A fire identification and early warning method and system based on neighborhood

Abstract

The present invention relates to a neighborhood-based fire identification and early warning method and system, which has a host computer, several sets of combined sensor groups composed of receivers, temperature and humidity, CO, and smoke sensors. By obtaining the plane area, the combined sensor group is placed according to the effective range of the sensor; the neighborhood group is delineated; the sensor data is obtained by the combined sensor group, input into the GA_BP neural network optimization model in the host computer, and the judgment result is output; the RSSI value of the current combined sensor group is calculated; the combined sensor group for auxiliary judgment is determined based on this, and the combined sensor group data for auxiliary judgment is read; the auxiliary combined sensor group data is fused, the sensor data value is determined by the RSSI value ratio, and the correction is performed to determine the final sensor data result; the corrected sensor data judgment result. Through reasonable layout, the overall layout is carried out using an appropriate number of sensors to ensure accurate identification of large-scale storage fires at low cost.

Description

Technical Field

[0001] The present invention belongs to the technical field of automatic identification, and in particular relates to a neighborhood-based fire identification and early warning method and system. Background Art

[0002] Fire warning has been widely used in large buildings and other scenarios, especially large warehouses. However, the following problems exist: 1. Large warehouses store a large number of items, and the placement of items leads to data transmission obstruction. Due to specific reasons such as light and location, signal transmission is obstructed or errors occur. In tobacco storage warehouses, the use of a small number of sensors such as temperature and humidity, CO, and smoke sensors is prone to missed detections. A large number of sensors leads to high costs and large and slow data processing, which is not suitable for normal inspections in real-time scenarios. 2. With the development of artificial intelligence technology, intelligent algorithms have been widely popularized, but traditional artificial intelligence algorithms are not particularly suitable for current scenarios and require adaptive changes to improve accuracy. In order to solve the above technical problems, there is an urgent need for a fire identification and warning system and method with a reasonable number and layout to quickly and accurately alarm. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a neighborhood-based fire identification and warning method and system, which solves the above-mentioned technical problems through a reasonable number of sensor layouts and neural network identification and warning judgments adapted to the current scene.

[0004] In order to achieve the above-mentioned object of the invention, the technical solution of the present invention is implemented in the following manner:

[0005] A neighborhood-based fire identification and warning method includes a host computer, several sets of combined sensor groups consisting of receivers, temperature and humidity, CO, and smoke sensors, and is specifically performed by the following steps:

[0006] Step S1: Obtain the plane area and place the combined sensor group s according to the effective range of the sensor i ,i=1,2,...,n;

[0007] Furthermore, according to the specific installation conditions, the minimum density is the minimum effective transmission identification range of the temperature and humidity, CO, and smoke sensors, and the specific density is determined according to actual needs;

[0008] Step S2: Delineate neighborhood groups, with tangent or intersecting combined sensor groups as a group q j , j = 1, 2, ..., n;

[0009] Step S3: The combined sensor group acquires sensor data, inputs the data into the GA_BP neural network optimization model in the host computer, and outputs the judgment result;

[0010] Furthermore, the GA_BP neural network optimization model includes:

[0011] Step S31: reading data;

[0012] Step S32: Data preprocessing; n =x n -x min / x max -x min ; where x max To read the maximum number in the neighborhood group data set, x min To read the minimum number in the neighborhood group data set;

[0013] Step S33: Setting the optimal number of hidden layer nodes, In the formula, m is the number of nodes in the input layer, and n is the number of nodes in the output layer;

[0014] Step S34: Initialize the number of evolutions, population size, crossover probability, and mutation probability;

[0015] The selection operation adopts the linear sorting method. The individuals in the population are first sorted according to the value of fitness, and then all individuals are assigned a serial number. The best individual is N, and the probability of being selected is P. max , the worst individual number is 1, and the probability of being selected is P min , choose to calculate the probability as follows:

[0016]

[0017] On the basis of real number coding for individuals, the real number crossover method is used for crossover processing. k and chromosome a l The crossover operation method at position r is as follows:

[0018] a kr =a kr (1-b)+a lr b

[0019] a lr =a lr (1-b)+a kr b, where b is a random number between the interval [0,1];

[0020] Select the rth gene a of the i-th individual ir To mutate, the mutation operation method is as follows:

[0021]

[0022] In the formula, a max Gene a irThe upper bound of min Gene a ir The lower bound of

[0023]

[0024] g is the current iteration number, G max is the maximum number of evolutions; t is a random number between the interval [0,1];

[0025] Step S35: Perform real number encoding on the population and calculate the fitness value:

[0026]

[0027] Where: n is the number of network output nodes; y i is the expected output of the ith node of the BP neural network; o i is the predicted output of the i-th node;

[0028] Step S36: looping steps S31-S35 until the optimal initial weights and thresholds are obtained; constructing the corresponding relationship between the fitness value and the result of the GA_BP neural network;

[0029] Furthermore, the corresponding relationship is adjusted according to the actual data calculation result;

[0030] Step S37: outputting a judgment result according to the corresponding relationship: one of normal, smoldering, and open flame;

[0031] Step S4: Determine whether the current state is normal, smoldering, or open flame; if it is determined to be smoldering or open flame, calculate the RSSI value of the current combined sensor group and set it as RSSI 0 And the neighborhood group combined sensor group RSSI value, set to RSSI 1 , calculate |RSSI 1 -RSSI 0 |<α, wherein α is determined according to the number of combined sensor groups, and it is best to include 3 combined sensor groups; accordingly, the combined sensor group for auxiliary judgment is determined, and the combined sensor group data for auxiliary judgment is read, and step S3 is executed;

[0032] Step S5: fusing the auxiliary combined sensor group data, determining the sensor data value by RSSI value ratio, and performing correction to determine the final sensor data result.

[0033] Furthermore, the RSSI value ratio is calculated as follows: n / (RSSI 0 +RSSI 1 +...+RSSI n );

[0034] Furthermore, the correction calculation method is:

[0035] Sensor data*(RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n )).

[0036] Step S6: After the correction, the sensor data is input to execute step S3. When the judgment result shows either smoldering or open flame again, the current state is determined to be abnormal and a fire alarm is issued. When the judgment result shows no smoldering or open flame, an alarm is issued according to the error elimination.

[0037] The present invention also provides a neighborhood-based fire identification and warning system, which has a host computer, several sets of combined sensor groups consisting of receivers, temperature and humidity, CO, and smoke sensors. The system realizes fire identification and warning functions through the host computer and the combined sensor group:

[0038] Get the plane area and place the combined sensor group according to the effective range of the sensor i ,i=1,2,...,n;

[0039] Furthermore, according to the specific installation conditions, the minimum density is the minimum effective transmission identification range of the temperature and humidity, CO, and smoke sensors, and the specific density is determined according to actual needs;

[0040] Furthermore, the neighborhood groups are defined, and the tangent or intersecting combined sensor groups are a group q j , j = 1, 2, ..., n;

[0041] Furthermore, the combined sensor group acquires sensor data, inputs it into the GA_BP neural network optimization model in the host computer, and outputs the judgment result;

[0042] Furthermore, the GA_BP neural network optimization model includes:

[0043] Read module: read data;

[0044] Preprocessing module: data preprocessing; x n =x n -x min / x max -x min ; where x max To read the maximum number in the neighborhood group data set, x min To read the minimum number in the neighborhood group data set;

[0045] Setting module: Setting the optimal number of hidden layer nodes, In the formula, m is the number of nodes in the input layer, n is the number of nodes in the output layer,

[0046] Evolution module: initialize the number of evolutions, population size, crossover probability, and mutation probability;

[0047] The selection operation adopts the linear sorting method. The individuals in the population are first sorted according to the value of fitness, and then all individuals are assigned a serial number. The best individual is N, and the probability of being selected is P. max , the worst individual number is 1, and the probability of being selected is P min , choose to calculate the probability as follows:

[0048]

[0049] On the basis of real number coding for individuals, the real number crossover method is used for crossover processing. k and chromosome a l The crossover operation method at position r is as follows:

[0050] a kr =a kr (1-b)+a lr b

[0051] a lr =a lr (1-b)+a kr b, where b is a random number between the interval [0,1];

[0052] Select the rth gene a of the i-th individual ir To mutate, the mutation operation method is as follows:

[0053]

[0054] In the formula, a max Gene a ir The upper bound of min Gene a ir The lower bound of

[0055]

[0056] g is the current iteration number, G max is the maximum number of evolutions; t is a random number between the interval [0,1];

[0057] Encoding module: Encode the population with real numbers, and the fitness value is:

[0058]

[0059] Where: n is the number of network output nodes; y iis the expected output of the ith node of the BP neural network; o i is the predicted output of the i-th node;

[0060] Cycle module: Cycle evolution until the optimal initial weight and threshold are obtained; construct the corresponding relationship between the fitness value and result of the GA_BP neural network;

[0061] Furthermore, the corresponding relationship is adjusted according to the actual data calculation result;

[0062] Judgment module: outputs the judgment result according to the corresponding relationship: normal, smoldering, open flame;

[0063] Correction module: Determine whether the current state is normal, smoldering, or open flame; if it is smoldering or open flame, calculate the RSSI value of the current combined sensor group and set it as RSSI 0 And the neighborhood group combined sensor group RSSI value, set to RSSI 1 , calculate |RSSI 1 -RSSI 0 |<α, wherein α is determined according to the number of combined sensor groups, and it is best to include 3 combined sensor groups; accordingly, the combined sensor group for auxiliary judgment is determined, and the combined sensor group data for auxiliary judgment is read, and step S3 is executed;

[0064] Fusion module: fuses the auxiliary combined sensor group data, determines the sensor data value based on the RSSI value ratio, and makes corrections to determine the final sensor data result.

[0065] Furthermore, the RSSI value ratio is calculated as follows: n / (RSSI 0 +RSSI 1 +...+RSSI n );

[0066] Furthermore, the correction calculation method is:

[0067] Sensor data*(RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n )).

[0068] Output module: After correction, the sensor data is input to execute step S3. When the judgment result shows either smoldering or open flame again, the current state is determined to be abnormal and a fire alarm is issued. When the judgment result shows no smoldering or open flame, an alarm is issued according to the error elimination.

[0069] Beneficial effects of the present invention: 1. Through reasonable layout, using an appropriate number of sensors for overall layout, it is ensured that while conducting a comprehensive inspection, the delay and error caused by signal obstruction are avoided, and accurate identification of fires in large warehouses is guaranteed at a low cost. 2. The data processed by a reasonable number of sensor groups is relatively small, and the processing is fast, achieving full-time normal online. 3. Sensor data based on the neighborhood can ensure data accuracy and timeliness, and optimize the traditional GA-BP network to improve accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more specifically illustrate the implementation methods of the present invention and the technical solutions in the prior art, the relevant drawings will be briefly introduced below, and various aspects of the present invention will be described in detail in the following text with reference to the drawings. Please note that the introduction of these drawings is only to more clearly illustrate the implementation methods of the present invention or the prior art solutions. The order of illustrations and the proportional relationship between objects shown in the drawings may not be consistent with the actual situation. The same marks may be used in the drawings to identify the same or similar parts or components. Those skilled in the art should understand that the purpose of the drawings is to help readers better understand the implementation methods of the present invention or the prior art solutions, rather than to limit the present invention. Therefore, the specific implementation methods of the present invention are not limited to the forms shown in these drawings, but should be understood and applied with reference to the overall description of the present invention. In the drawings:

[0071] Figure 1 A schematic diagram of the steps of a neighborhood-based fire identification and early warning method according to an embodiment of the present invention;

[0072] Figure 2 A simplified structural diagram of a neighborhood-based fire identification and early warning method according to an embodiment of the present invention;

[0073] 1. Host computer; 2. Combined sensor group; 3. Receiver; 4. CO sensor; 5. Temperature and humidity sensor; 6. Smoke sensor; 7. Plane range. DETAILED DESCRIPTION

[0074] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0075] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0076] like Figure 1 As shown, a neighborhood-based fire identification and warning method includes a host computer, several sets of combined sensor groups consisting of receivers, temperature and humidity, CO, and smoke sensors, and is specifically performed by the following steps:

[0077] Step S1: Obtain the plane area and place the combined sensor group s according to the effective range of the sensor i ,i=1,2,...,n;

[0078] Specifically, according to the specific installation conditions, the minimum density is the tangency generated by the minimum effective transmission and recognition range of the temperature and humidity, CO, and smoke sensors, and the specific density is determined according to actual needs; if the effective transmission and recognition range of a sensor in the sensor is 10 meters, for a warehouse of 100*100 square meters, 25 sensor groups are required; the receiver is used to receive and send data and is wirelessly connected to the host computer;

[0079] Step S2: Delineate neighborhood groups, with tangent or intersecting combined sensor groups as a group q j , j = 1, 2, ..., n; tangency or intersection is determined by the minimum effective transmission identification range as the radius;

[0080] Step S3: The combined sensor group acquires sensor data, inputs the data into the GA_BP neural network optimization model in the host computer, and outputs the judgment result; the GA algorithm is used as a genetic algorithm to optimize the BP neural network parameters and determine the best value;

[0081] Specifically, the GA_BP neural network optimization model includes:

[0082] Step S31: reading data;

[0083] Step S32: Data preprocessing; n =x n -x min / x max -x min ; where x max To read the maximum number in the neighborhood group data set, x min To read the minimum number in the neighborhood group data set;

[0084] Step S33: Setting the optimal number of hidden layer nodes, In the formula, m is the number of nodes in the input layer, and n is the number of nodes in the output layer. j is used as a constant term, and the number of domain groups is used for more accurate optimization.

[0085] Step S34: Initialize the number of evolutions, population size, crossover probability, and mutation probability;

[0086] The selection operation adopts the linear sorting method. The individuals in the population are first sorted according to the value of fitness, and then all individuals are assigned a serial number. The best individual is N, and the probability of being selected is P. max , the worst individual number is 1, and the probability of being selected is P min , choose to calculate the probability as follows:

[0087]

[0088] On the basis of real number coding for individuals, the real number crossover method is used for crossover processing. k and chromosome a l The crossover operation method at position r is as follows:

[0089] a kr =a kr (1-b)+a lr b

[0090] a lr =a lr (1-b)+a kr b, where b is a random number between the interval [0,1];

[0091] Select the rth gene a of the i-th individual ir To mutate, the mutation operation method is as follows:

[0092]

[0093] In the formula, a max Gene a ir The upper bound of min Gene a ir The lower bound of

[0094]

[0095] g is the current iteration number, G max is the maximum number of evolutions; t is a random number between the interval [0,1];

[0096] Step S35: Real number encoding is performed on the population, and the fitness value is:

[0097]

[0098] Where: n is the number of network output nodes; y i is the expected output of the ith node of the BP neural network; o i is the predicted output of the i-th node; using i as a constant means using i data for optimization to further accurately optimize the fitness value;

[0099] Step S36: looping steps S31-S35 until the optimal initial weights and thresholds are obtained; constructing the corresponding relationship between the fitness value and the result of the GA_BP neural network;

[0100] Specifically, the corresponding relationship is adjusted according to the actual data calculation results; standard fire data can be used first in the preliminary construction of the neural network model, and a real-time data set can be constructed according to the real-time monitoring data for optimization;

[0101] Step S37: outputting a judgment result according to the corresponding relationship: one of normal, smoldering, and open flame; in particular, outputting one of the judgment results according to the interval corresponding to the fitness value;

[0102] Step S4: Determine whether the current state is normal, smoldering, or open flame; if the host computer determines that it is in the smoldering or open flame state, it communicates with the sensor group and calculates the RSSI value of the current combined sensor group and sets it as RSSI 0 And the neighborhood group combined sensor group RSSI value, set to RSSI 1 , calculate |RSSI 1 -RSSI 0 |<α, wherein α is determined according to the number of combined sensor groups, and it is best to include 3 combined sensor groups; accordingly, the combined sensor group for auxiliary judgment is determined, and the combined sensor group data for auxiliary judgment is read, and step S3 is executed;

[0103] Step S5: fusing the auxiliary combined sensor group data, determining the sensor data value by RSSI value ratio, and performing correction to determine the final sensor data result.

[0104] Specifically, the RSSI value ratio is calculated as follows: RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n );

[0105] Specifically, the correction calculation method is:

[0106] Sensor data*(RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n )).

[0107] Step S6: After the correction, the sensor data is input to execute step S3. When the judgment result shows either smoldering or open flame again, the current state is determined to be abnormal and a fire alarm is issued. When the judgment result shows no smoldering or open flame, an alarm is issued according to the error elimination.

[0108] like Figure 2 As shown, the present invention also provides a neighborhood-based fire identification and warning system, which has a host computer, several sets of combined sensor groups consisting of receivers, temperature and humidity, CO, and smoke sensors. The system realizes fire identification and warning functions through the host computer and the combined sensor group:

[0109] Get the plane area and place the combined sensor group according to the effective range of the sensor i ,i=1,2,...,n;

[0110] Specifically, according to the specific installation conditions, the minimum density is the tangency produced by the minimum effective transmission identification range of the temperature and humidity, CO, and smoke sensors, and the specific density is determined according to actual needs;

[0111] Specifically, a neighborhood group is defined, with the tangent or intersecting combined sensor group as a group q j , j = 1, 2, ..., n;

[0112] Specifically, the combined sensor group acquires sensor data, inputs it into the GA_BP neural network optimization model in the host computer, and outputs the judgment result;

[0113] Specifically, the GA_BP neural network optimization model includes:

[0114] Read module: read data;

[0115] Preprocessing module: data preprocessing; x n =x n -x min / x max -x min ; where x max To read the maximum number in the neighborhood group data set, x min To read the minimum number in the neighborhood group data set;

[0116] Setting module: Setting the optimal number of hidden layer nodes, In the formula, m is the number of nodes in the input layer, n is the number of nodes in the output layer,

[0117] Evolution module: initialize the number of evolutions, population size, crossover probability, and mutation probability;

[0118] The selection operation adopts the linear sorting method. The individuals in the population are first sorted according to the value of fitness, and then all individuals are assigned a serial number. The best individual is N, and the probability of being selected is P. max , the worst individual number is 1, and the probability of being selected is P min , choose to calculate the probability as follows:

[0119]

[0120] On the basis of real number coding for individuals, the real number crossover method is used for crossover processing. k and chromosome a l The crossover operation method at position r is as follows:

[0121] a kr =a kr (1-b)+a lr b

[0122] a lr =a lr (1-b)+a kr b, where b is a random number between the interval [0,1];

[0123] Select the rth gene a of the i-th individual ir To mutate, the mutation operation method is as follows:

[0124]

[0125] In the formula, a max Gene a ir The upper bound of min Gene a ir The lower bound of

[0126]

[0127] g is the current iteration number, G max is the maximum number of evolutions; t is a random number between the interval [0,1];

[0128] Encoding module: Encode the population with real numbers, and the fitness value is:

[0129]

[0130] Where: n is the number of network output nodes; y i is the expected output of the ith node of the BP neural network; o i is the predicted output of the i-th node;

[0131] Cycle module: Cycle evolution until the optimal initial weight and threshold are obtained; construct the corresponding relationship between the fitness value and result of the GA_BP neural network;

[0132] Specifically, the corresponding relationship is adjusted according to the actual data calculation result;

[0133] Judgment module: outputs the judgment result according to the corresponding relationship: normal, smoldering, open flame;

[0134] Correction module: Determine whether the current state is normal, smoldering, or open flame; if it is smoldering or open flame, calculate the RSSI value of the current combined sensor group and set it as RSSI 0 And the neighborhood group combined sensor group RSSI value, set to RSSI 1 , calculate |RSSI 1 -RSSI 0 |<α, wherein α is determined according to the number of combined sensor groups, and it is best to include 3 combined sensor groups; accordingly, the combined sensor group for auxiliary judgment is determined, and the combined sensor group data for auxiliary judgment is read, and step S3 is executed;

[0135] Fusion module: fuses the auxiliary combined sensor group data, determines the sensor data value based on the RSSI value ratio, and makes corrections to determine the final sensor data result.

[0136] Specifically, the RSSI value ratio is calculated as follows: RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n );

[0137] Specifically, the correction calculation method is:

[0138] Sensor data*(RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n )).

[0139] Output module: After correction, the sensor data is input to execute step S3. When the judgment result shows either smoldering or open flame again, the current state is determined to be abnormal and a fire alarm is issued. When the judgment result shows no smoldering or open flame, an alarm is issued according to the error elimination.

[0140] Through the present invention, the following are achieved: 1. Through reasonable layout, an appropriate number of sensors are used for overall layout to ensure that while conducting a comprehensive inspection, the signal is not blocked and the resulting delay and error are avoided, and accurate identification of fires in large warehouses is guaranteed at a low cost. 2. The data processed by a reasonable number of sensor groups is relatively small, and the processing is fast, achieving full-time normal online. 3. Sensor data based on the neighborhood can ensure data accuracy and timeliness, and optimize the traditional GA-BP network to improve accuracy.

[0141] The above description is only a partial specific implementation of the present invention and is not limited thereto. The protection scope of the present invention should cover any changes or substitutions that can be easily thought of by anyone familiar with the art. As long as these changes or substitutions fall within the technical scope disclosed by the present invention, they should be regarded as within the protection scope of the present invention. Therefore, no matter how the future development is, any technical improvement, change or substitution within the technical scope of the present invention should be regarded as part of the protection scope of the present invention.

Claims

1. A neighborhood-based fire identification and early warning method, comprising a host computer, several sets of combined sensor groups consisting of receivers, temperature and humidity sensors, CO, and smoke sensors, It is characterized in that The specific steps are as follows: Step S1: Obtain the plane area and place the combined sensor group s according to the effective range of the sensor i ,i=1,2,...,n; Step S2: Delineate neighborhood groups, with tangent or intersecting combined sensor groups as a group q j , j = 1, 2, ..., n; Step S3: The combined sensor group acquires sensor data, inputs the data into the GA_BP neural network optimization model in the host computer, and outputs the judgment result; The GA_BP neural network optimization model performs the following steps: Step S31: reading data; Step S32: Data preprocessing; n =x n -x min / x max -x min ; where x max To read the maximum number in the neighborhood group data set, x min To read the minimum number in the neighborhood group data set; Step S33: Set the optimal number of hidden layer nodes, l< +j; where m is the number of nodes in the input layer, and n is the number of nodes in the output layer; Step S34: Initialize the number of evolutions, population size, crossover probability, and mutation probability; The selection operation adopts the linear sorting method. The individuals in the population are first sorted according to the value of fitness, and then all individuals are assigned a serial number. The best individual is N, and the probability of being selected is P. max , the worst individual number is 1, and the probability of being selected is P min , choose to calculate the probability as follows: On the basis of real number coding for individuals, the real number crossover method is used for crossover processing. k and chromosome a l The crossover operation method at position r is as follows: , where: b is a random number between the interval [0,1]; Select the rth gene of the ith individual To mutate, the mutation operation method is as follows: In the formula, For gene The upper bound of For gene The lower bound of ； g is the current iteration number, is the maximum number of evolutions; t is a random number between the interval [0,1]; Step S35: Perform real number encoding on the population and calculate the fitness value: Where: n is the number of network output nodes; is the expected output of the i-th node of the BP neural network; is the predicted output of the i-th node; Step S36: looping steps S31-S35 until the optimal initial weights and thresholds are obtained; constructing the corresponding relationship between the fitness value and the result of the GA_BP neural network; Step S37: outputting a judgment result according to the corresponding relationship: one of normal, smoldering, and open flame; Step S4: Determine whether the current state is normal, smoldering, or open flame; if it is determined to be smoldering or open flame, calculate the RSSI value of the current combined sensor group and set it as RSSI 0 And the neighborhood group combined sensor group RSSI value, set to RSSI 1 , calculate |RSSI 1 -RSSI 0 |<α, wherein α is determined according to the number of combined sensor groups; based on this, the combined sensor group for auxiliary judgment is determined, and the combined sensor group data for auxiliary judgment is read, and step S3 is executed; Step S5: fusing the data of the combined sensor group, determining the sensor data value by the RSSI value ratio, and making corrections to determine the final sensor data result; Step S6: After the correction, the sensor data is input to execute step S3. When the judgment result shows either smoldering or open flame again, the current state is determined to be abnormal and a fire alarm is issued. When the judgment result shows no smoldering or open flame, an alarm is issued according to the error elimination.

2. A neighborhood-based fire identification and early warning method according to claim 1, It is characterized in that According to the specific installation conditions, the lowest density is the tangency produced by the minimum effective transmission recognition range of the temperature and humidity, CO, and smoke sensors.

3. A neighborhood-based fire identification and early warning method according to claim 2, It is characterized in that The calculation method of the RSSI value ratio in step S5 is: RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n ); The calculation method of the correction in step S5 is: sensor data (RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n )).

4. A neighborhood-based fire identification and early warning method according to claim 2, It is characterized in that A neighborhood group is formed by incorporating three combined sensor groups.

5. A neighborhood-based fire identification and warning system, It is characterized in that The system has a host computer, several sets of combined sensor groups consisting of receivers, temperature and humidity, CO, and smoke sensors. The system realizes fire identification and early warning functions through the host computer and the combined sensor group: Layout device: Get the plane area and place the combined sensor group according to the effective range of the sensor i ,i=1,2,...,n; Neighborhood delineation device: Delineate neighborhood groups, with tangent or intersecting combined sensor groups as a group q j , j = 1, 2, ..., n; Neural network judgment module: The combined sensor group obtains sensor data, inputs the GA_BP neural network optimization model in the host computer, and outputs the judgment result; The neural network determination module comprises: Read module: read data; Preprocessing module: data preprocessing; x n =x n -x min / x max -x min ; where x max To read the maximum number in the neighborhood group data set, x min To read the minimum number in the neighborhood group data set; Setting module: Setting the optimal number of hidden layer nodes, l< +j; where m is the number of nodes in the input layer, n is the number of nodes in the output layer, Evolution module: initialize the number of evolutions, population size, crossover probability, and mutation probability; The selection operation adopts the linear sorting method. The individuals in the population are first sorted according to the value of fitness, and then all individuals are assigned a serial number. The best individual is N, and the probability of being selected is P. max , the worst individual number is 1, and the probability of being selected is P min , choose to calculate the probability as follows: On the basis of real number coding for individuals, the real number crossover method is used for crossover processing. k and chromosome a l The crossover operation method at position r is as follows: , where: b is a random number between the interval [0,1]; Select the rth gene of the ith individual To mutate, the mutation operation method is as follows: In the formula, For gene The upper bound of For gene The lower bound of ； g is the current iteration number, is the maximum number of evolutions; t is a random number between the interval [0,1]; Encoding module: Encode the population with real numbers, and the fitness value is: Where: n is the number of network output nodes; is the expected output of the i-th node of the BP neural network; is the predicted output of the i-th node; Cycle module: Cycle evolution until the optimal initial weight and threshold are obtained; construct the corresponding relationship between the fitness value and result of the GA_BP neural network; Judgment module: outputs the judgment result according to the corresponding relationship: normal, smoldering, open flame; Correction module: Determine whether the current state is normal, smoldering, or open flame; if it is smoldering or open flame, calculate the RSSI value of the current combined sensor group and set it as RSSI 0 And the neighborhood group combined sensor group RSSI value, set to RSSI 1 , calculate |RSSI 1 -RSSI 0 |<α, wherein α is determined according to the number of combined sensor groups; based on this, the combined sensor group for auxiliary judgment is determined, and the data of the combined sensor group for auxiliary judgment is read and input into the neural network judgment module; Fusion module: fuses the data of the combined sensor group, determines the sensor data value according to the RSSI value ratio, and makes corrections to determine the final sensor data result; Output module: The corrected sensor data is input into the neural network judgment module. When the judgment result shows either smoldering or open flame again, the current state is determined to be abnormal and a fire alarm is issued. When the judgment result shows no smoldering or open flame, the alarm is eliminated according to the error.

6. A neighborhood-based fire identification and warning system according to claim 5, It is characterized in that According to the specific installation conditions, the lowest density is the tangency produced by the minimum effective transmission recognition range of the temperature and humidity, CO, and smoke sensors.

7. A neighborhood-based fire identification and warning system according to claim 6, It is characterized in that The calculation method of the RSSI value ratio in the correction module is: RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n ); The calculation method of correction in the correction module is: sensor data (RSSI n / (RSSI 0 +RSSI 1 +...+RSSI n )).

8. A neighborhood-based fire identification and warning system according to claim 6, It is characterized in that A neighborhood group is formed by incorporating three combined sensor groups.

Images