CN115496353A - Intelligent risk assessment method for compressed natural gas filling station - Google Patents

Abstract

The invention relates to an intelligent risk assessment method for a compressed natural gas filling station; the method comprises the following construction steps: step 1, constructing a wedge wave multi-core support vector machine; step 2, determining an optimized wedge wave support vector machine optimized by an improved locust algorithm; step 3, determining a risk assessment index system of the compressed natural gas filling station; and 4, collecting data related to a risk index system of the compressed natural gas filling station and determining input sample data by utilizing an expert evaluation method and a questionnaire method. And 5, dividing the risk level of the compressed natural gas filling station into five grades and taking the risk grade as output sample data of the evaluation model. Step 6, collecting related risk assessment data of the compressed natural gas filling station and determining a training sample and a testing sample; and 7, training the evaluation model by using the training data sample and evaluating the risk level of the compressed natural gas filling station to be evaluated by using the trained evaluation model. The petrochemical enterprise oil refining device energy management system optimization method utilizing the structure is provided. The method and the device can improve the precision and the efficiency of risk assessment of the compressed natural gas filling station, and further can improve the safety management level of the compressed natural gas filling station.

Description

An Intelligent Risk Assessment Method for Compressed Natural Gas Filling Stations

technical field

The invention relates to an intelligent risk assessment method for a compressed natural gas filling station, in particular to an intelligent risk assessment method for a compressed natural gas filling station based on a wedge-shaped multi-core support vector machine optimized by an improved locust algorithm.

Background technique

Compressed natural gas has the characteristics of low cost, no pollution, and convenient use. Compressed natural gas filling stations often store a certain amount of natural gas through storage tanks. In recent years, the number of compressed natural gas filling stations has been increasing. The rapid growth of the compressed natural gas industry has led to site fragmentation and regulatory difficulties. In addition, the management work in many places is relatively weak, and the control methods are backward. Therefore, many new compressed natural gas filling stations have many loopholes and problems in business management, equipment monitoring, safety production, etc. Due to corrosion or material defects, natural gas storage tanks may leak, resulting in fires, explosions and other accidents. Safety production accidents occur from time to time, causing casualties and property losses. Therefore, the safety issues of compressed natural gas filling stations are increasingly attracting public attention. In view of the high risk status of compressed natural gas filling stations, it is necessary to carry out hazard identification risk analysis on possible risk factors in the production and operation of compressed natural gas filling stations. Combined with the statistical analysis results of accidents, the qualitative evaluation method is used to conduct a preliminary risk analysis on the production equipment of compressed natural gas filling stations, find out the causes of common accidents in compressed natural gas filling stations, and propose technical management measures for the safe production and operation of compressed natural gas filling stations. The CNG filling station risk assessment study will ensure the safety of the CNG filling station and its surrounding facilities.

In recent years, with the vigorous development of artificial intelligence technology, machine learning algorithms based on artificial intelligence have become more and more mature. The machine learning algorithm based on artificial intelligence can automatically obtain information and release it in real time to prevent major or serious damage. In addition, it has the advantages of low cost, low power consumption, and reliable information. Due to the strong complexity, nonlinearity, uncertainty and real-time nature of the risk assessment of compressed natural gas filling stations, the use of traditional mathematical models for risk assessment has certain limitations. The traditional evaluation method is subject to randomness and ambiguity, the operation is relatively complicated, and it lacks self-learning ability. The non-linear processing ability of support vector machine is realized by "kernel mapping" method. For kernel mapping, the kernel function must satisfy the Mercer condition. For example, the Gaussian kernel function is a widely used kernel function that shows good mapping performance when analyzing nonlinear problems. However, for existing kernel functions, SVM cannot approximate any function on a certain L2 ₍ R) subspace, because existing kernel functions cannot generate a complete set of basis on this subspace by translation.

Contents of the invention

The present invention aims at the problems existing in the above-mentioned prior art, and provides an intelligent risk assessment method for compressed natural gas refueling stations based on wedge-shaped multi-core support vector machines optimized by the locust algorithm. The invention can improve the precision and efficiency of the risk assessment of the compressed natural gas filling station, and further can improve the safety management level of the compressed natural gas filling station.

Technical scheme of the present invention comprises the steps:

Step 1: Build Wedgelet Multi-Core Support Vector Machine

Combining the Wedgelet transform which is robust to changes in different scales and directions and the Support Vector Machine, the Wedgelet SVM is established. Aiming at the blindness problem in the evaluation of single-kernel function support vector machine when processing machine learning tasks with multi-feature sets, the multi-core support vector machine is generated by the weighted linear addition of local and global kernel functions to classify the data, further improving the classification accuracy , so as to improve the risk assessment accuracy of compressed natural gas filling stations.

Step 1-1: Construct the combined kernel function, the corresponding formula is as follows:

In the formula, λ _i represents the weight coefficient, _ki (x, y) represents the single kernel function, and the expression is as follows:

In the formula, c represents the scale factor, d _i and e _i represent the translation factors.

Step 1-2: Define the regression function as follows:

表示映射函数,B表示补偿因子。In the formula, ω represents the weight variable,

Indicates the mapping function, and B indicates the compensation factor.

Steps 1-3: Define the objective function and boundary conditions as follows:

In the formula, L represents the number of samples, and D represents the penalty factor.

Steps 1-4: Use the Lagrangian duality to construct the Lagrangian objective function and corresponding boundary conditions as follows:

θ_i,

χ_i表示拉格朗日算子。In the formula,

θ _i ,

χ _i represents the Lagrangian operator.

According to formula (5), we can get:

ε_i的偏导，并且将结果回代至式(9)得到如下方程：Calculation formula (7) for w, B,

The partial derivative of ε _i , and substitute the result back into formula (9) to get the following equation:

The minimization equation can be derived through optimization as follows:

Step 1-5: Get the multi-core support vector machine by introducing the multi-core function, as shown below

Steps 1-6: Determine the decision function of the Wedgelet multi-core SVM as follows:

Step 2: Determine the improved locust algorithm to optimize wedge wave support vector machine

In order to improve the risk assessment effect of Wedgelet multi-core support vector machine, the improved locust optimization algorithm with better global optimization performance and higher convergence accuracy is applied to the optimization of penalty factors, kernel function parameters and weights of Wedgelet multi-core support vector machine.

Step 2-1: Initialize the basic parameters of the improved locust optimization algorithm, including population size, space dimension, maximum number of iterations and initial position.

Step 2-2: Update the position of the locust group according to the formula:

In the formula, T _d represents the target position of the locust swarm, η represents the attenuation coefficient, ω _l represents the weight coefficient, t represents the current iteration number, η _max and η _min represent the maximum and minimum attenuation factors. s(·) represents the interaction force function between locust populations, as shown below:

In the formula, f represents the attraction strength parameter, and r represents the attraction scale parameter.

Step 2-3: Adjust the position of a single locust using the Levy flight local search strategy as follows:

X＝X+10×s _ts L X (19)

In the formula, L represents the Levy flight step length, and s _ts represents the threshold function, which can be used to control the flight method and change probability of locusts.

L is calculated using the following formula:

L=μ/|v| ^1/β (20)

In the formula, β represents a constant between 0 and 2, and the parameters μ and v represent parameters that obey the normal distribution.

Step 2-4: When individual locusts search for the current optimal position, the original position is replaced; if no optimal solution is found, use the linearly decreasing parameters to randomly jump out of the strategy, as shown below:

P _i =(2-2rand(0,1))·P _i (21)

In the formula, P _i represents the i-th locust. When a new position P _i is found, if the new position is better than the current position, the new position replaces the current position.

Step 2-5: In order to make the attenuation coefficient decrease at a faster rate in the early stage of algorithm execution, ensure that the individual locusts in the population can quickly approach the optimal goal, and improve the convergence speed of the algorithm; is reduced, so that individual locusts can carefully search the space and avoid the algorithm from falling into local optimum. The attenuation factor η is adjusted using the decrement coefficient update strategy as follows:

In the formula, n represents the current iteration coefficient, and N represents the maximum number of iterations.

Step 3: Determine the risk assessment index system for compressed natural gas filling stations

Through the detailed collection and analysis of data on the design, construction, operation, operation, leakage, defect, personnel, social and economic aspects of compressed natural gas filling stations, the risk influencing factors of compressed natural gas filling stations are systematically, comprehensively and qualitatively identified. According to the risk formation mechanism, the risk index system of compressed natural gas filling stations is determined. Determine the first-level indicators and second-level indicators of the risk assessment system for natural gas filling stations.

Step 4: Collect data related to the risk index system of compressed natural gas filling stations, determine the first-level index value through expert evaluation method and questionnaire method, and determine the input sample data.

Step 5: Divide the risk level of compressed natural gas filling stations into five levels, namely level I (score range [0.90, 1.00]), level II (score range [0.75, 0.90]), level III (score range The interval is [0.60,0.75)), grade IV (score interval is [0.45,0.60)) and grade V (score interval is [0,0.45)). The risk level is used as the output sample data of the evaluation model.

Step 6: Collect data related to the risk assessment of compressed natural gas filling stations, and divide the data into two parts, namely training data samples and test data samples.

Step 7: Use the training data samples to train the evaluation model, and then use the trained evaluation model to evaluate the risk level of the compressed natural gas filling station to be evaluated, and determine the safety status of the compressed natural gas filling station to be evaluated.

The advantages and effects of the present invention are as follows:

The wedge-shaped multi-core support vector machine optimized by the improved locust algorithm constructed by the present invention can evaluate the risk of compressed natural gas filling stations, which can effectively improve the accuracy and efficiency of risk assessment of compressed natural gas filling stations, thereby improving the compression of natural gas filling stations. The effectiveness of the station risk assessment provides a strong theoretical basis for the development of risk prevention and control measures for compressed natural gas filling stations, and has a relatively broad development prospect.

Description of drawings

Figure 1 Training time for different risk assessment methods

detailed description

Example

Technical scheme of the present invention comprises the steps:

Step 1: Build Wedgelet Multi-Core Support Vector Machine

Combining the Wedgelet transform which is robust to changes in different scales and directions and the Support Vector Machine, the Wedgelet SVM is established. Aiming at the blindness problem in the evaluation of single-kernel function support vector machine when processing machine learning tasks with multi-feature sets, the multi-core support vector machine is generated by the weighted linear addition of local and global kernel functions to classify the data, further improving the classification accuracy , so as to improve the risk assessment accuracy of compressed natural gas filling stations.

Step 1-1: Construct the combined kernel function, the corresponding formula is as follows:

In the formula, λ _i represents the weight coefficient, _ki (x, y) represents the single kernel function, and the expression is as follows:

In the formula, c represents the scale factor, d _i and e _i represent the translation factors.

Step 1-2: Define the regression function as follows:

表示映射函数,B表示补偿因子。In the formula, ω represents the weight variable,

Indicates the mapping function, and B indicates the compensation factor.

Steps 1-3: Define the objective function and boundary conditions as follows:

In the formula, L represents the number of samples, and D represents the penalty factor.

Steps 1-4: Use the Lagrangian duality to construct the Lagrangian objective function and corresponding boundary conditions as follows:

θ_i,

χ_i表示拉格朗日算子。In the formula,

θ _i ,

χ _i represents the Lagrangian operator.

According to formula (5), we can get:

ε_i的偏导，并且将结果回代至式(9)得到如下方程：Calculation formula (7) for w, B,

The partial derivative of ε _i , and substitute the result back into formula (9) to get the following equation:

The minimization equation can be derived through optimization as follows:

Step 1-5: Get the multi-core support vector machine by introducing the multi-core function, as shown below

Steps 1-6: Determine the decision function of the Wedgelet multi-core SVM as follows:

Step 2: Determine the improved locust algorithm optimization optimization wedge wave support vector machine

In order to improve the risk assessment effect of Wedgelet multi-core support vector machine, the improved locust optimization algorithm with better global optimization performance and higher convergence accuracy is applied to the optimization of penalty factors, kernel function parameters and weights of Wedgelet multi-core support vector machine.

Step 2-1: Initialize the basic parameters of the improved locust optimization algorithm, including population size, space dimension, maximum number of iterations and initial position.

Step 2-2: Update the position of the locust group according to the formula:

In the formula, T _d represents the target position of the locust swarm, η represents the attenuation coefficient, ω _l represents the weight coefficient, t represents the current iteration number, η _max and η _min represent the maximum and minimum attenuation factors. s(·) represents the interaction force function between locust populations, as shown below:

In the formula, f represents the attraction strength parameter, and r represents the attraction scale parameter.

Step 2-3: Adjust the position of a single locust using the Levy flight local search strategy as follows:

X＝X+10×s _ts L X (19)

In the formula, L represents the Levy flight step length, and s _ts represents the threshold function, which can be used to control the flight method and change probability of locusts.

L is calculated using the following formula:

L=μ/|v| ^1/β (20)

In the formula, β represents a constant between 0 and 2, and the parameters μ and v represent parameters that obey the normal distribution.

Step 2-4: When individual locusts search for the current optimal position, the original position is replaced; if no optimal solution is found, use the linearly decreasing parameters to randomly jump out of the strategy, as shown below:

P _i =(2-2rand(0,1))·P _i (21)

In the formula, P _i represents the i-th locust. When a new position P _i is found, if the new position is better than the current position, the new position replaces the current position.

Step 2-5: In order to make the attenuation coefficient decrease at a faster rate in the early stage of algorithm execution, ensure that the individual locusts in the population can quickly approach the optimal goal, and improve the convergence speed of the algorithm; is reduced, so that individual locusts can carefully search the space and avoid the algorithm from falling into local optimum. The attenuation factor η is adjusted using the decrement coefficient update strategy as follows:

In the formula, n represents the current iteration coefficient, and N represents the maximum number of iterations.

Step 3: Determine the risk assessment index system for compressed natural gas filling stations

Through the detailed collection and analysis of data on the design, construction, operation, operation, leakage, defect, personnel, social and economic aspects of compressed natural gas filling stations, the risk influencing factors of compressed natural gas filling stations are systematically, comprehensively and qualitatively identified. According to the risk formation mechanism, the risk index system of compressed natural gas filling stations is determined. Determine the first-level indicators and second-level indicators of the risk assessment system for natural gas filling stations.

Step 4: Collect data related to the risk index system of compressed natural gas filling stations, determine the first-level index value through expert evaluation method and questionnaire method, and determine the input sample data.

Step 5: Divide the risk level of compressed natural gas filling stations into five levels, namely level I (score range [0.90, 1.00]), level II (score range [0.75, 0.90]), level III (score range The interval is [0.60,0.75)), grade IV (score interval is [0.45,0.60)) and grade V (score interval is [0,0.45)). The risk level is used as the output sample data of the evaluation model.

Step 6: Collect data related to the risk assessment of compressed natural gas filling stations, and divide the data into two parts, namely training data samples and test data samples.

Step 7: Use the training data samples to train the evaluation model, and then use the trained evaluation model to evaluate the risk level of the compressed natural gas filling station to be evaluated, and determine the safety status of the compressed natural gas filling station to be evaluated.

A specific example is as follows:

In the embodiment, a compressed natural gas filling station is selected for risk assessment. The scale of the compressed natural gas filling station is as follows: the filling capacity is 1000 standard cubic meters per hour, and the daily filling capacity is 10000 standard cubic meters. The compressed natural gas refueling station mainly supplies gas for small and medium-sized vehicles.

The parameters of the improved locust algorithm are set as follows: l=2.0, f=1.0, b _min ＝0.35, b _max ＝0.90, η _min ＝0.40, η _max ＝0.95, the population size is 350, and the maximum number of iterations is 400.

Using questionnaire method and expert investigation method to obtain 50 sets of relevant data for the risk assessment of the compressed natural gas filling station, the first 40 sets of data are used as training samples, and the last 10 sets of data are used as test samples.

In order to verify the effectiveness of the evaluation method proposed by the present invention, the locust algorithm (BSWSVM-LA) optimized by B-spline wavelet support vector machine and the support vector machine (SVM-PSA) optimized by particle swarm algorithm are also used to carry out risk analysis on training samples. Evaluate. The accuracy of different evaluation models was evaluated using mean absolute error (MAE) and mean absolute percentage error (MAPE). The risk assessment results of compressed natural gas filling stations based on different models are shown in Table 1.

Table 1 CNG risk assessment results based on different models

The training time of different models is shown in Fig. 1. It can be seen from Figure 1 that the training time of the evaluation method proposed by the present invention is 6.7s, which is the smallest among the three methods. Therefore, the risk evaluation method proposed by the present invention can improve the efficiency of risk evaluation of compressed natural gas filling stations.

Three evaluation methods trained by the training samples are used to evaluate the risk of the test samples, and the comparison results are shown in Table 2.

Table 2 Risk assessment results of test samples based on three methods

According to the calculation results in Table 2, it can be seen that the MAE range of the evaluation method proposed by the present invention is 0.82 to 0.92, the MAE of BSWSVM-LA is 3.27 to 3.68, and the MEA range of SVM-PSA is 4.17 to 4.57. In addition, the evaluation results of the evaluation methods proposed by the present invention are all correct, BSWSVM-LA evaluation results have 4 errors, and SVM-PSA evaluation results have 7 errors, therefore, the proposed WMKSVM-ILA has the highest evaluation correctness among the three models .

The analysis results show that the proposed risk assessment method for compressed natural gas filling stations based on the improved locust algorithm to optimize the wedge wave multi-core support vector machine can obtain the best results, and it is a risk assessment method with high practical value.

Claims (3)

1. A compressed natural gas filling station intelligent risk assessment method, is characterized in that comprising the steps: Step 1: Build Wedgelet Multi-Core Support Vector Machine Combining the Wedgelet transform which is robust to changes in different scales and directions and the Support Vector Machine, the Wedgelet SVM is established. Aiming at the blindness problem in the evaluation of single-kernel function support vector machine when processing machine learning tasks with multi-feature sets, the multi-core support vector machine is generated by the weighted linear addition of local and global kernel functions to classify the data, further improving the classification accuracy , so as to improve the risk assessment accuracy of compressed natural gas filling stations. Step 2: Determine the improved locust algorithm optimization optimization wedge wave support vector machine In order to improve the risk assessment effect of Wedgelet multi-core support vector machine, the improved locust optimization algorithm with better global optimization performance and higher convergence accuracy is applied to the optimization of penalty factors, kernel function parameters and weights of Wedgelet multi-core support vector machine. Step 3: Determine the risk assessment index system for compressed natural gas filling stations Through the detailed collection and analysis of data on the design, construction, operation, operation, leakage, defect, personnel, social and economic aspects of compressed natural gas filling stations, the risk influencing factors of compressed natural gas filling stations are systematically, comprehensively and qualitatively identified. According to the risk formation mechanism, the risk index system of compressed natural gas filling stations is determined. Determine the first-level indicators and second-level indicators of the risk assessment system for natural gas filling stations. Step 4: Collect data related to the risk index system of compressed natural gas filling stations, determine the first-level index value through expert evaluation method and questionnaire method, and determine the input sample data. Step 5: Divide the risk level of compressed natural gas filling stations into five levels, namely level I (score range [0.90, 1.00]), level II (score range [0.75, 0.90]), level III (score range The interval is [0.60,0.75)), grade IV (score interval is [0.45,0.60)) and grade V (score interval is [0,0.45)). The risk level is used as the output sample data of the evaluation model. Step 6: Collect data related to the risk assessment of compressed natural gas filling stations, and divide the data into two parts, namely training data samples and test data samples. Step 7: Use the training data samples to train the evaluation model, and then use the trained evaluation model to evaluate the risk level of the compressed natural gas filling station to be evaluated, and determine the safety status of the compressed natural gas filling station to be evaluated. 2. A kind of compressed natural gas filling station intelligent risk assessment method according to claim 1, it is characterized in that wedge wave multi-core support vector machine construction steps are as follows: Step 1-1: Construct the combined kernel function, the corresponding formula is as follows:

In the formula, λ _i represents the weight coefficient, _ki (x, y) represents the single kernel function, and the expression is as follows:

In the formula, c represents the scale factor, d _i and e _i represent the translation factors. Step 1-2: Define the regression function as follows:

In the formula, ω represents the weight variable,

Indicates the mapping function, and B indicates the compensation factor. Steps 1-3: Define the objective function and boundary conditions as follows:

In the formula, L represents the number of samples, and D represents the penalty factor. Steps 1-4: Use the Lagrangian duality to construct the Lagrangian objective function and corresponding boundary conditions as follows:

In the formula,

θ _i ,

χ _i represents the Lagrangian operator. According to formula (5), we can get:

Calculation formula (7) for w, B,

The partial derivative of ε _i , and substitute the result back into formula (9) to get the following equation:

The minimization equation can be derived through optimization as follows:

Step 1-5: Get the multi-core support vector machine by introducing the multi-core function, as shown below

Steps 1-6: Determine the decision function of the Wedgelet multi-core SVM as follows:

3. A kind of compressed natural gas filling station intelligent risk assessment method according to claim 1, it is characterized in that improving locust algorithm analysis specific steps are as follows: Step 2-1: Initialize the basic parameters of the improved locust optimization algorithm, including population size, space dimension, maximum number of iterations and initial position. Step 2-2: Update the position of the locust group according to the formula:

In the formula, T _d represents the target position of the locust swarm, η represents the attenuation coefficient, ω _l represents the weight coefficient, t represents the current iteration number, η _max and η _min represent the maximum and minimum attenuation factors. s(·) represents the interaction force function between locust populations, as shown below:

In the formula, f represents the attraction strength parameter, and r represents the attraction scale parameter. Step 2-3: Adjust the position of a single locust using the Levy flight local search strategy as follows: X＝X+10×s _ts L X (19) In the formula, L represents the Levy flight step length, and s _ts represents the threshold function, which can be used to control the flight method and change probability of locusts. L is calculated using the following formula: L=μ/|v| ^1/β (20) In the formula, β represents a constant between 0 and 2, and the parameters μ and v represent parameters that obey the normal distribution. Step 2-4: When individual locusts search for the current optimal position, the original position is replaced; if no optimal solution is found, use the linearly decreasing parameters to randomly jump out of the strategy, as shown below: P _i =(2-2rand(0,1))·P _i (21) In the formula, P _i represents the i-th locust. When a new position P _i is found, if the new position is better than the current position, the new position replaces the current position. Step 2-5: In order to make the attenuation coefficient decrease at a faster rate in the early stage of algorithm execution, ensure that the individual locusts in the population can quickly approach the optimal goal, and improve the convergence speed of the algorithm; is reduced, so that individual locusts can carefully search the space and avoid the algorithm from falling into local optimum. The attenuation factor η is adjusted using the decrement coefficient update strategy as follows:

In the formula, n represents the current iteration coefficient, and N represents the maximum number of iterations.

Images