature sensor, smoke sensor, CO sensor, and flame sensor.
Further, the BP neural network employs a single hidden layer neural network.
Compared with the prior art, the invention has the following beneficial effects:
the method for reducing the fire false alarm rate based on the GA-BP neural network algorithm obtains the fire related sensor data; taking sensor data as the input of a BP neural network, and determining the structure of the BP neural network; determining the weight w and the bias b of the BP neural network by adopting a genetic algorithm, and training the BP neural network according to the weight w and the bias b to obtain a GA-BP neural network model; and the GA-BP neural network model judges whether to early warn fire or not according to the real-time sensor data. The error probability is reduced through the joint judgment of multiple sensors; the BP neural network is used as a classifier, fire sample data is used for training, and the model trained by the known sample data improves the robustness of the invention; initializing the weight w and the bias b of the BP neural network by adopting a GA algorithm, and avoiding the BP neural network from falling into a local optimal solution; the parallel and pipeline characteristics of the FPGA are fully utilized, the decision time is reduced, and the real-time performance is very high.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only one embodiment of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
FIG. 1 is a flow chart of a method for reducing fire false alarm rate based on GA-BP neural network algorithm of the present invention;
FIG. 2 is a schematic diagram of the sigmoid activation function curve of the present invention;
FIG. 3 is a schematic representation of the sigmoid activation function piecewise curve of the present invention;
FIG. 4 is a schematic diagram of a module pipeline structure of a GA-BP neural network in the FPGA of the present invention;
FIG. 5 is a flowchart of the operation of the hidden layer computation module of the present invention;
FIG. 6 is a flow chart of the operation of the output layer calculation module of the present invention;
FIG. 7 is a flow chart of the present invention for real-time data determination and alerting.
Detailed Description
The technical solutions in the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in FIG. 1, the method for reducing the fire false alarm rate based on the GA-BP neural network algorithm provided by the invention comprises the following steps:
s1, acquiring fire related sensor data such as a temperature sensor, a smoke sensor, a CO sensor and a flame sensor, converting the sensor data into digital quantity through an AD converter, taking the converted quantity data as the input of a BP neural network, namely a characteristic vector, and determining the structure of the BP neural network, wherein the step of determining the structure of the BP neural network comprises the following steps:
in the embodiment, data input by 4 sensors, namely a temperature sensor, a smoke sensor, a CO sensor and a flame sensor, are used as input data of an input layer, wherein the input layer comprises 4 neurons; only 1 output out, namely whether to alarm or not is judged according to the out, so that the output layer is 1 neuron;
determining the number of neurons of the hidden layer; the number of the neurons of the hidden layer can be changed, the BP neural network adopts a single hidden layer neural network, the number of the neurons of the hidden layer is determined to be 3-15 neurons according to the characteristics of the neural network, and the convergence speed and the error of a neural network model formed by the numbers of the neurons of different hidden layers in the range are compared to determine the optimal number of the neurons of the hidden layer.
And S2, determining the weight w and the bias b of the BP neural network by adopting a genetic algorithm, and training the BP neural network according to the weight w and the bias b to obtain a GA-BP neural network model.
Under general conditions, the structural form of the neural network is completely parallel in hardware, and some sequential processing is inevitably existed, so that the parallel structure most suitable for the hardware is required to be selected to complete the optimal mapping of the neural network structure in the FPGA, parallel operation is adopted among neurons in the same layer, and the pipeline technology (serial) of the FPGA is adopted among layers (between a data acquisition module and a hidden layer calculation module, and between a hidden layer calculation module and an output layer calculation module), namely the invention adopts a mode of combining the parallel structure and the serial structure to realize the GA-BP neural network model in the FPGA. Therefore, in the application, the fire situation can be judged by converting the output of the real-time acquisition sensor into a digital quantity as a characteristic vector and calculating the weight w and the offset b of the cured BP neural network to obtain the output (which is represented as 0 or 1).
S3, the GA-BP neural network model is solidified on the FPGA (terminal) in a code writing mode, when the output of the GA-BP neural network model on the FPGA is 1, an alarm is given, and when the output of the GA-BP neural network model on the FPGA is 0, the GA-BP neural network model belongs to a normal state, and no alarm is given. The BP neural network is realized by adopting the FPGA because the good parallelism of the BP neural network conforms to the operation mode of the neural network, so that the decision delay of the system can be greatly reduced, and the system performance is obviously improved.
And S4, judging whether to early warn fire or not by the GA-BP neural network model according to the real-time sensor data acquired by the FPGA.
With continued reference to fig. 1, the global optimal solution cannot be accurately derived using only the GA algorithm, while the local solution is easily derived using only the BP neural network. And GA iteration is firstly carried out to solve, the solutions of w and b are fixed in a certain range, and then the BP neural network is used for accurately solving the solutions of w and b, so that the optimal solution can be obtained by combining the genetic algorithm and the neural network, and therefore the S2 specifically comprises the following steps:
s21, training the BP neural network according to the weight w and the bias b to obtain an error, and taking the error as a fitness value of the genetic algorithm (namely, solving the minimum value of the BP neural network error by using the genetic algorithm);
s22, coding the weight w and the bias b according to the fitness value, iteratively selecting, intersecting and mutating the fitness value through a genetic algorithm until the fitness value meets a set end condition, wherein the end condition is the minimum value of the fitness (namely the minimum value of the error), and then decoding to obtain the first optimized weight w1And bias b1；
S23, optimizing the first weight w1And bias b1Putting the weight W into a BP neural network, performing iterative error calculation, training and updating the weight W and the bias b until a finishing condition is met (the iterative training is carried out for a certain number of times or the error rate is reduced to a certain value), and obtaining a second optimized weight w2And bias b2At this time, the second optimized weight w2And bias b2For global optimal solution, according to the weight w of the second optimization2And bias b2And establishing a GA-BP neural network model.
And S24, performing approximate processing on the S-type activation function in the GA-BP neural network model, and converting the S-type activation function into a piecewise function of multiplication and addition operation, wherein each piecewise function is a linear function. The S-type activation function contains an index, so that the S-type activation function is difficult to directly realize by using an FPGA and needs to be subjected to linearization processing, therefore, an S-type activation function curve is shown in figure 2, an S-type activation function piecewise curve is shown in figure 3, the S-type activation function piecewise curve consists of 7 linear interval functions and can replace the original S-type activation function, the function is divided into 7 sections as shown in figure 3, and the mathematical expression of the piecewise function is shown in the following table 1.
Table 1 piecewise function mathematical expression:
Figure BDA0002526387550000071
since the FPGA can perform only fixed-point data processing, it is necessary to convert floating-point numbers such as input data, weight, offset, and the like into fixed-point numbers.
And (3) solidifying the whole GA-BP neural network model onto an FPGA (field programmable gate array) (terminal) in a code writing mode, knowing the weight w and the bias b, and performing module-based operation. The invention adopts a mode of combining a parallel structure and a serial structure to realize the BP neural network in the FPGA, parallel operation is adopted among all layers of neurons, the BP neural network is divided into a hidden layer computing module and an output layer computing module because the BP neural network is only a single hidden layer network, and the FPGA pipeline technology is adopted among the modules. The output layer module processes the data processed by the last clock hidden layer module.
As shown in fig. 4, the GA-BP neural network model performs modular operation on the FPGA, and the module design of the GA-BP neural network model according to the pipeline includes: the data acquisition module, the hidden layer calculation module and the output layer calculation module advance in a flowing manner according to the data acquisition module, the hidden layer calculation module and the output layer calculation module. The working principle of the GA-BP neural network model for performing modular operation on the FPGA comprises the following steps:
when the first clock signal of the FPGA starts, the real-time sensor data acquired by the data acquisition module is stored in the register in real time.
The second clock starts, as shown in FIG. 5, the hidden layer computation module will convert the real-time sensor data xiAnd w between the input layer and the hidden layer(1) i,jAnd bjZ is obtained by calculation of the formula (1)jThen, z is further substitutedjObtaining the output y of the hidden layer through an S-type functionjAnd update xi(ii) a The formula (1) is:
zj＝xi*w(1) i,j+bj(1)
in the formula (1), xiIndicates the ith information (smoke, CO, temperature, etc.) of the input layer (i ═ 1,2,3,4), w(1) i,jRepresenting the weight from the jth node of the hidden layer to the ith node of the input layer, bjThreshold, z, representing the jth node of the hidden layerjIs an intermediate variable of the jth node of the hidden layer, yjIs the output of the jth hidden layer ( j 1, 2.., number of hidden layer neurons).
Third clock Start, as shown in FIG. 6, the output layer computation Module uses the hidden layer output yjAnd w between the hidden layer and the output layer(2) j，kAnd bkZ is obtained by calculation of the formula (2)kThen, z is further substitutedkObtaining the output out of the BP neural network through the S function, wherein the output out is 0, represents no alarm, and represents alarm when the output out is 1, outputting a voltage signal, and updating xiAnd yj(ii) a For example, out is 1, and outputs high level, and the alarm device is connected with a buzzer and a light alarm, and simultaneously sends a signal to a master console for warning; the formula (2) is:
zk＝yj*w(2) j，k+bk(2)
in the formula (2), yjIs the output of the jth hidden layer ( j 1, 2.., number of hidden layer neurons), w(2) j,kRepresenting the weight from the jth node of the hidden layer to the kth node of the output layer, bkThreshold, z, representing the kth node of the output layerkIs the intermediate variable (k ═ 1) at the kth node of the output layer, and out is the output of the output layer, i.e., the output of the entire GA-BP neural network.
Updating x when the fourth and the following clock signals arrivei，yjAnd out, real-time monitoring of the surrounding environment, as shown in fig. 7. Therefore, the FPGA is used as a terminal, the GA-BP neural network model is used for reducing the false alarm rate, meanwhile, the surrounding environment can be rapidly monitored in real time, the monitoring can be realized by using three clock signal intervals during starting, and then, each clock signal is monitored, so that the real-time performance is greatly improved compared with a single chip microcomputer and the like.
In summary, the method for reducing the fire false alarm rate based on the GA-BP neural network algorithm reduces the error probability through the joint judgment of multiple sensors; the BP neural network is used as a classifier, fire sample data is used for training, and the model trained by the known sample data improves the robustness of the invention; initializing the weight w and the bias b of the BP neural network by adopting a GA algorithm, and avoiding the BP neural network from falling into a local optimal solution; the parallel and pipeline characteristics of the FPGA are fully utilized, the decision time is reduced, and the real-time performance is very high.
The above disclosure is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or modifications within the technical scope of the present invention, and shall be covered by the scope of the present invention.

Claims (9)

1. A method for reducing fire false alarm rate based on GA-BP neural network algorithm is characterized in that: the method comprises the following steps:
acquiring fire related sensor data;
taking the sensor data as the input of a BP neural network, and determining the structure of the BP neural network;
determining the weight w and the bias b of the BP neural network by adopting a genetic algorithm, and training the BP neural network according to the weight w and the bias b to obtain a GA-BP neural network model;
and the GA-BP neural network model judges whether to early warn fire or not according to the real-time sensor data.
2. The GA-BP neural network algorithm-based method for reducing fire false alarm rate according to claim 1, wherein: and transplanting the GA-BP neural network model to a terminal, and carrying out fire early warning through the terminal.
3. The GA-BP neural network algorithm-based method for reducing fire false alarm rate according to claim 1, wherein: determining the weight w and the bias b of the BP neural network by adopting a genetic algorithm, training the BP neural network according to the weight w and the bias b, and obtaining a GA-BP neural network model, wherein the GA-BP neural network model comprises the following steps:
training the BP neural network according to the weight w and the bias b to obtain an error, and taking the error as a fitness value;
iteratively selecting, crossing and mutating the fitness value through a genetic algorithm until the fitness value meets a set end condition to obtain a first optimized weight w1And bias b1；
Weighting w of the first optimization1And bias b1Putting the weight into a BP neural network, performing iterative calculation on errors, updating the weight and the offset until an ending condition is met, and obtaining a second optimized weight w2And bias b2According to the second optimized weight w2And bias b2And establishing a GA-BP neural network model.
4. A method for reducing fire false alarm rate based on GA-BP neural network algorithm according to claim 3, wherein: and the S-shaped activation function in the GA-BP neural network model adopts approximate processing to convert the S-shaped activation function into a piecewise function, and each piecewise function is a linear function.
5. A method for reducing fire false alarm rate based on GA-BP neural network algorithm according to claim 2, wherein: and applying the GA-BP neural network model in a terminal by adopting a mode of combining a parallel structure and a serial structure.
6. A method for reducing fire false alarm rate based on GA-BP neural network algorithm according to claim 5, wherein: and the terminal adopts an FPGA, and alarms when the output of the GA-BP neural network model on the FPGA is 1, and does not alarm when the output of the GA-BP neural network model on the FPGA is 0.
7. A method for reducing fire false alarm rate based on GA-BP neural network algorithm according to claim 5, wherein: the GA-BP neural network model is subjected to modular operation on the FPGA, and the GA-BP neural network model is subjected to module division, and the method comprises the following steps: the device comprises a data acquisition module, a hidden layer calculation module and an output layer calculation module;
the data acquisition module is used for storing the acquired real-time sensor data into a register in real time;
the hidden layer calculation module is used for calculating the real-time sensor data xiAnd w between the input layer and the hidden layer(1) i,jAnd bjZ is obtained by calculation of the formula (1)jThen, z is addedjObtaining the output y of the hidden layer through an S-type functionjAnd update xi(ii) a The formula (1) is:
zj＝xi*w(1) i,j+bj(1)
in the formula (1), xiI represents the ith information of the input layer, i is 1,2,3, 4; w is a(1) i,jRepresenting the weight from the jth node of the hidden layer to the ith node of the input layer, bjThreshold, z, representing the jth node of the hidden layerjIs an intermediate variable of the jth node of the hidden layer, yjIs the output of the jth hidden layer, j is 1, 2.. and j is the number of neurons in the hidden layer;
the output layer calculation module is used for calculating the yjAnd w between the hidden layer and the output layer(2) j，kAnd bkZ is obtained by calculation of the formula (2)kThen, z is addedkObtaining the output out of the BP neural network through the S function, wherein the out is 0 and represents no alarm; out is 1, representing an alarm, outputs a voltage signal, and updates xiAnd yj(ii) a The formula (2) is:
zk＝yj*w(2) j，k+bk(2)
in the formula (2), yjIs the output of the jth hidden layer, j is 1, 2.. and j is the number of neurons in the hidden layer; w is a(2) j,kRepresenting the weight from the jth node of the hidden layer to the kth node of the output layer, bkThreshold, z, representing the kth node of the output layerkIs an intermediate change of the kth node of the output layerThe quantity, out, is the output of the output layer, i.e. the output of the whole GA-BP neural network.
8. The GA-BP neural network algorithm-based method for reducing fire false alarm rate according to claim 1, wherein: the sensor data includes: temperature sensor, smoke sensor, CO sensor, and flame sensor.
9. The GA-BP neural network algorithm-based method for reducing fire false alarm rate according to claim 1, wherein: the BP neural network adopts a single hidden layer neural network.
CN202010505476.6A 2020-06-05 2020-06-05 Method for reducing fire false alarm rate based on GA-BP neural network algorithm Pending CN111783940A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010505476.6A CN111783940A (en) 2020-06-05 2020-06-05 Method for reducing fire false alarm rate based on GA-BP neural network algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010505476.6A CN111783940A (en) 2020-06-05 2020-06-05 Method for reducing fire false alarm rate based on GA-BP neural network algorithm

Publications (1)

Publication Number Publication Date
CN111783940A true CN111783940A (en) 2020-10-16

Family

ID=72754677

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010505476.6A Pending CN111783940A (en) 2020-06-05 2020-06-05 Method for reducing fire false alarm rate based on GA-BP neural network algorithm

Country Status (1)

Country Link
CN (1) CN111783940A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113298244A (en) * 2021-04-21 2021-08-24 上海安路信息科技股份有限公司 Neural network post-processing implementation method, device, terminal and medium in target detection
CN113642109A (en) * 2021-08-17 2021-11-12 湖南大学 Multi-sensor scheme evaluation and optimization system and method based on BP neural network

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105513293A (en) * 2016-03-03 2016-04-20 广西民族大学 Passive temperature difference alarm system
CN106934451A (en) * 2015-12-29 2017-07-07 扬州大学 A kind of BP neural network WSN forest fire crime prevention systems based on ant group optimization
CN109978005A (en) * 2019-02-25 2019-07-05 深圳市中电数通智慧安全科技股份有限公司 A kind of fire alarm method, device, storage medium and terminal device
CN110738006A (en) * 2019-09-19 2020-01-31 天津大学 High-precision resistance measurement algorithm based on GA-BP neural network algorithm
CN111127811A (en) * 2019-12-25 2020-05-08 上海船舶电子设备研究所(中国船舶重工集团公司第七二六研究所) Fire early detection alarm system and method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106934451A (en) * 2015-12-29 2017-07-07 扬州大学 A kind of BP neural network WSN forest fire crime prevention systems based on ant group optimization
CN105513293A (en) * 2016-03-03 2016-04-20 广西民族大学 Passive temperature difference alarm system
CN109978005A (en) * 2019-02-25 2019-07-05 深圳市中电数通智慧安全科技股份有限公司 A kind of fire alarm method, device, storage medium and terminal device
CN110738006A (en) * 2019-09-19 2020-01-31 天津大学 High-precision resistance measurement algorithm based on GA-BP neural network algorithm
CN111127811A (en) * 2019-12-25 2020-05-08 上海船舶电子设备研究所(中国船舶重工集团公司第七二六研究所) Fire early detection alarm system and method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CHAOYI等: "Prediction model for the water jet falling point in fire extinguishing based on a GA-BP neural network", 《PLOS ONE》 *
张文杰: "飞机货舱复合烟雾探测方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
杨景明等: "基于FPGA的BP神经网络硬件实现及改进", 《计算机工程与设计》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113298244A (en) * 2021-04-21 2021-08-24 上海安路信息科技股份有限公司 Neural network post-processing implementation method, device, terminal and medium in target detection
CN113298244B (en) * 2021-04-21 2023-11-24 上海安路信息科技股份有限公司 Neural network post-processing implementation method, device, terminal and medium in target detection
CN113642109A (en) * 2021-08-17 2021-11-12 湖南大学 Multi-sensor scheme evaluation and optimization system and method based on BP neural network
CN113642109B (en) * 2021-08-17 2023-07-04 湖南大学 Multi-sensor scheme evaluation and optimization system and method based on BP neural network

Similar Documents

Publication Publication Date Title
CN113126676B (en) Livestock and poultry house breeding environment parameter intelligent control system
CN111832825B (en) Wind power prediction method and system integrating long-term memory network and extreme learning machine
CN111767517B (en) BiGRU multi-step prediction method, system and storage medium applied to flood prediction
CN113031555B (en) Intelligent purification system for harmful gas in environment of livestock and poultry house
CN111783940A (en) Method for reducing fire false alarm rate based on GA-BP neural network algorithm
CN112881601B (en) Moisture detecting system based on cloud platform
Liang et al. A combined model based on CEEMDAN, permutation entropy, gated recurrent unit network, and an improved bat algorithm for wind speed forecasting
CN111814956A (en) Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction
CN112330951A (en) Method for realizing road network traffic data restoration based on generation of countermeasure network
CN113219871B (en) Curing room environmental parameter detecting system
CN115630101A (en) Hydrological parameter intelligent monitoring and water resource big data management system
CN111474094A (en) Intelligent dust concentration detection system
CN114662791A (en) Long time sequence pm2.5 prediction method and system based on space-time attention
CN114911185A (en) Security big data Internet of things intelligent system based on cloud platform and mobile terminal App
CN111292121A (en) Garden load prediction method and system based on garden image
CN112911533B (en) Temperature detection system based on remove end App
Alfredo et al. Time series forecasting of significant wave height using gru, cnn-gru, and lstm
CN113283642A (en) Poultry feed detection and formula system
CN110545208B (en) Network traffic prediction method based on LSTM
CN114386672B (en) Environment big data Internet of things intelligent detection system
CN115659201A (en) Gas concentration detection method and monitoring system for Internet of things
CN116822722A (en) Water level prediction method, system, device, electronic equipment and medium
CN115217152B (en) Method and device for predicting open-close deformation of immersed tunnel pipe joint
CN114970745B (en) Intelligent security and environment big data system of Internet of things
CN114267422B (en) Method and system for predicting surface water quality parameters, computer equipment and storage medium

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20201016