Lightning protection safety risk assessment method

Abstract

The invention relates to a lightning protection safety risk assessment method comprising the steps that a first lightning disaster assessment model and a second lightning disaster assessment model are constructed, the first lightning disaster assessment model adopts a deep neural network, and the second lightning disaster assessment model adopts the combination of an analytic hierarchy process and a fuzzy evaluation method to obtain a risk assessment value; and finally determining an assessment risk level according to the risk assessment value. Inputting all historical thunder disaster data into a second thunder disaster assessment model to obtain a first assessment risk level; inputting each piece of historical thunder disaster data into the trained first thunder disaster assessment model to obtain a second assessment risk level; correcting parameters of the second thunder disaster assessment model according to the second assessment risk level and the first assessment risk level; and finally, acquiring to-be-evaluated lightning protection safety data, and inputting the to-be-evaluated lightning protection safety data into the corrected second lightning disaster evaluation model to obtain a lightning protection safety risk evaluation grade. According to the method, the evaluation accuracy and the evaluation efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of lightning protection, and in particular to a lightning protection safety risk assessment method. Background Art

[0002] At present, the assessment of lightning disaster risk is mainly aimed at a certain area or a group of buildings. The indicators selected for the assessment are relatively simple and limited.

[0003] In order to improve the accuracy of lightning disaster risk assessment, a Chinese invention patent with patent number ZL202110539370.2 (authorization announcement number CN113327022B) discloses a lightning protection safety risk management system and method, which includes: a data acquisition module for remotely collecting lightning protection safety performance data of each lightning protection component of the project to be evaluated; a lightning fixed-point monitoring module for real-time monitoring of lightning activities at lightning monitoring stations in a set area; a lightning dynamic risk assessment module for dynamically analyzing and evaluating the background laws of lightning activities, lightning activity paths and the lightning disaster risks of the project to be evaluated based on the data monitored in real time by the lightning fixed-point monitoring module, and providing lightning protection recommendations; an early warning module for issuing early warning information based on the early warning range of lightning at the location of the project to be evaluated.

[0004] Although the above-mentioned lightning protection safety risk management system can monitor lightning activities in combination with real-time meteorological detection data, dynamically assess lightning risks, and better play the role of lightning disaster assessment, the above-mentioned lightning protection safety risk management system has the following limitations: The above-mentioned lightning protection safety risk management system requires real-time monitoring of lightning activities at lightning monitoring stations in a set area. This management method increases the workload of lightning protection safety risks, makes the lightning protection safety risk management method complicated and time-consuming, and reduces the management efficiency of lightning protection safety risks. Therefore, further improvements to the existing technology are needed. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a lightning protection safety risk assessment method with improved assessment accuracy and efficiency in view of the above-mentioned prior art.

[0006] The technical solution adopted by the present invention to solve the above technical problems is: a lightning protection safety risk assessment method, characterized by comprising the following steps:

[0007] Step 1: Collect historical lightning disaster data;

[0008] Step 2: construct a first lightning disaster assessment model. The first lightning disaster assessment model uses a deep neural network, and uses part of the historical lightning disaster data to form a training set. The training set is used in batches to train the constructed first lightning disaster assessment model to obtain the first lightning disaster assessment model after training.

[0009] Step 3: construct a second lightning disaster assessment model. The second lightning disaster assessment model uses a combination of hierarchical analysis method and fuzzy evaluation method to obtain a risk assessment value, and finally confirms the assessment risk level according to the risk assessment value; input all historical lightning disaster data into the second lightning disaster assessment model to obtain the first assessment risk level corresponding to each historical lightning disaster data;

[0010] Step 4: input each historical lightning disaster data into the first lightning disaster assessment model trained in step 2 to obtain a second assessment risk level corresponding to each historical lightning disaster data;

[0011] Step 5: According to the comparison result between the second assessment risk level and the first assessment risk level, it is determined whether the parameters of the second lightning disaster assessment model need to be corrected, and finally a corrected second lightning disaster assessment model is obtained;

[0012] Step 6: Collect the lightning protection safety data to be evaluated, and input the lightning protection safety data to be evaluated into the revised second lightning disaster assessment model to obtain the lightning protection safety risk assessment level.

[0013] In order to improve the training speed of the first lightning disaster assessment model, the following steps are also included before training the constructed first lightning disaster assessment model in step 2:

[0014] The historical lightning disaster data were cleaned, feature extracted and manually labeled to obtain the main hazard factors of lightning disasters, which were used as the input of the first lightning disaster assessment model.

[0015] Preferably, the first lightning disaster assessment model adopts a convolutional neural network, a recurrent neural network or a BP neural network.

[0016] In the above scheme, the specific risk assessment process of the second lightning disaster assessment model in step 3 is:

[0017] Step 3-1: Determine a set of evaluation index factors that affect the lightning protection safety risk level based on all historical lightning disaster data;

[0018] Step 3-2, using the analytic hierarchy process to calculate the weight of each evaluation index factor in the evaluation index factor set;

[0019] Step 3-3, using the fuzzy evaluation method to perform fuzzy comprehensive evaluation on each evaluation index factor, and calculating the comprehensive evaluation value of each evaluation index factor;

[0020] Step 3-4: Use the risk assessment matrix and the fuzzy comprehensive evaluation results of each assessment risk to calculate the risk assessment value, and confirm the assessment risk level according to each risk assessment value.

[0021] Specifically, the fuzzy comprehensive evaluation process in step 3-3 includes establishing the membership function of each evaluation index factor, and multiplying each evaluation index factor with the corresponding membership function to obtain a total fuzzy evaluation matrix; in addition, multiplying the weight of each evaluation index factor with the total fuzzy evaluation matrix to obtain a comprehensive evaluation value of each evaluation index factor.

[0022] As an improvement, the specific steps in step 5 are:

[0023] If the second assessment risk level in step 5 is different from the first assessment risk level, the weight of at least one assessment index factor and the membership function of at least one assessment index factor in the second lightning disaster assessment model are modified;

[0024] If the second assessment risk level in step 5 is the same as the first assessment risk level, there is no need to modify the second lightning disaster assessment model.

[0025] Compared with the prior art, the advantages of the present invention are: by constructing the first lightning disaster assessment model and the second lightning disaster assessment model, on the one hand, since the first lightning disaster assessment model adopts a deep neural network, the deep neural network learns the mapping relationship between input data and output data, and can predict the corresponding assessment risk level according to the input lightning disaster data, the first lightning disaster assessment model after training can guide the second lightning disaster assessment model to perform parameter correction, thereby improving the assessment accuracy of the second lightning disaster assessment model; on the other hand, since the second lightning disaster assessment model adopts a combination of hierarchical analysis method and fuzzy evaluation method, this method has a fast calculation speed compared to the assessment method of deep neural network. Therefore, the lightning protection safety risk assessment method in this application improves the assessment accuracy and efficiency, and there is no need to monitor the lightning activities of the lightning monitoring stations in the set area in real time. It is only necessary to regularly update the first lightning disaster assessment model and the second lightning disaster assessment model according to historical data. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a specific flow chart of the lightning protection safety risk assessment method in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The present invention is further described in detail below with reference to the accompanying drawings.

[0028] like Figure 1 As shown, the lightning protection safety risk assessment method in this embodiment includes the following steps:

[0029] Step 1: Collect historical lightning disaster data;

[0030] Step 2: construct a first lightning disaster assessment model. The first lightning disaster assessment model uses a deep neural network, and uses part of the historical lightning disaster data to form a training set. The training set is used in batches to train the constructed first lightning disaster assessment model to obtain the first lightning disaster assessment model after training.

[0031] In this embodiment, the first lightning disaster assessment model adopts a convolutional neural network, a recurrent neural network or a BP neural network;

[0032] In this embodiment, the following steps are also included before training the constructed first lightning disaster assessment model:

[0033] The historical lightning disaster data are cleaned, feature extracted and manually labeled to obtain the main hazard factors of lightning disasters, and the main hazard factors of lightning disasters are used as the input of the first lightning disaster assessment model constructed;

[0034] In this way, the training speed of the first lightning disaster assessment model can be improved by preprocessing the lightning disaster data;

[0035] Step 3: construct a second lightning disaster assessment model. The second lightning disaster assessment model uses a combination of hierarchical analysis method and fuzzy evaluation method to obtain a risk assessment value, and finally confirms the assessment risk level according to the risk assessment value; input all historical lightning disaster data into the second lightning disaster assessment model to obtain the first assessment risk level corresponding to each historical lightning disaster data;

[0036] The specific risk assessment process of the second lightning disaster assessment model is as follows:

[0037] Step 3-1: Determine a set of evaluation index factors that affect the lightning protection safety risk level based on all historical lightning disaster data;

[0038] Step 3-2, using the analytic hierarchy process to calculate the weight of each evaluation index factor in the evaluation index factor set;

[0039] Step 3-3, using the fuzzy evaluation method to perform fuzzy comprehensive evaluation on each evaluation index factor, and calculating the comprehensive evaluation value of each evaluation index factor;

[0040] In this embodiment, the fuzzy comprehensive evaluation process includes establishing the membership function of each evaluation index factor, and multiplying each evaluation index factor with the corresponding membership function to obtain a total fuzzy evaluation matrix; in addition, multiplying the weight of each evaluation index factor with the total fuzzy evaluation matrix to obtain a comprehensive evaluation value of each evaluation index factor;

[0041] Step 3-4: Use the risk assessment matrix and the fuzzy comprehensive evaluation results of each assessment risk to calculate the risk assessment value, and confirm the assessment risk level according to each risk assessment value.

[0042] Step 4: input each historical lightning disaster data into the first lightning disaster assessment model trained in step 2 to obtain a second assessment risk level corresponding to each historical lightning disaster data;

[0043] Step 5: According to the comparison result between the second assessment risk level and the first assessment risk level, it is determined whether the parameters of the second lightning disaster assessment model need to be corrected, and finally a corrected second lightning disaster assessment model is obtained;

[0044] In this embodiment, if the second assessment risk level is different from the first assessment risk level, the weight of at least one assessment index factor and the membership function of at least one assessment index factor in the second lightning disaster assessment model are modified;

[0045] If the second assessment risk level is the same as the first assessment risk level, there is no need to modify the second lightning disaster assessment model;

[0046] Step 6: Collect the lightning protection safety data to be evaluated, and input the lightning protection safety data to be evaluated into the revised second lightning disaster assessment model to obtain the lightning protection safety risk assessment level.

[0047] This embodiment also relates to a computer-readable storage medium on which a computer program is stored. When the computer program is run by a processor, the lightning protection safety risk assessment method as described above is executed.

[0048] To facilitate understanding of the specific evaluation method of the second lightning disaster evaluation model in the present application, it is described in detail in this embodiment, specifically:

[0049] 1. Establish evaluation levels

[0050] Four evaluation levels are adopted: excellent (low risk), good (certain risk), medium (relatively high risk) and poor (high risk). The corresponding percentages are 0-25 points, 25-50 points, 50-75 points, and 75-100 points, and the intervals are as follows: excellent (low risk) (75, 100]; good (certain risk) [50, 75]; medium (relatively high risk) [25, 50]; poor (high risk) [0, 25):

[0051] x is the final calculated score

[0052] After statistics on the scores given by each expert, it was found that the frequency of low scores (i.e. 0-20 points) and high scores (i.e. 80-100 points) was extremely low. Therefore, adjustments were made based on the original evaluation level segmentation, and these two segments were cancelled. The four evaluation levels were re-divided into corresponding segments. The final results were excellent, good, medium, and poor, corresponding to percentage scores of 65-80 points, 50-65 points, 35-50 points, and 20-35 points, respectively. The interval expressions were excellent (low risk) (65, 80]; good (certain risk) [50, 65]; medium (higher risk) [35, 50]; and poor (high risk) [20, 35).

[0053] x is the final calculated score

[0054] The above-mentioned classification of evaluation levels is prior art and will not be elaborated here;

[0055] 2. Determine the set of evaluation index factors

[0056] A two-level indicator factor set is used. The first level is the large indicator factor set, which is 4 in total. The second level is the subdivided indicator factor set of each of the four large indicator factor sets, which are 5, 6, 5 and 4 corresponding detailed indicator factor sets respectively. The detailed content and subordinate relationship of each indicator factor set are as follows:

[0057]

[0058] The first level and the second level of each index factor set mentioned above are index factors summarized by those skilled in the art based on historical lightning disaster data in the prior art;

[0059] 3. Use the analytic hierarchy process (AHP) to determine the weights of the large indicator factor set

[0060] 3.1) Using the 1-9 scale method, the importance scale of each evaluation index factor is judged by combining the expert scores, and the judgment matrix M = [M ij ] 4*4

[0061] Among them, M ij Represents the relative importance of the i-th and j-th evaluation index factors, i, j∈[1,4]

[0062] 3.2) Perform consistency check on the generated judgment matrix M

[0063] 3.2.1) Calculate the maximum eigenvalue λ of the judgment matrix max

[0064] 3.2.2) Calculate the consistency index CI of the judgment matrix

[0065]

[0066] Where n is the matrix order, here n = 4

[0067] 3.2.3) Obtain the average random consistency index RI by looking up the corresponding matrix order, and calculate the consistency ratio CR

[0068] Order n 3 4 5 6 7 8 9 10 RI 0.52 0.89 1.12 1.26 1.36 1.41 1.46 1.49

[0069]

[0070] 3.2.4) Determine whether the consistency ratio CR is ≤ 0.1. If so, proceed to the next step; if not, return to step 3.1 to adjust the importance scale until the consistency ratio of the judgment matrix passes the test.

[0071] 3.3) Calculate the relative weights of the major index factors

[0072] After the judgment matrix passes the consistency test, the geometric mean method is used to obtain the weight vector ω of the relative weights of the major indicator factors. i The calculation formula is as follows:

[0073]

[0074] The weights of the major index factors are ω1, ω2, ω3, ω4 respectively.

[0075] 4. Use fuzzy evaluation method (FHE) to calculate the final evaluation level

[0076] 4.1) Experts are invited to subjectively score each refined indicator factor under the refined indicator factor set under the large indicator factor set. The scoring interval is the overall evaluation level interval, that is, [20,80]. After the scoring is completed, assuming that n experts participate in the scoring, the scoring result of each expert is recorded as S i , (i=1,2,…,n), then the final score of a refinement index factor is:

[0077]

[0078] 4.2) Establish the membership function of the index factor score

[0079] The comprehensive evaluation grade score is segmented, and the triangular cross membership function is considered. The characteristic of this function is to avoid the rough estimation of only one grade membership in a certain scoring interval, so that the final score can be more accurately mapped to the corresponding actual risk level. In addition, considering that the proportion of the evaluation grades on the two end sides should be relatively small, that is, the proportion of the excellent (low risk) and poor (high risk) evaluation grades is relatively small, the specific function is constructed as follows:

[0080]

[0081] 4.3) Obtain the membership degree of each refinement index factor score to each evaluation index

[0082] Obtain the membership vector (row vector) of each factor to the four evaluation levels. Taking the first refined indicator in the first large indicator factor set as an example, its membership vector is recorded as R 1j =[r 11 ,r 12 ,r 13 ,r 14 ], where j = 1, 2, 3, 4, corresponding to the membership of the four evaluation levels of excellent, good, medium and poor respectively.

[0083] The same membership vector calculation is performed on other refined index factors under the large index factor set, and all membership vectors are merged row by row to obtain the membership vector matrix of the large index factor set, which is recorded as the fuzzy evaluation matrix R of the large index factor set:

[0084] Where n is the number of refined index factors under the large index factor set, here n = 5

[0085] 4.4) Obtain the membership degree of a large index factor set to the evaluation level

[0086] 4.4.1) Repeat the analytic hierarchy process (AHP) to determine the weights of each detailed indicator factor under a large indicator factor set. Taking the first large indicator factor set as an example, note that n = 5 here, the weights of each detailed indicator factor are ω 11 ,ω 12 ,ω 13 ,ω 14 ,ω 15 .

[0087] 4.4.2) Multiply the weights (row vectors) of each refined index factor obtained by step 4.3 with the fuzzy evaluation matrix R of the large index factor set to obtain the large index factor membership result matrix b:

[0088]

[0089] where b 11 ,b 12 ,b 13 ,b 14 The membership of this large index factor set to the evaluation grades of excellent, good, medium and poor is

[0090] 4.5) Obtain the overall evaluation level

[0091] 4.5.1) Repeat 4.4 for the four large index factor sets to obtain the membership result matrix of each index factor, and combine the rows to obtain the total fuzzy evaluation matrix B:

[0092] Where n is the number of large index factors, here n = 4

[0093] 4.5.2) Multiply the weight of the large index factor by the total fuzzy evaluation matrix to obtain the final membership result matrix V:

[0094]

[0095] Among them, V1, V2, V3, and V4 correspond to the total membership of the four evaluation levels of excellent, good, medium, and poor respectively;

[0096] .6) Final evaluation grade calculation

[0097] The evaluation results are displayed in numerical form, and the final score X can be calculated by the following formula:

[0098]

[0099] Among them, D j It represents the index score designed after considering the "penalty effect". When the total membership degree tends to be poor, the index score is set to the lower limit of the interval, which is 20; when the total membership degree tends to be excellent, the index score is set to the upper limit of the interval, which is 80; when the total membership degree tends to be medium, the index score tends to the lower limit of the interval, which is 40; when the total membership degree tends to be good, the index score tends to the upper limit of the interval, which is 60; therefore, the total index score vector D j is [80,60,40,20] T ;

[0100] Substitute the final score X calculated last into the evaluation grade function to obtain the final evaluation grade.

[0101] This embodiment also involves a lightning protection safety risk assessment system, including a computer-readable storage medium, which stores a computer program and can be read and executed by a processor. When the computer program is executed by the processor, the lightning protection safety risk assessment method as described above is implemented.

[0102] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A lightning protection safety risk assessment method, characterized in that The steps include: Step 1: Collect historical lightning disaster data; Step 2: construct a first lightning disaster assessment model. The first lightning disaster assessment model uses a deep neural network, and uses part of the historical lightning disaster data to form a training set. The training set is used in batches to train the constructed first lightning disaster assessment model to obtain the first lightning disaster assessment model after training. Step 3: construct a second lightning disaster assessment model. The second lightning disaster assessment model uses a combination of hierarchical analysis method and fuzzy evaluation method to obtain a risk assessment value, and finally confirms the assessment risk level according to the risk assessment value; input all historical lightning disaster data into the second lightning disaster assessment model to obtain the first assessment risk level corresponding to each historical lightning disaster data; Step 4: input each historical lightning disaster data into the first lightning disaster assessment model trained in step 2 to obtain a second assessment risk level corresponding to each historical lightning disaster data; Step 5: According to the comparison result between the second assessment risk level and the first assessment risk level, it is determined whether the parameters of the second lightning disaster assessment model need to be corrected, and finally a corrected second lightning disaster assessment model is obtained; Step 6: Collect the lightning protection safety data to be evaluated, and input the lightning protection safety data to be evaluated into the revised second lightning disaster assessment model to obtain the lightning protection safety risk assessment level.

2. The lightning protection safety risk assessment method according to claim 1 is characterized in that: Before training the constructed first lightning disaster assessment model in step 2, the following steps are also included: The historical lightning disaster data were cleaned, feature extracted and manually labeled to obtain the main hazard factors of lightning disasters, which were used as the input of the first lightning disaster assessment model.

3. The lightning protection safety risk assessment method according to claim 2 is characterized in that: The first lightning disaster assessment model adopts a convolutional neural network, a recurrent neural network or a BP neural network.

4. The lightning protection safety risk assessment method according to any one of claims 1 to 3, characterized in that: The specific risk assessment process of the second lightning disaster assessment model in step 3 is: Step 3-1: Determine a set of evaluation index factors that affect the lightning protection safety risk level based on all historical lightning disaster data; Step 3-2, using the analytic hierarchy process to calculate the weight of each evaluation index factor in the evaluation index factor set; Step 3-3, using the fuzzy evaluation method to perform fuzzy comprehensive evaluation on each evaluation index factor, and calculating the comprehensive evaluation value of each evaluation index factor; Step 3-4: Use the risk assessment matrix and the fuzzy comprehensive evaluation results of each assessment risk to calculate the risk assessment value, and confirm the assessment risk level according to each risk assessment value.

5. The lightning protection safety risk assessment method according to claim 4 is characterized in that: The fuzzy comprehensive evaluation process in step 3-3 includes establishing the membership function of each evaluation index factor, and multiplying each evaluation index factor with the corresponding membership function to obtain a total fuzzy evaluation matrix; in addition, multiplying the weight of each evaluation index factor with the total fuzzy evaluation matrix to obtain a comprehensive evaluation value of each evaluation index factor.

6. The lightning protection safety risk assessment method according to claim 5 is characterized in that: The specific steps in step 5 are: If the second assessment risk level in step 5 is different from the first assessment risk level, the weight of at least one assessment index factor and the membership function of at least one assessment index factor in the second lightning disaster assessment model are modified; If the second assessment risk level in step 5 is the same as the first assessment risk level, there is no need to modify the second lightning disaster assessment model.

Images