A coal and gas outburst risk identification method based on PSO-CSA

Abstract

The present application relates to the field of intelligent coal mine gas sequence prediction, in particular to a coal and gas outburst risk identification method based on PSO-CSA, which is as follows: the present application uses clonal selection algorithm (PSO) to identify the risk of coal and gas outburst, uses particle swarm optimization algorithm (PSO) to improve the demand of CSA for the identification of coal and gas outburst risk, accelerates the convergence speed, improves the global search ability, eliminates the shock in the later stage of operation, improves the success rate of identification, and effectively identifies the outburst risk. The PSO is introduced into the mutation process of CSA, so that the mutation is no longer dependent on a large number of calculations of binary encoding and decoding, and the antibodies generated in the mutation process can also achieve the purpose of showing high affinity. A coal and gas outburst risk identification method based on PSO-CSA is established, which effectively identifies the risk of coal and gas outburst, has beneficial effects, and can be used to guide coal mine engineering practice.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent coal mine gas sequence prediction, in particular to a coal and gas outburst risk identification method based on PSO-CSA. BACKGROUND

[0002] At present, the shallow coal resources are consumed seriously, with the increase of mining depth, the ground stress and gas pressure are getting bigger and bigger, the influence of coal and rock mechanics factors such as ground stress on coal and gas outburst is not clear at present, and the mechanism of coal and gas outburst is still in the hypothesis stage, coal and gas outburst is still the primary threat factor in coal mine production, and the disaster problem that coal mine safety researchers need to solve urgently.

[0003] In the process of coal and gas outburst prevention and control, monitoring and prediction are the most important links, timely perception, timely identification and prediction can take preventive measures in time to eliminate coal and gas outburst in the germination stage, clonal selection algorithm (CSA) only pays attention to the influence of antibody and antigen affinity on B lymphocyte differentiation, is not affected by the affinity between antibodies, the algorithm only selects and clones the strongest stimulated individual, and has the advantages of autonomous learning, identification and memory, which meets the needs of coal and gas outburst risk abnormal detection and identification, but it also has problems such as oscillation, insufficient convergence and easy to fall into local optimum. Particle swarm optimization algorithm (PSO) has the advantages of fast convergence speed and no need of repeated encoding and decoding, which is used to optimize the mutation process of CSA and improve the deficiencies in the mutation process of CSA. Thus, a coal and gas outburst risk identification method based on PSO-CSA is established to identify the risk of coal and gas outburst. SUMMARY

[0004] The present application aims to provide a coal and gas outburst risk identification method based on PSO-CSA, which solves the problems of uncertain mutation convergence direction and slow calculation speed of CSA, realizes global optimization of coal and gas outburst risk identification problem, further improves the calculation efficiency and accuracy, and provides a new idea and method for solving the risk identification of coal and gas outburst.

[0005] The purpose of the present application can be realized by the following technical scheme: a coal and gas outburst risk identification method based on PSO-CSA, the specific steps of the coal and gas outburst risk identification method based on PSO-CSA are as follows:

[0006] Step 1: collect the index data of local coal and gas outburst in coal mine underground, including drill cuttings S, drill cuttings gas desorption K1, and coal gas diffusion initial velocity As antigens, the data are uniformly processed and divided into training set and identification set;

[0007] Step 2: The clonal selection algorithm is optimized using a particle swarm optimization algorithm to establish a PSO-CSA coal and gas outburst risk identification model;

[0008] Step 3: Based on step 1, the training set is input into PSO-CSA to train the algorithm, stimulate the algorithm to produce antibodies, match antibodies with high affinity as memory cells, and update the memory cell set;

[0009] Step 4: Input the identification set data into the algorithm, calculate the antibody with the highest affinity to the antigen in the antibody, and determine whether it is greater than the critical value , output the identification result of the input antigen leading to the outburst;

[0010] Step 5: If there is un-identified data, re-perform step 4 to output the coal and gas outburst risk identification result.

[0011] Further, the specific steps of step 1 are:

[0012] Step 1.1: Take the initial velocity of gas emission of coal , drill cuttings S, drill cuttings gas desorption K1, three characteristic indexes of coal and gas outburst danger;

[0013] Step 1.2: The collected gas emission initial velocity , drill cuttings S, drill cuttings gas desorption K1, three characteristic indexes of coal and gas outburst danger are processed by normalization to obtain membership values, which are in the interval [0, 1]. The normalization processing standard is that 0.1 is used to represent data without outburst danger; 0.6 is used to represent general outburst danger; and 1 is used to represent serious outburst danger or obvious outburst danger, which exceeds the critical value;

[0014] According to the normalization principle, the model identifies the outburst danger on site according to the following rules:

[0015] When the identification result is less than 0.35, it is considered to have no outburst danger; when the identification result is between [0.35, 0.8], it is output as general outburst danger; and when the identification result is greater than 0.8, it is determined to have serious outburst danger.

[0016] Step 1.3: The data after normalization is divided into training set and identification set.

[0017] Further, the specific steps of step 2 are:

[0018] Step 2.1, the specific steps of CSA are:

[0019] Within the predictive space of the algorithm, the string length of the antibody and antigen genotypes is L, S represents the appropriate coordinate axis of the immune space, and the required variables in the CSA are defined first:

[0020] • Ag: antigen;

[0021] • Ab: available antibody table (Ab∈S N×L , Ab=Ab (r) ∪Ab (m) ) ;

[0022] • Ab (m) : memory antibody table (Ab (m) ∈S m×L , m≤N) ;

[0023] • Ab (r) : remaining antibody table (Ab (r) ∈S m×L , r=N-m) ;

[0024] • Ag (M) : successfully identified antigen group (Ag (M) ∈S M×L ) ;

[0025] • f j : affinity vector related to antigen Ag j ;

[0026] • Ab (H) : n antibodies in Ab with the maximum affinity value to Ag j (Ab (H) ∈S n×L , n≤N) ;

[0027] • C j : population composed of N c clones in Ab (H） (C j ∈S Nc×L ) ;

[0028] • C j* : population transformed from C j after affinity maturation;

[0029] • Ab (d) : d low-affinity antibodies in Ab (r） replaced by d molecules in C j* (Ab (d) ∈S d×L , d≤r) ;

[0030] • Ab (G) : memory antibody from C j*antibodies in the set Ab

[0031] The detailed steps of CSA algorithm are as follows:

[0032] Step 2.1.1: Randomly select an antigen Ag j (Ag j ∈ Ag), let it stimulate all antibodies in the set Ab = Ab (r) ∪ Ab (m) (r + m = N), and take the affinity of antibodies to antigen as the solution of the objective function by setting the optimized function g(x), each antibody Ab i represents an element of the input space;

[0033] Step 2.1.2: Calculate the affinity vector f j of N antibodies in Ab

[0034] Step 2.1.3: Select the n antibodies with the highest affinity to the antigen Ag j in Ab (H) ∈ S n×L , n ≤ N) to form a new set;

[0035] Step 2.1.4: The antibodies in the set Ab (H） will produce new clones according to their respective affinities in a set C j , Ab (H） The higher the affinity of the n antibodies in Ab j to the antigen, the more their own clones;

[0036] Step 2.1.5: All antibodies in the set C j* undergo a mutation process related to affinity to produce a set of mature clones C j* , the higher the affinity, the lower the antibody mutation rate;

[0037] Step 2.1.6: Calculate the affinity f j of the mature clone set C j* to the antigen Ag j ;

[0038] Step 2.1.7: Re-select the antibody with the highest affinity to Ag j from the clones in the set C j* and place it in the memory cell set Ab (m） , if the affinity of this antibody to the antigen Ag j is greater than the original memory cell, it will be replaced, if an antibody group is used to determine multiple optimal solutions of a problem, two variables are determined:

[0039] (1) Set n=N, that is, all antibodies in Ab are selected for cloning in step 2.1.3;

[0040] (2) Determine Ab (n） Number of antibody clones:

[0041]

[0042] In the formula N c It is a collection of clones C j The total number of antibodies, β is an influence parameter, round(x) is the rounding function, for the antibody with the maximum affinity, i.e., if i=1, β=1, N=100, then this antibody needs 100 clones, and the next antibody needs 50 clones.

[0043] Step 2.1.8: C j* d antibodies replace Ab (r) Collection of neutralizing antigens Ag j d antibodies with the minimum affinity.

[0044] After all M antigens have undergone steps 2.1.1 to 2.1.8 once, the algorithm has completed one generation. After step 2.1.3, the n antibodies with the highest affinity will be sorted from highest to lowest affinity, and their specific number of clones will be calculated using the following formula:

[0045]

[0046] Step 2.2: Since CSA uses a binary algorithm for data identification, output, and calculation, the algorithm needs to be re-decoded each time the affinity between antigen and antibody is calculated, thus increasing the computational load and time. Simultaneously, the mutation mechanism of the algorithm increases antibody diversity by randomly altering the feature vectors of existing cloned antibodies, producing the side effect of destroying antibodies with high affinity. This mutation process also increases the computational load. Therefore, the algorithm is optimized based on the original clone selection algorithm, with the following specific steps:

[0047] The convergence optimization of the algorithm makes it adaptive to high-probability mutations; that is, as the number of iterations increases, β in the algorithm decreases, and the decreasing formula is as follows:

[0048]

[0049] In the formula: i represents the ranking of the antibody in the permutation, k represents the number of iterations, and the parameter Satisfying 0< <1;

[0050] At the same time, the affinity of the cloned new antibody is calculated, the antibody clone is updated, the newly generated antibody replaces the original low affinity antibody, and the algorithm converges upward in the cloning process, and the affinity of the newly generated antibody is above the average affinity of the current population; if it is below the average, the cloned antibody is discarded, and a new antibody is randomly generated;

[0051]

[0052]

[0053]

[0054] When the generated antibody meets the antibody of formula (16), it is saved, if the continuously generated antibodies do not meet the requirements, or the iteration number exceeds the set threshold, the cloning is stopped, and the original affinity antibody is used;

[0055] To prevent the algorithm from appearing "premature" mutation in the running process, y index is used for detection, and the value =1 is set as the critical value, when y≤ , it is judged that the algorithm enters "premature" state, the cloning of the antibody with β assignment is suspended, and the method of suddenly magnifying the mutation probability to k times of the original probability is used to introduce new population to increase the algorithm to jump out of local optimum and converge; when y returns to y> , the algorithm continues according to β assignment.

[0056]

[0057] In the formula: f max is the maximum affinity of the antibody in this iteration, f min is the minimum affinity.

[0058] Step 2.3, the PSO optimizes the CSA, and the specific steps are as follows:

[0059] In the optimization process, each antibody in the set C j with high affinity is taken as a particle in space, and the initial speed v0 of each particle is randomly generated, so the initial speed of the i th antibody is v i0 , and the characteristic vector H vi0 =(h v1 ,h v2 ,…,h vq ) T of each particle speed is (M)The greater the affinity between the antibody and the antibody (Ab), the better the determination of the antibody's location. The iteration count for each antibody mutation is set to Q, and each particle iterates to the q-th generation position. The optimal location found is denoted as P. q Then P q =f max Where q = {1, 2, ..., Q}, according to this calculation method, the entire particle swarm is updated to the Qth generation, and the optimal position found is P. iq P iq =f max PSO is based on P 0q and P iq The speed of updating itself v in And the position, until q=Q, to achieve its optimized mutation of CSA, the process is described as follows:

[0060]

[0061]

[0062] in, The inertial weights, learning factors r1 and r2 are non-negative constants, and the number of random variations of r1, r2 ∈ [0, 1], constitute the antibody set C after PSO optimization of the variant antibodies. j* Furthermore, the total number of aggregated antibodies after mutation remains unchanged, still being n.

[0063] Furthermore, the specific steps of step 4 are as follows:

[0064] Set thresholds for no-prominence and prominence risks, the number of iterations, and the memory cell replacement threshold for the algorithm, starting with no prominence risks. If set Ab... (m) The highest affinity of memory cells in the adenosine monophosphate (A) for Ab is greater than 100%. , If ∈(0,1), then the identification is marked as normal, that is... If no prominent danger is found, the affinity of all memory cells in the Ab set is calculated. The memory cell with the highest affinity to Ab is determined to determine whether the Ab set data has a prominent danger, thereby achieving the purpose of identification.

[0065] The recognition success rate is calculated as follows:

[0066]

[0067] Among them, T suc (0≤T) suc ≤100) represents the successfully identified antigen set Ag (M)The number of antigens in the set, where T is the total number of Ag antigens in the recognition set.

[0068] The beneficial effects of this invention are:

[0069] This invention improves upon the CSA (Coal and Gas Outburst Analysis) for identifying coal and gas outbursts, accelerating its convergence speed, enhancing its global search capability, eliminating late-stage operational oscillations, and increasing the success rate of identification, thereby effectively identifying outburst anomalies. PSO (Polymerase Optimization) optimizes the CSA mutation process by leveraging its advantages of fast convergence speed and the elimination of repeated encoding and decoding, thus addressing the shortcomings of the CSA cloning mutation process. Based on this combination, a PSO-CSA-based method for identifying coal and gas outburst hazards is established. Introducing PSO into the CSA mutation process eliminates the reliance on extensive binary encoding and decoding calculations, while still achieving the goal of high affinity for antibodies generated during the mutation process. Attached Figure Description

[0070] The invention will now be further described with reference to the accompanying drawings.

[0071] Figure 1 CSA process;

[0072] Figure 2 PSO-CSA model training process;

[0073] Figure 3 Application process of coal and gas outburst risk identification method based on PSO-CSA;

[0074] Figure 4 Relationship between cloning rate u and recognition rate;

[0075] Figure 5 Relationship between iteration number d and recognition rate;

[0076] Figure 6 The relationship between the memory cell replacement threshold e and the recognition rate;

[0077] Figure 7 Is there a threshold for prominent risks? Relationship with recognition rate. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Please see Figure 1As shown, the present application is a coal and gas outburst risk identification method based on PSO-CSA, and the specific steps of the coal and gas outburst risk identification method based on PSO-CSA are as follows:

[0080] Step 1: Collect the index data of local coal and gas outburst in the coal mine, including drill cuttings S, drill cuttings gas desorption K1, and coal gas diffusion initial velocity As an antigen, the data is processed by normalization and divided into a training set and an identification set, and the specific steps are as follows:

[0081] Step 1.1: Take the coal gas diffusion initial velocity , drill cuttings S, and drill cuttings gas desorption K1 as three characteristic indexes of coal and gas outburst danger;

[0082] Step 1.2: The collected gas diffusion initial velocity , drill cuttings S, and drill cuttings gas desorption K1 are processed by normalization, and the membership values are obtained, which are in the interval [0, 1]. The normalization processing standard is that 0.1 is used to represent data without outburst danger; 0.6 is used to represent data with general outburst danger; and 1 is used to represent data with serious outburst danger or obvious outburst danger that exceeds the critical value;

[0083] According to the normalization principle, the model identifies the outburst danger on site according to the following rules:

[0084] When the identification result is less than 0.35, it is considered to have no outburst danger; when the identification result is between [0.35, 0.8], it is output as a general outburst danger result; and when the identification result is greater than 0.8, it is determined to have serious outburst danger.

[0085] Step 1.3: Divide the data processed by normalization into a training set and an identification set;

[0086] The data required for verification in the specific reference example comes from the selected coal seam 3 # Coal seam, due to the presence of many structures near the coal seam, coal and gas outburst has occurred many times, therefore, identifying and evaluating the outburst danger of this coal seam can verify the feasibility and value of the model, and the selected data is shown in Table 1:

[0087] Table 1 3 # Coal seam collects original data

[0088]

[0089] The original data collected in Table 1 is normalized to obtain the membership values in Table 2, which are in the interval [0, 1]. According to the normalization processing standard, 0.1 is used to represent no outstanding danger, 0.6 is used to represent general outstanding danger, and 1 is used to represent serious or obvious outstanding danger for data exceeding the critical value. According to this normalization principle, the model identifies the outstanding danger on site according to the following rules: for the identification result less than 0.35, it is considered as no outstanding danger; for the identification result between [0.35, 0.8], it is output as general outstanding danger; and for the result greater than 0.8, it is considered as serious outstanding danger;

[0090] Table 2 3 # Coal seam data normalization processing

[0091]

[0092] According to the calculation and analysis of different index data of the coal mine, different weights are used for the three indexes and the identification correlation degree to improve the success rate of identifying the outstanding danger;

[0093] Step 2: Use the particle swarm optimization algorithm to optimize the clonal selection algorithm, and establish a coal and gas outburst risk identification model based on PSO-CSA, the specific steps are as follows:

[0094] Step 2.1, the specific steps of CSA are as follows:

[0095] In the prediction space of the algorithm, the length of the string representing the genotype of the antibody and antigen is L, and S represents the appropriate coordinate axis of the immune space. First, define the required variables in CSA:

[0096] •Ag: antigen;

[0097] •Ab: available antibody table (Ab∈S N×L ,Ab=Ab (r) ∪Ab (m) );

[0098] •Ab (m) : memory antibody table (Ab (m) ∈S m×L ,m≤N);

[0099] •Ab (r) : remaining antibody table (Ab (r) ∈S m×L ,r=N-m);

[0100] •Ag (M) : successfully identified antigen group (Ag (M) ∈SM×L );

[0101] •f j : and antigen Ag j The relevant affinity vector;

[0102] •Ab (H) :Ab and Ag j There are n antibodies (Abs) with the highest affinity. (H) ∈S n×L (n≤N);

[0103] •C j Ab (H） N c A population of clones (C) j ∈S Nc×L );

[0104] •C j* :C j A group that has evolved after reaching a certain level of affinity;

[0105] •Ab (d) Ab (r） The d low-affinity antibodies were C j* It contains d molecules that are substituted (Ab) (d) ∈S d×L ,d≤r);

[0106] •Ab (G) : The memory antibody prepared from C j* Antibodies in;

[0107] The specific steps of the CSA algorithm are as follows:

[0108] Step 2.1.1: Randomly select an antigen Ag j (Ag j ∈Ag), so that it stimulates the antibody set Ab=Ab (r) ∪Ab (m) All antibodies in (r+m=N); by setting the optimal function g(x), the affinity of the antibody for the antigen is regarded as the solution of the objective function, and each antibody Ab... i Represents an element of an input space;

[0109] Step 2.1.2: Calculate the affinity vector f of the N antibodies in Ab. j ;

[0110] Step 2.1.3: Select Ab to neutralize antigen Ag j The n antibodies with the highest affinity form a new set (Ab). (H) ∈S n×L (n≤N);

[0111] Step 2.1.4: Ab (H） The antibodies in the set will produce new clones in a set ratio according to their respective affinities, and the set of clones is C j , Ab (H） The higher the affinity of the n antibodies and the antigen, the more their own clones;

[0112] Step 2.1.5: Set C j All antibodies in the set C j* The higher the affinity, the lower the antibody mutation rate;

[0113] Step 2.1.6: Calculate the mature clone set C j* The affinity f j of the antigen Ag j* ;

[0114] Step 2.1.7: Re-select the clones in set C j* The antibody with the highest affinity to Ag j is placed in the memory cell set Ab (m） If the affinity of this antibody to the antigen Ag j is greater than the original memory cell, it is replaced. If an antibody group is used to determine multiple optimal solutions to a problem, two variables are determined:

[0115] (1) Set n=N, that is, all antibodies in Ab are selected for cloning in step 2.1.3;

[0116] (2) Determine the number of clones of antibodies in Ab (n） :

[0117]

[0118] Where N c is the total number of antibodies in the clone set C j , β is an influence parameter, and round(x) is the rounding function. The antibody with the highest affinity is i=1, β=1, N=100, which requires 100, and the next ranked antibody requires 50 clones.

[0119] Step 2.1.8: d antibodies in C j* replace the d antibodies in Ab (r)） with the lowest affinity to the antigen Ag j .

[0120] When all the M antigens are executed once the above step 2.1.1 to step 2.1.8 process, then the algorithm performs a generation, after step 2.1.3, n affinity highest antibody will be sorted according to the affinity from high to low, by the following formula to calculate their specific clone quantity:

[0121]

[0122] Step 2.2: since the CSA to the data in the data set and output and calculation all use binary algorithm, so the algorithm in each time to antigen and antibody affinity calculation all need to decode, then the amount of calculation increases, the calculation time increases; at the same time, the algorithm of variation is realized by changing the original clone antibody feature vector, increase the antibody diversity, produce the side effect of destroying the high affinity antibody, and with the increase of mutation process, the calculation amount, on the basis of the original clone selection algorithm, its own optimization, the specific steps are:

[0123] The convergence optimization of the algorithm makes the algorithm adapt to the high probability of variation, that is, with the increase of the number of iterations, the β in the algorithm decreases, and its decreasing formula is:

[0124]

[0125] In the formula: i represents the rank of the antibody in the arrangement, k represents the number of iterations, and the parameter Satisfies 0 <1;

[0126] At the same time, the affinity of the new antibody is calculated, the antibody clone is updated, the new antibody replaces the original low affinity antibody, so that the algorithm converges upward in the cloning process, and the affinity of the new antibody is above the average affinity of the current population; If it is lower than the average, the cloned antibody is discarded, and a new antibody is randomly generated;

[0127]

[0128]

[0129]

[0130] When the generated antibody meets the antibody of formula (27), it is saved, and if the continuously generated antibodies do not meet the requirements, or the number of iterations exceeds the set threshold, the cloning is stopped, and the original affinity antibody is used;

[0131] To prevent the algorithm from appearing "premature" mutation during running, use y index to detect and set value =1 as critical value, when , judge the algorithm into "premature" state, suspend the use of β assignment of antibody clone, while using the method of suddenly amplifying the mutation probability to the original probability k times, introduce new population to increase the algorithm to jump out of local optimum, and convergence; when y returns to y> , then the algorithm continues according to the β assignment.

[0132]

[0133] In the formula: f max is the maximum affinity of antibody in this iteration, f min is the minimum affinity.

[0134] Step 2.3, PSO optimizes CSA, the specific steps are as follows:

[0135] In the optimization process, each antibody with high affinity in the set C j is regarded as a particle in space, and the initial velocity v0 of each particle is randomly generated. Therefore, the initial velocity of the i th antibody is v i0 , and the feature vector H vi0 of each particle velocity is v1 , h v2 , …, h vq ) T When the affinity between antigen Ag (M) and Ab is greater, it is determined that the position of the antibody is better. At the same time, the iteration number of each antibody mutation is set to Q, and each particle is iterated to the q th generation position. The optimal position searched is P q , P q =f max , wherein q={1, 2, …, Q}. According to this calculation method, the entire particle swarm is updated to the Q th generation, and the optimal position searched is P iq , P iq =f max . PSO updates its own velocity v 0q and position according to P iq and P in , until q=Q, and realizes the optimization of CSA mutation. The process is described as follows:

[0136]

[0137]

[0138] wherein, is the inertia weight, learning factors r1 and r2 are non-negative constants, r1, r2 ∈ [0, 1] are random change numbers, and the mutated antibodies after PSO optimization constitute the antibody set C j* , and the total number of the mutated set antibodies remains unchanged, still n;

[0139] Step 3: Based on step 1, input the training set into PSO-CSA to train the algorithm, stimulate the algorithm to produce antibodies, match antibodies with high affinity as memory cells, and update the memory cell set;

[0140] Step 4: Input the recognition set data into the algorithm, calculate the antibody with the highest affinity to the antigen, and judge whether it is greater than the critical value , output the recognition result of the input antigen leading to the degree of outburst, and the specific steps are:

[0141] Set the threshold values of the algorithm for no outburst danger and outburst danger, the number of iterations, and the replacement threshold value of the memory cells. Starting from no outburst, if the highest affinity of the memory cells in the set Ab (m） is greater than , ∈ (0, 1), the recognition label is normal, that is , it is judged that there is no outburst danger, the affinity of all memory cells in the Ab set is calculated, and whether the memory cell with the highest affinity to Ab has outburst danger is the outburst danger of the Ab set data, thereby achieving the purpose of recognition;

[0142] Table 3: Association weight of each index

[0143]

[0144] In this example, CSA and PSO-CSA methods are used to identify the index data of coal and gas outburst danger, and the recognition results of the two algorithms are compared to analyze the improvement effect of PSO on the recognition result of CSA algorithm before and after optimization.

[0145] The selected data is normalized, and the first 300 data is used as the training set to train the model, and then the last 100 data is used as the recognition set to identify the data exceeding the outburst index critical value using CSA, as abnormal data processing leading to outburst. The recognition threshold value of no outburst danger is set to 0.35 and 0.80, the number of iterations d is 100, and the memory cell replacement threshold value e is 0.99. The abnormal data recognition success rate is calculated under different cloning rates, iteration numbers, and threshold memory cell replacement conditions, and the results are shown in Figures 4-7 the grid point line graph;

[0146] Under the same data and setting conditions, the CSA algorithm is optimized by using PSO, and the result is as shown in the dotted line in the figure, wherein the identification success rate is calculated as follows: Figures 5-7

[0147]

[0148] Wherein, T suc (0≤T suc ≤100) is the number of antigens Ag (M)） in the successfully identified antigen set, and T is the total number of antigens in the identified set Ag;

[0149] In order to compare the improvement effect of the identification result of the CSA algorithm by PSO before and after optimization, the relationship between the cloning rate u, the iteration number d, the memory cell replacement threshold e, the presence or absence of the dangerousness threshold and the identification rate is shown in the figure Figures 4-7 It can be seen that the optimization effect of PSO on CSA is significant, and the identification accuracy is obviously higher than that before optimization.

[0150] Step 5: If there is un-identified data, step 4 is performed again, and the coal and gas outburst dangerousness identification result is output.

[0151] The above content is only an example and description of the structure of the present application, and those skilled in the art can make various modifications or supplements or use similar ways to replace the described specific embodiments, as long as they do not deviate from the structure of the present application or exceed the scope defined by the present application, which shall belong to the protection scope of the present application.​

Claims

1. A PSO-CSA-based coal and gas outburst risk identification method, characterized in that, The specific steps of the coal and gas outburst risk identification method based on PSO-CSA are as follows: Step 1: Collect the index data of local coal and gas outburst in coal mine underground, including drilling cuttings S, drilling cuttings gas desorption K1, coal gas diffusion initial velocity As an antigen, the data is normalized and divided into training set and identification set; Step 2: The particle swarm optimization algorithm is used to optimize the clonal selection algorithm, and a coal and gas outburst risk identification model based on PSO-CSA is established; Step 3: Based on step 1, the training set is input into PSO-CSA to train the algorithm, stimulate the algorithm to generate antibodies, match the antibodies with high affinity as memory cells, and update the memory cell set; Step 4: input the recognition set data into the algorithm, calculate the antibody with the highest affinity to the antigen in the antibody, determine whether it is greater than the critical value , output the recognition result of the input antigen leading to the degree of prominence; Step 5: If there is un-identified data, step 4 is re-performed, and the coal and gas outburst risk identification result is output; The specific steps of step 2 are as follows: The specific steps of CSA are as follows: In the prediction space of the algorithm, the length of the string representing the genotype of the antibody and antigen is L, and S represents the appropriate coordinate axis of the immune space. First, define the required variables in CSA: • Ag: antigen; • Ab: available antibody table (Ab e S N×L , Ab = Ab (r) ∪ Ab (m) ); • Ab (m) : memory antibody table (Ab (m) ∈ S m×L , m≤ N) ; • Ab (r) : remaining antibody table (Ab (r) ∈ S m×L , r = N - m) ; • Ag (M) : a group of antigens (Ag (M) ∈ S M×L ) that are successfully recognized; •f j : and antigen Ag j related affinity vector; • Ab (H) : Ab and Ag j n antibodies (Ab (H) ∈ S n×L ,n≤ N) with a maximum affinity • C j : Ab (H） in N c clones consisting of a population (C j ∈ S Nc×L ); • C j* : C j The population that is transformed after affinity maturation; • Ab (d) : Ab (r） d low affinity antibodies are C j* d molecules are substituted (Ab (d) ∈ S d×L ,d≤r); • Ab (G) : Preparation of antibodies from C j* : Preparation of antibodies from C The specific steps of the CSA algorithm are as follows: Step 2.1.1: Randomly select one antigen Ag j (Ag j ∈ Ag), let it stimulate the antibody set Ab = Ab (r) ∪ Ab (m) (r + m = N) of all antibodies; by setting an optimized function g(x), the affinity of antibodies to antigens is regarded as the solution of the objective function, each antibody Ab i represents an element of the input space; Step 2.1.2: Calculate affinity vectors f for N antibodies in Ab j ; Step 2.1.3: Select antibodies Ab that neutralize antigen Ag j The n antibodies with the highest affinity form a new set (Ab (H) ∈ S n×L ,n≤ N) Step 2.1.4: Ab (H） The antibodies of the collection are used to create new clones in a set ratio according to their respective affinities, to form a collection C of clones j , Ab (H） The higher the affinity of the n antibodies and the antigen, the more clones they have of themselves; Step 2.1.5: Set C j All antibodies in Set C are generated through an affinity-related variation process j* The higher the affinity, the lower the antibody variation rate; Step 2.1.6: Calculate the mature clone set C j* and antigen Ag j affinity f j* ; Step 2.1.7: Re-selecting the set C j* neutralizing Ag j The antibody with the highest affinity is put into the memory cell set Ab (m） If the affinity of this antibody to the antigen Ag j is greater than the existing memory cell, it is replaced. If a population of antibodies is used to determine multiple optimal solutions to a problem, two additional variables are determined: (1) Set n=N, that is, all antibodies in Ab are selected for cloning in step 2.1.3; (2) determining the number of clones of Ab (n） in the middle of the antibody: ； where N c is the total number of antibodies in the collection C j , β is an impact parameter, and round(x) is the rounding function. The antibody with the highest affinity is the one with N = 100, the next one with N = 50, and so on. Step 2.1.8: C j* d antibodies of the set and the antigen Ag (r) d antibodies of the set and the antigen Ag j d antibodies of the set and the antigen Ag When all M antigens have performed the above steps 2.1.1 to 2.1.8, the algorithm performs a generation. After step 2.1.3, the n antibodies with the highest affinity are sorted from high to low according to the affinity, and their specific clone quantities are calculated by the following calculation formula: ； Step 2.2: Since CSA uses a binary algorithm for data recognition and output in the data set and calculation, the algorithm needs to be re-decoded each time the affinity of the antigen and antibody is calculated, which increases the calculation amount and the calculation time. At the same time, the mutation of the algorithm is realized by randomly changing the feature vector of the original cloned antibody, which increases the diversity of the antibody and has the side effect of destroying the high-affinity antibody. The mutation process increases the calculation amount, and the algorithm is optimized based on the original clonal selection algorithm. The specific steps are as follows: Optimize the convergence of the algorithm, so that the algorithm is self-adaptive to high probability mutation, that is, as the number of iterations increases, the beta in the algorithm decreases, and its decreasing formula is: ； where i denotes the rank of the antibody in the ranking, k denotes the iteration number, and the parameters satisfies 0 <1; At the same time, the affinity of the new cloned antibody is calculated, the antibody cloning is updated, the newly generated antibody replaces the original low-affinity antibody, and the algorithm converges upward in the cloning process. The affinity of the newly generated antibody should be higher than the average affinity of the current population. If it is lower than the average, the cloned antibody is discarded and a new antibody is randomly generated; ； ； ； When the generated antibody meets the antibody of formula (6), it is saved. If the continuously generated antibodies do not meet the requirements, or the number of iterations exceeds the set threshold, the cloning is stopped, and the original affinity antibody is used; To prevent the algorithm from "premature" mutation during running, the y index is used for detection, and the value is set =1 as the critical value, when y≤ , it is judged that the algorithm enters the "premature" state, the clone of the antibody using β assignment is suspended, a new population is introduced using the method of suddenly amplifying the mutation probability to k times of the original probability to increase the algorithm to jump out of the local optimum and converge; when y returns to y> , the algorithm continues according to β assignment; ； wherein: f max is the maximum affinity of the antibody for the iteration, f min is the minimum affinity; Step 2.3: PSO optimizes CSA, and the specific steps are as follows: During the optimization process, each antibody set with high affinity C j As particles within space, each particle's initial velocity v0 is randomly generated. Therefore, the initial velocity of the i-th antibody is vi. i0 The eigenvector H of each particle's velocity vi0 =(h v1 ,h v2 ,…,h vq ) T When antigen Ag (M) The greater the affinity between the antibody and the antibody (Ab), the better the determination of the antibody's location. The iteration count for each antibody mutation is set to Q, and each particle iterates to the q-th generation position. The optimal location found is denoted as P. q Then P q =f max Where q = {1, 2, ..., Q}, according to this calculation method, the entire particle swarm is updated to the Qth generation, and the optimal position found is P. iq P iq =f max PSO is based on P 0q and P iq The speed of updating itself v in And the position, until q=Q, to achieve its optimized mutation of CSA, the process is described as follows: ； ； wherein, is the inertia weight, learning factors r1 and r2 are non-negative constants, r1, r2 ∈ [0, 1] are random change numbers, and the mutated antibodies after PSO optimization constitute the antibody set C j* , and the total number of the mutated set antibodies remains unchanged, still n.

2. The coal and gas outburst risk identification method based on PSO-CSA according to claim 1, characterized in that, The specific steps of step 1 are as follows: Step 1.1: Collecting the initial velocity of gas emission of coal , drill cuttings quantity S, drill cuttings gas desorption K1 three coal and gas outburst dangerous characteristic indexes; Step 1.2: Initial velocity of gas emission collected The three characteristic indexes of coal and gas outburst danger, i.e., the drilling cuttings amount S, the drilling cuttings gas desorption K1, are subjected to uniformization processing to obtain the membership values, the intervals of which are all in [0, 1]. The uniformization processing standard used is that 0.1 is used to represent the data without outburst danger; 0.6 is used to represent the data with general outburst danger; 1 is used to represent the data with serious outburst danger or obvious outburst danger, which is collected and exceeds the critical value. According to the normalization principle, the model identifies the outburst risk on site according to the following rules: When the identification result is less than 0.35, it is considered to have no outburst risk; when the identification result is between [0.35, 0.8], the result of general outburst risk is output; when the identification result is greater than 0.8, it is determined to have serious outburst risk; Step 1.3: The data after homogenization is divided into training set and identification set.

3. The PSO-CSA based coal and gas outburst risk identification method according to claim 1, characterized in that, The specific steps of step 4 are as follows: The algorithm sets the threshold of non-outstanding danger and outstanding danger, the number of iterations, and the replacement threshold of memory cells. Starting from non-outstanding, if the highest affinity of the memory cells in the set Ab (m) is greater than , ∈(0, 1), the identification mark is normal, that is , it is judged that there is no outstanding danger, the affinity of all memory cells in the Ab set is calculated, and whether the memory cell with the highest affinity of Ab has outstanding danger is the outstanding danger of the Ab set data, thereby achieving the identification purpose. The identification success rate is calculated as follows: ； where T suc (0≤T suc ≤100) is the number of antigens in the successfully identified antigen set Ag (M) T is the total number of antigens in the recognition set Ag.

Images