An energy security risk assessment method and system considering multi-dimensional indicators

Abstract

The present invention discloses an energy security risk assessment method that considers multidimensional indicators, comprising the following steps: S1. Constructing a multidimensional energy security indicator system based on supply security, usage security, environmental security, and economic security; S2. Constructing an energy security assessment model using an autoencoder-BP neural network-entropy weighted TOPSIS approach; S3. Calculating energy sub-item evaluation indices based on the multidimensional energy security indicator system; S4. Improving the maximum entropy model and assessing energy security risks; S5. Determining grading standards for each evaluation indicator based on the multidimensional energy security indicator system, defining corresponding grade boundaries, and classifying risk levels; S6. Assessing the risk level of the area to be assessed based on the improved maximum entropy model. This method achieves a multidimensional, comprehensive analysis and assessment of energy security.

Description

Technical Field

[0001] The present invention relates to the field of energy security assessment, and in particular to an energy security risk assessment method and system considering multi-dimensional indicators. Background Art

[0002] Against the backdrop of global carbon reduction, environmental governance concepts have shifted, shifting from radical to incremental emissions reductions. Energy system operational security risks are increasing, necessitating the establishment of a clean, low-carbon, safe, and efficient energy security system. This requires a multi-dimensional, comprehensive analysis and assessment of energy security. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide an energy security risk assessment method and system that considers multi-dimensional indicators, which can realize multi-dimensional and comprehensive analysis and assessment of energy security.

[0004] In order to solve the above technical problems, a technical solution adopted by the present invention is:

[0005] An energy security risk assessment method considering multi-dimensional indicators includes the following steps:

[0006] S1. Construct a multi-dimensional energy security indicator system based on the dimensions of supply security, use security, environmental security, and economic security;

[0007] S2. Construct an energy security evaluation model based on autoencoder-BP neural network-entropy weight TOPSIS;

[0008] S3. Based on the multi-dimensional energy security indicator system, calculate the energy sub-item evaluation index respectively;

[0009] S4. Improve the maximum entropy model and evaluate energy security risks;

[0010] S5. Based on the multi-dimensional energy security indicator system, determine the grading standards for each evaluation indicator, divide the corresponding grade boundaries, and divide the risk levels;

[0011] S6. Evaluate the risk level of the area to be tested based on the improved maximum entropy model.

[0012] In order to solve the above technical problems, another technical solution adopted by the present invention is:

[0013] An energy security risk assessment system considering multi-dimensional indicators is used to implement the above-mentioned energy security risk assessment method considering multi-dimensional indicators, including:

[0014] The data input module is used to transmit the data of each dimension and each indicator and other parameters to the storage module;

[0015] A storage module is used to build a multi-dimensional energy security indicator system based on the four security dimensions of supply security, usage security, environmental security, and economic security;

[0016] An energy security evaluation index calculation module is used to construct an energy security evaluation model of autoencoder-BP neural network-entropy weight TOPSIS, and calculate the energy sub-item evaluation index based on the multi-dimensional energy security indicator system;

[0017] The energy security risk assessment module is used to improve the maximum entropy model and evaluate energy security risks, and evaluate the risk level of the area to be tested based on the improved maximum entropy model.

[0018] The beneficial effects of this invention include providing a multi-dimensional energy security risk assessment method and system. This system constructs a comprehensive energy security indicator system based on four dimensions: supply security, usage security, environmental security, and economic security. This system is used to assess the energy security level of a sample, enabling a multi-dimensional, comprehensive analysis and assessment of energy security, providing valuable support for energy security and energy structure upgrades and optimization. Furthermore, an improved maximum entropy model is used to derive the energy security level distribution pattern, reflecting the energy security risk level of each sample. Based on the energy security index measurement and energy security risk assessment results, the system further explores the key influencing factors of energy security risk. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of an energy security risk assessment method considering multi-dimensional indicators according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the system structure of an energy security risk assessment system that considers multi-dimensional indicators according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of an autoencoding method for energy security risk assessment considering multi-dimensional indicators according to an embodiment of the present invention;

[0022] Figure 4 This is a BP neural network evaluation flow chart of an energy security risk assessment method considering multi-dimensional indicators according to an embodiment of the present invention;

[0023] Figure 5 It is an energy security index BP neural network training of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0024] Figure 6 is the relative error of energy security index prediction in an energy security risk assessment method considering multi-dimensional indicators according to an embodiment of the present invention;

[0025] Figure 7 It is a comparison between the actual value and the predicted value of the energy security index of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0026] Figure 8 It is a supply security index BP neural network training of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0027] Figure 9 is the relative error of supply security index prediction in an energy security risk assessment method considering multi-dimensional indicators according to an embodiment of the present invention;

[0028] Figure 10 It is a comparison between the actual value and the predicted value of the supply security index of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0029] Figure 11 An energy security risk assessment method considering multi-dimensional indicators according to an embodiment of the present invention uses a security index BP neural network training;

[0030] Figure 12 It is a relative error of the safety index prediction of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0031] Figure 13 The present invention provides an energy security risk assessment method that considers multi-dimensional indicators, and compares the actual value and predicted value of the security index.

[0032] Figure 14 It is an environmental safety index BP neural network training of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0033] Figure 15 is a relative error of environmental safety index prediction in an energy security risk assessment method considering multi-dimensional indicators according to an embodiment of the present invention;

[0034] Figure 16 It is a comparison between the actual value and the predicted value of the environmental safety index of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0035] Figure 17 It is an economic security index BP neural network training of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention;

[0036] Figure 18 is the relative error of economic security index prediction in an energy security risk assessment method considering multi-dimensional indicators according to an embodiment of the present invention;

[0037] Figure 19 This is a comparison between the actual value and predicted value of the economic security index of an energy security risk assessment method considering multi-dimensional indicators in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0039] Glossary:

[0040] Maximum Entropy Principle: The Maximum Entropy Principle is a criterion for selecting the statistical properties of random variables that best match the objective situation. It is also known as the Maximum Information Principle. The probability distribution of random quantities is difficult to measure. Generally, only various means (such as mathematical expectations and variances) or values ​​under certain known conditions (such as peak values ​​and the number of values) can be measured. There can be many, even infinite, distributions that match these values. Typically, one of these distributions has the highest entropy.

[0041] Autoencoder method: An unsupervised machine learning algorithm that can reduce the dimensionality of data. Its basic principle is to use a neural network encoder to obtain a new set of low-dimensional data from the input high-dimensional data, and then use a neural network decoder to output a set of high-dimensional data, in an attempt to reconstruct the original input. Figure 2 By minimizing the reconstruction error between input and output data, the model is trained to obtain the weights of each edge in the neural network. The resulting encoding can be considered the main features of the input data. This algorithm model can be applied to the calculation of indicator weights.

[0042] The BP neural network model is a multi-layer feedforward network trained using the back propagation error algorithm. It is one of the most widely used neural network models. BP networks can learn and store a large number of input-output pattern mappings without having to reveal the mathematical equations that describe these mappings in advance.

[0043] Please refer to Figure 1 The embodiment of the present invention provides an energy security risk assessment method considering multi-dimensional indicators, including the following steps:

[0044] S1. Construct a multi-dimensional energy security indicator system based on the dimensions of supply security, use security, environmental security, and economic security;

[0045] S2. Construct an energy security evaluation model based on autoencoder-BP neural network-entropy weight TOPSIS;

[0046] S3. Based on the multi-dimensional energy security indicator system, calculate the energy sub-item evaluation index respectively;

[0047] S4. Improve the maximum entropy model and evaluate energy security risks;

[0048] S5. Based on the multi-dimensional energy security indicator system, determine the grading standards for each evaluation indicator, divide the corresponding grade boundaries, and divide the risk levels;

[0049] S6. Evaluate the risk level of the area to be tested based on the improved maximum entropy model.

[0050] As can be seen from the foregoing description, the present invention provides a multi-dimensional energy security risk assessment method. This method constructs a comprehensive energy security indicator system based on four dimensions: supply security, usage security, environmental security, and economic security. This system is used to assess the energy security level of a sample, enabling a multi-dimensional, comprehensive analysis and assessment of energy security, providing valuable support for energy security and energy structure upgrades and optimization. Furthermore, an improved maximum entropy model is used to derive the energy security level distribution pattern, reflecting the energy security risk level of each sample. Based on the energy security index measurement and energy security risk assessment results, the key influencing factors of energy security risk are further explored.

[0051] Furthermore, step S1 is specifically as follows:

[0052] S101. Energy structure indicators in the supply security dimension include energy composition, energy types, and energy uses;

[0053] The energy composition is measured by fossil energy consumption, with the proportion of coal as a specific indicator;

[0054] The energy types mentioned are measured by the diversity index, with the energy production diversity index (SWI) as a specific indicator: Where, P i represents the share of the i-th energy in the total;

[0055] The energy use is measured by the power generation structure, with the proportion of thermal power generation as a specific indicator;

[0056] S102. Energy structure indicators using the security dimension include energy demand, consumption structure, and consumption sensitivity;

[0057] The energy demand is measured by energy usage, with per capita energy usage as a specific indicator;

[0058] The consumption structure is measured by structural stability and structural differences. The structural stability is measured by the energy consumption diversity index, and the structural difference is measured by the energy use carbon emission coefficient CE:

[0059]

[0060] Where, Pi represents the share of the i-th energy use in the total energy consumption, Coef i represents the carbon emission coefficient of the i-th energy source;

[0061] The consumption sensitivity is measured by elasticity, and the specific indicator is energy consumption elasticity index = energy growth rate / economic growth rate;

[0062] S103. Energy structure indicators in the environmental security dimension include carbon emissions, air pollution, and greenhouse gas impacts;

[0063] The specific indicator of the carbon emission factor is CO2 emissions, the specific indicator of the atmospheric pollution factor is SO2 emissions; the specific indicator of the greenhouse effect factor is CO2 emissions / total energy consumption;

[0064] S104. Energy structure indicators for economic security include supply economy and energy accessibility;

[0065] The supply economy is measured by the production-consumption ratio and external dependence. The specific indicator of the production-consumption ratio is energy production / energy consumption, and the specific indicator of external dependence is the proportion of fossil energy imports to GDP.

[0066] The energy accessibility is measured by energy price and availability. The specific indicator of energy price is gasoline price, and the specific indicator of availability is per capita household electricity consumption.

[0067] From the above description, it can be seen that the present invention constructs a multi-dimensional and comprehensive energy security indicator system from the four dimensions of supply security, use safety, environmental safety and economic security, which is used to evaluate the energy security level of samples and realize multi-dimensional and comprehensive analysis and evaluation of energy security.

[0068] Furthermore, the BP neural network model in step S2 is specifically:

[0069] Determine the number of nodes and connection weights of each layer of the BP neural network, initialize the hidden layer threshold and output layer threshold, and set the learning rate and neuron activation function;

[0070] Calculating a hidden layer output using the input vector, the connection weights, and the hidden layer threshold;

[0071] Calculating the output layer output using the connection weights, the output layer threshold, and the hidden layer output;

[0072] The output layer output and the expected output are used to perform error calculation, weight update and threshold update in sequence, and it is determined whether the iteration is completed. If so, the calculation is completed; otherwise, the step of calculating the hidden layer output is returned to be executed.

[0073] From the above description, it can be seen that the neural network BP model has shown good results in various evaluation and prediction fields due to its ability to approximate nonlinear functions. It has excellent performance in engineering applications and has been successfully applied in various fields. Therefore, the present invention chooses to establish a BP neural network model to realize the weights on the autoencoder edges.

[0074] Furthermore, step S2 is specifically as follows:

[0075] Using a BP neural network model to obtain the weights of the edges in the autoencoder neural network, and using the weights to evaluate the energy security index;

[0076] The energy security index obtained by entropy weight TOPSIS evaluation is used as the input layer to supervise the BP neural network learning and construct an autoencoder-BP neural network-entropy weight TOPSIS energy security index evaluation model.

[0077] As can be seen from the above description, based on the energy security indicator system, specific measurement indicators can be used as input data in the autoencoder, and the third-level indicators, second-level indicators, and first-level indicators can be used as hidden layers in the neural network. The edges in the neural network are established based on the hierarchical relationship structure between the indicators. Entropy-weighted TOPSIS uses the entropy weight method to perform relevant calculations, quantify the weights of different indicators, and then use the TOPSIS method for analysis, which has the advantage of making the evaluation results more realistic.

[0078] Furthermore, step S3 is specifically as follows:

[0079] The indexes of the four security dimensions of supply security, consumption security, environmental security and economic security are used as the number of neurons in the four-layer input layer of the BP neural network model, and the energy security index is set as the output neuron, and the energy sub-item evaluation index is calculated respectively.

[0080] From the above description, it can be seen that based on the multi-dimensional energy security indicator system, the energy sub-item evaluation index is calculated separately to achieve a multi-dimensional and comprehensive analysis and evaluation of energy security.

[0081] Furthermore, step S4 is specifically as follows:

[0082] There are n energy security samples with m evaluation indicators forming a reference series:

[0083] x j ={x j (i)|i=1,…,m; j=1,…,n}

[0084] Suppose there are m evaluation indicators and t energy security evaluation standards forming a comparative series:

[0085] x h ={x h(i)|h=1,…,t;i=1,…,m}

[0086] Δ h (i)=|x j (i)-x h (i)|

[0087] The difference between two series is expressed by the grey correlation coefficient ξ i (x j , x h ) is expressed as:

[0088]

[0089] In the formula, hi min Δ h (i) represents the minimum difference between the two levels, hi max Δ h (i) represents the maximum difference between the two levels, ρ represents the resolution coefficient, Δ h (i) indicates the difference between two levels;

[0090] Calculate the correlation degree by combining the weights of each indicator:

[0091] Where Z i Indicates the weight of each indicator;

[0092] Uncertainty H j It can be expressed as:

[0093] Where u hj The probability that the jth sample belongs to the hth level of energy security is expressed as follows:

[0094]

[0095]

[0096] Construct the Lagrangian function:

[0097]

[0098] By calculating the uncertainty H when the information entropy is maximum j Derivative, solve the final evaluation result is:

[0099]

[0100] Where w i It represents the generalized distance between the ith indicator and the evaluation criteria at all levels. B is a constant, which represents the number of information entropies.

[0101] From the above description, we can see that the improved maximum entropy model is used to derive the distribution law of energy security levels, reflecting the energy security risk level status of each sample. Based on the energy security index measurement and energy security risk assessment results, the key influencing factors of energy security risks are further explored.

[0102] Furthermore, the corresponding level boundaries described in step S5 are specifically:

[0103] Based on the mean of each subsystem, calculate P 75 、P 50 、P 25 The quantile value serves as a reference benchmark for energy security severity grading;

[0104] The P 75 75% of energy security incidents in the region have a severity higher than this value, while 25% of energy security incidents have a severity lower than this value, indicating that the risk level is at a high level;

[0105] The P 50 50% of energy security incidents in the region are more severe than this value, and the other 50% are less severe than this value, indicating that the risk level is at a medium level;

[0106] The P 25 This means that 25% of energy security incidents in the region have a severity higher than this value, and another 75% of energy security incidents have a severity lower than this value, indicating that the risk level is at a low level.

[0107] From the above description, it can be seen that based on the multi-dimensional energy security indicator system, the grading standards of various evaluation indicators are determined, the corresponding grade boundaries are divided, and the risk levels are divided.

[0108] Furthermore, step S6 further includes:

[0109] Construct an obstacle factor model to identify and modify the obstacle factors of various evaluation indicators;

[0110] The construction of the obstacle factor model is specifically as follows:

[0111] The obstacle degree O is used to measure the influence of the i-th indicator in the j-th sample on the index. ij Expressed as:

[0112]

[0113] Where m is the number of index evaluation indicators; indicator weight ω i Indicates factor contribution F i Measures the contribution of the i-th indicator to the index; I ijIndicates the difference between the actual value of the i-th indicator in the j-th sample and the optimal value, and the indicator deviation I ij Expressed as (1-r ij ), then r ij is the value after index standardization; set the factor contribution F i , Index Deviation I ij , Obstacle Degree O uj Three basic variables;

[0114] The obstacle degree O of the jth sample uth system uj for:

[0115] Where u i is the number of indicators of the u-th system.

[0116] As can be seen from the above description, the calculated indicator barrier represents the degree of obstruction posed by the measured indicator to energy security development, while the calculated subsystem barrier represents the degree of obstruction posed by the measured subsystem to energy security development. The calculation formula shows that the subsystem barrier is calculated by summing the barrier values ​​of the indicators within that subsystem. Furthermore, a larger barrier value indicates a greater degree of obstruction to energy security development. Targeted corrections to subsystems or indicators with high barrier values ​​can effectively improve energy security.

[0117] Please refer to Figure 2 Another embodiment of this embodiment provides an energy security risk assessment system that considers multi-dimensional indicators, which is used to implement the above-mentioned energy security risk assessment method that considers multi-dimensional indicators, including:

[0118] The data input module is used to transfer the data of each dimension and each indicator and other parameters to the storage module.

[0119] A storage module is used to build a multi-dimensional energy security indicator system based on the four security dimensions of supply security, usage security, environmental security, and economic security;

[0120] An energy security evaluation index calculation module is used to construct an energy security evaluation model of autoencoder-BP neural network-entropy weight TOPSIS, and calculate the energy sub-item evaluation index based on the multi-dimensional energy security indicator system;

[0121] The energy security risk assessment module is used to improve the maximum entropy model and evaluate energy security risks, and evaluate the risk level of the area to be tested based on the improved maximum entropy model.

[0122] The above-mentioned energy security risk assessment method and system considering multi-dimensional indicators of the present invention can realize multi-dimensional and comprehensive analysis and assessment of energy security. The following is an explanation through specific implementation methods:

[0123] Example 1

[0124] In this embodiment, based on the characteristics of the autoencoder reconstructing the original input, the BP neural network model is used to train the weights of the edges in the autoencoder neural network to evaluate the energy security index. Given that the BP neural network model is a "supervised" deep learning method, the entropy weighted TOPSIS method is used to evaluate the energy security index as the input layer to supervise the BP neural network learning, thereby constructing an autoencoder-BP neural network-entropy weighted TOPSIS energy security index evaluation model.

[0125] Please refer to Figure 1 , an energy security risk assessment method considering multi-dimensional indicators, including the following steps:

[0126] S1. Construct a multi-dimensional energy security indicator system based on the dimensions of supply security, use security, environmental security, and economic security. Specifically:

[0127] S101. Energy structure indicators in the supply security dimension include energy composition, energy types, and energy uses;

[0128] The energy composition is measured by fossil energy consumption, with the proportion of coal as a specific indicator;

[0129] The energy types mentioned are measured by the diversity index, with the energy production diversity index (SWI) as a specific indicator: Where, P i represents the share of the i-th energy in the total;

[0130] The energy use is measured by the power generation structure, with the proportion of thermal power generation as a specific indicator;

[0131] Table 1 Supply security evaluation indicators

[0132]

[0133] S102. Energy structure indicators using the security dimension include energy demand, consumption structure, and consumption sensitivity;

[0134] The energy demand is measured by energy usage, with per capita energy usage as a specific indicator;

[0135] The consumption structure is measured by structural stability and structural differences. The structural stability is measured by the energy consumption diversity index, and the structural difference is measured by the energy use carbon emission coefficient CE:

[0136]

[0137] Where, P irepresents the share of the i-th energy use in the total energy consumption, Coef i represents the carbon emission coefficient of the i-th energy source;

[0138] The consumption sensitivity is measured by elasticity, and the specific indicator is energy consumption elasticity index = energy growth rate / economic growth rate;

[0139] Table 2 Safety evaluation indicators for use

[0140]

[0141] S103. Energy structure indicators in the environmental security dimension include carbon emissions, air pollution, and greenhouse gas impacts;

[0142] The specific indicator of the carbon emission factor is CO2 emissions, the specific indicator of the atmospheric pollution factor is SO2 emissions; the specific indicator of the greenhouse effect factor is CO2 emissions / total energy consumption;

[0143] Table 3 Environmental safety evaluation indicators

[0144]

[0145] S104. Energy structure indicators for economic security include supply economy and energy accessibility;

[0146] The supply economy is measured by the production-consumption ratio and external dependence. The specific indicator of the production-consumption ratio is energy production / energy consumption, and the specific indicator of external dependence is the proportion of fossil energy imports to GDP.

[0147] The energy accessibility is measured by energy price and availability. The specific indicator of energy price is gasoline price, and the specific indicator of availability is per capita household electricity consumption.

[0148] Table 4 Economic security evaluation indicators

[0149]

[0150] S2. Construct an energy security evaluation model based on autoencoder, BP neural network and entropy weight TOPSIS, specifically:

[0151] Please refer to Figure 3 and Figure 4, using a BP neural network model to obtain the weights of the edges in the autoencoder neural network, and using these weights to evaluate the energy security index. Based on the energy security indicator system, specific measurement indicators can be used as input data for the autoencoder, and the third-level indicators, second-level indicators, and first-level indicators can be used as hidden layers of the neural network. The edges in the neural network are established based on the hierarchical relationship structure between the indicators. The weights on the edges in the neural network obtained from model training can be used as weights for indicator aggregation, and the data dimensionality reduction results obtained by the autoencoder.

[0152] The energy security index obtained by entropy weight TOPSIS evaluation is used as the input layer to supervise the BP neural network learning and construct an autoencoder-BP neural network-entropy weight TOPSIS energy security index evaluation model.

[0153] Among them, the BP neural network model is specifically:

[0154] Determine the number of nodes and connection weights of each layer of the BP neural network, initialize the hidden layer threshold and output layer threshold, and set the learning rate and neuron activation function;

[0155] Calculating a hidden layer output using the input vector, the connection weights, and the hidden layer threshold;

[0156] Calculating the output layer output using the connection weights, the output layer threshold, and the hidden layer output;

[0157] The output layer output and the expected output are used to perform error calculation, weight update and threshold update in sequence, and it is determined whether the iteration is completed. If so, the calculation is completed; otherwise, the step of calculating the hidden layer output is returned to be executed.

[0158] The BP algorithm is a supervised training algorithm that requires the output of a neural network to guide the learning of the network model. The entropy-weighted TOPSIS method uses the entropy-weighted method to perform relevant calculations, quantifying the weights of different indicators. This method is then used for analysis, resulting in evaluation results that are more realistic. Therefore, this section uses the entropy-weighted TOPSIS method to assess energy security, using this method as input data for the BP neural network model.

[0159] Therefore, in this embodiment, in the conventional TOPSIS method, it is assumed that the weights of each indicator are consistent. However, this assumption is far from reality. The entropy-weighted TOPSIS method first quantifies the weights of different indicators through correlation calculations, and then uses the TOPSIS method for analysis, so that the results are more in line with reality. The specific steps of this method are as follows:

[0160] Assuming there are m evaluation objects and each object has n evaluation indicators, a judgment matrix is ​​established and standardized:

[0161] X=(x ij ) m×n (i=1,2,3,...,m; j=1,2,3,...,n),

[0162] Calculate the information entropy H and determine the weight ω:

[0163]

[0164] Weighted matrix calculation and determination of optimal solution and the worst solution

[0165] R=(r ij ) m×n ,r ij =ω j ·x i ' j (i=1,2,3,...,m; j=1,2,3,...,n),

[0166]

[0167] The Euclidean distance between the optimal solution and the worst solution for different items:

[0168]

[0169] Calculate the comprehensive evaluation index C i , the larger the value of the index, the greater the project advantage:

[0170]

[0171] S3. Based on the multi-dimensional energy security indicator system, calculate the energy sub-item evaluation index respectively, specifically:

[0172] The indexes of the four security dimensions of supply security, consumption security, environmental security and economic security are used as the number of neurons in the four-layer input layer of the BP neural network model, and the energy security index is set as the output neuron, and the energy sub-item evaluation index is calculated respectively.

[0173] In this embodiment, based on the established energy security index system, the indexes of the four dimensions of "supply security, consumption security, environmental security and economic security" are used as the number of neurons in the four-layer input layer of the BP model; the energy security index is the final evaluation result, so it is set as the output neuron; existing research and applications have shown that a neural network with one hidden layer can represent any continuous function with arbitrary precision. Since the reasonable determination of the number of network layers and the number of neurons in each layer can improve the accuracy of the BP neural network model, the hidden layer is set to one layer. In summary, the neural network structure is as follows: Figure 5 shown.

[0174] The closer the correlation coefficient is to 1, the higher the correlation between input and output. Figure 6 It can be seen that the goodness of fit of all test set samples is 0.9521. At the same time, the relative error remains between 0 and 0.1, the minimum error is close to 0, and the maximum error is close to 0.1, indicating that the prediction effect is relatively good. Figure 7 It can be seen that the trend of the actual value of the energy security index is generally consistent with the predicted value. Most of the predicted values ​​are higher than the value of entropy weight TOPSIS, and the extreme values ​​are basically fitted, which shows that the weights of each level adjusted by the BP neural network can measure the energy security index well.

[0175] Specifically, the supply security index is calculated as follows:

[0176] According to the index system, this embodiment uses the energy composition, energy type and energy use represented by the proportion of coal, energy production diversity index and thermal power development ratio as the three-layer input layer neurons of the supply security index BP model; the supply security index is the final evaluation result, so it is set as the output neuron; the hidden layer is set to one layer. In summary, the neural network structure is as follows Figure 8 shown.

[0177] The closer the correlation coefficient is to 1, the higher the correlation between input and output. Figure 9 It can be seen that the goodness of fit of all test set samples is 0.6870. At the same time, except for the extreme error values, most of the relative errors of the predictions are below 0.2, indicating that the prediction effect is relatively reasonable. Figure 10 It can be seen that the trends of the real value and the predicted value of supply security are basically consistent, and the trends of the real value and the predicted value of the supply security index are extremely consistent. When the index is closer to 1, the real value is greater than the predicted value, and when the index is closer to 0, the real value is mostly smaller than the predicted value. This shows that the weights of each level adjusted by the BP neural network can measure the supply security index more compactly.

[0178] The safety index is calculated as follows:

[0179] According to the index system, this embodiment uses energy demand, structural stability, structural difference and elasticity coefficient, which are characterized by per capita energy consumption, energy consumption diversity index, energy consumption carbon emission coefficient and energy consumption elasticity coefficient, as the number of neurons in the four-layer input layer of the safety index BP model; the safety index is the final evaluation result, so it is set as the output neuron; a neural network with one hidden layer can represent any continuous function with arbitrary precision, so the hidden layer is set to one layer. In summary, the neural network structure is as follows: Figure 11 shown.

[0180] The closer the correlation coefficient is to 1, the higher the correlation between input and output. Figure 12 As can be seen from the table, the goodness of fit of all test set samples is 0.9287. Except for one sample with a deviation exceeding 0.2, the relative errors of other samples are all controlled within 0.15, indicating that the prediction effect is relatively good. Figure 13 It can be seen that the actual value of the supply security index is highly fitted with the predicted value, and most of the values ​​overlap, which shows that the weights of each level adjusted by the BP neural network can well measure the use security index.

[0181] The environmental safety index is calculated as:

[0182] According to the indicator system, this embodiment uses carbon emissions, air pollution and environmental protection costs represented by CO2 emissions, SO2 emissions and CO2 emissions / total energy consumption as the three-layer input layer neurons of the BP model of environmental safety index; the environmental safety index is the final evaluation result, so it is set as the output neuron; a neural network with one hidden layer can represent any continuous function with arbitrary precision, so the hidden layer is set to one layer. In summary, the neural network structure is as follows: Figure 14 shown.

[0183] The closer the correlation coefficient is to 1, the higher the correlation between input and output. Figure 15 It can be seen that the goodness of fit of all test set samples is 0.5960, and the relative error remains between 0-0.03, indicating that the prediction effect is relatively good. Figure 16 It can be seen that the neural network has a good fit for extremely large values, which can greatly reflect the trend of the index. For the gaps between other values, it measures the gaps more compactly and does not affect the ranking overall. This shows that the weights of each level adjusted by the BP neural network can well measure the environmental safety index.

[0184] The economic security index is calculated as:

[0185] According to the indicator system, in this embodiment, the production-sales ratio, external dependence, energy price and availability, which are represented by energy production / energy consumption, the proportion of fossil energy import value in GDP, gasoline prices and per capita electricity consumption, are used as the number of neurons in the four-layer input layer of the economic security index BP model; the economic security index is the final evaluation result, so it is set as the output neuron; a neural network with one hidden layer can represent any continuous function with arbitrary precision, so the hidden layer is set to one layer. In summary, the neural network structure is as follows: Figure 17 shown.

[0186] The closer the correlation coefficient is to 1, the higher the correlation between input and output. Figure 18It can be seen that the goodness of fit of all test set samples is 0.9454. At the same time, the relative error remains between 0-0.2, and most of them are controlled within 0.1, which shows that the prediction effect is relatively good. Figure 19 It can be seen that, except for the four extreme values ​​where the predicted values ​​are slightly lower than the true values, the predicted values ​​of the economic security index are basically consistent with the true values, and the values ​​are highly fitted, which shows that the weights of each level adjusted by the BP neural network can measure the economic security index very well.

[0187] S4. Improve the maximum entropy model and assess energy security risks, specifically:

[0188] In this embodiment, based on the maximum entropy theory, π is defined as the number of n ) and satisfies x i (i=1, 2, ..., n) are independent of each other.

[0189] Therefore, every x has a non-negative probability π(x) such that:

[0190] Assumptions is an estimate of π, then The entropy of

[0191]

[0192] Where ln is the natural logarithm, It can be seen that it is non-negative and its maximum value is lnn.

[0193] Preferably, based on the maximum entropy model, a maximum entropy model suitable for energy security risk assessment is constructed.

[0194] There are n energy security samples with m evaluation indicators forming a reference series:

[0195] x j ={x j (i)|i=1,…,m; j=1,…,n}

[0196] Suppose there are m evaluation indicators and t energy security evaluation standards forming a comparative series:

[0197] x h ={x h (i)|h=1,…,t;i=1,…,m}

[0198] Δ h (i)=|x j (i)-x h (i)|

[0199] The difference between two series is expressed by the grey correlation coefficient ξ i (x j , x h ) is expressed as:

[0200]

[0201] In the formula, hi min Δ h (i) represents the minimum difference between the two levels, hi max Δ h (i) represents the maximum difference between the two levels, ρ represents the resolution coefficient, Δ h (i) indicates the difference between two levels;

[0202] Calculate the correlation degree by combining the weights of each indicator:

[0203] Where Z i Indicates the weight of each indicator;

[0204] Due to the statistical volatility of monitoring values, the evaluation of energy security is fuzzy, so the uncertainty parameter is introduced. j It can be expressed as:

[0205] Where u hj It represents the probability that the jth sample belongs to the hth level of energy security, which is used to evaluate the energy security level. It minimizes the sum of the generalized distances between the sample and the evaluation criteria at each level and maximizes the information entropy as follows:

[0206]

[0207]

[0208] Construct the Lagrangian function:

[0209]

[0210] By calculating the uncertainty H when the information entropy is maximum j Derivative, solve the final evaluation result is:

[0211]

[0212] Where w i It represents the generalized distance between the ith indicator and the evaluation criteria at all levels. B is a constant, which represents the number of information entropies.

[0213] S5. Based on the multi-dimensional energy security indicator system, determine the grading standards for each evaluation indicator, divide the corresponding grade boundaries, and divide the risk levels;

[0214] Among them, the corresponding level boundaries are divided as follows: based on the mean of each subsystem, calculate P 75 、P 50 、P 25 The quantile value serves as a reference benchmark for energy security severity grading;

[0215] The P 75 75% of energy security incidents in the region have a severity higher than this value, while 25% of energy security incidents have a severity lower than this value, indicating that the risk level is at a high level;

[0216] The P 50 50% of energy security incidents in the region are more severe than this value, and the other 50% are less severe than this value, indicating that the risk level is at a medium level;

[0217] The P 25 This means that 25% of energy security incidents in the region have a severity higher than this value, and another 75% of energy security incidents have a severity lower than this value, indicating that the risk level is at a low level.

[0218] In this embodiment, based on the relevant standards of the country, industry and local areas or methods, the methods of determining standards in relevant literature are cited, and the characteristics of the samples are taken into consideration to determine the grading standards of various evaluation indicators. 75 、P 50 、P 25 Percentile values ​​are used as reference values ​​for energy security severity classification. 75 It means that the severity of 75% of energy security incidents in a country or province will be higher than this value, while the severity of 25% of energy security incidents will be lower than this value, which represents a lower level of energy security severity; P 50 This means that 50% of energy security incidents in a country or province will be less severe than this value, while the other 50% will be more severe than this value, thus representing a medium level of energy security. See Table 4-2 for a detailed classification.

[0219] Table 4-1 Level limits of energy security assessment indicators

[0220]

[0221] Table 4-2 Severity Reference Levels

[0222]

[0223] This example simulated 53 observation points according to the severity reference level classification (see Appendix 4-1 for details) and evaluated the energy security risks of 11 samples based on the improved maximum entropy model. The results are shown in Table 4-3.

[0224] Table 4-3 Energy security risk assessment results

[0225]

[0226]

[0227] S6. Evaluate the risk level of the area to be tested based on the improved maximum entropy model.

[0228] In this example, the index evaluation process is influenced by the constituent indicators. While there is theoretically an optimal configuration for the constituent indicators, achieving this in practice is often difficult. Therefore, the weak indicators in the index formation are considered as impediments. Based on this approach, identifying impediments and addressing them in a targeted manner will help improve the index.

[0229] Specifically, step S6 further includes:

[0230] Construct an obstacle factor model to identify and modify the obstacle factors of various evaluation indicators;

[0231] The construction of the obstacle factor model is specifically as follows:

[0232] The obstacle degree O is used to measure the influence of the i-th indicator in the j-th sample on the index. ij Expressed as:

[0233]

[0234] Where m is the number of index evaluation indicators; indicator weight ω i Indicates factor contribution F i Measures the contribution of the i-th indicator to the index; I ij Indicates the difference between the actual value of the i-th indicator in the j-th sample and the optimal value, and the indicator deviation I ij Expressed as (1-r ij ), then r ij is the value after index standardization; set the factor contribution F i , Index Deviation I ij , Obstacle Degree O uj Three basic variables;

[0235] The obstacle degree O of the jth sample uth system uj for:

[0236] Where u iis the number of indicators of the u-th system.

[0237] Specifically, the calculated indicator barrier represents the degree of obstruction posed by the measured indicator to energy security development, while the calculated subsystem barrier represents the degree of obstruction posed by the measured subsystem to energy security development. The calculation formula shows that the subsystem barrier is calculated by summing the barrier values ​​of the indicators within that subsystem. Furthermore, a larger barrier factor indicates a greater degree of obstruction to energy security development. Targeted corrections to subsystems or indicators with high barrier values ​​can effectively improve energy security.

[0238] In this embodiment, to further examine the specific impact of the four subsystems of "supply security, use safety, environmental safety, and economic security" and the indicators under each system on the energy security index of Fujian Province, this embodiment will conduct an obstacle degree diagnostic analysis based on the obstacle degree factor model and analyze the obstacle degree of each subsystem year by year.

[0239] Table 5 Obstacle degree of energy security subsystem in Fujian Province (%)

[0240]

[0241] Table 5 shows that each subsystem has a different impact on Fujian Province's energy transition. In terms of barrier intensity, the economic security system has an average barrier intensity of approximately 26%, significantly higher than the other three subsystems, which are only around 20%. Energy affordability, or price, is a comprehensive reflection of supply and demand. With the advent of the global energy crisis, prices of all energy sources have risen, exacerbating global energy supply and demand. For example, the surge in coal prices in China in 2020 exacerbated power rationing. Therefore, Fujian Province needs to focus on the side effects of energy price fluctuations and actively participate in the development of a unified national market, leveraging market-based measures to mitigate energy risks. Looking at the year-on-year changes in each subsystem, only the energy use system shows an upward trend in barrier intensity, while the energy supply security, economic security, and environmental security subsystems show a downward trend. Generally speaking, economic security reflects energy security through the perspective of energy availability. The reduction in barrier intensity in the economic security system indicates that Fujian Province's energy security is continuously improving, with significant room for improvement. The decline in the environmental safety system's barrier level indicates significant progress in Fujian Province's energy conservation and emission reduction efforts. Although the barrier level for this subsystem remains high, the downward trend is clear. The decline in the energy production safety system's barrier level indicates that Fujian Province's energy structure is gradually upgrading, energy utilization efficiency is continuously improving, and energy production methods are steadily optimizing. The energy use system, the only subsystem to experience an increase in barrier level, saw a relatively small increase. However, this also suggests that Fujian Province's energy consumption structure has not been optimized in tandem with improvements in other subsystems. This will be an area of ​​focus for Fujian Province in its future energy security efforts. Based on the above analysis, Fujian Province's energy security level is gradually improving, and significant progress has been made in pollution control and emission reduction.

[0242] Table 6 Obstacle levels of energy security indicators in Fujian Province (%)

[0243]

[0244] Table 6 shows that the main obstacles to Fujian Province's energy security have changed dynamically over time, but some influencing factors have persisted. This also demonstrates the continuity, multifaceted nature, and dynamism of the factors influencing energy security. This section focuses on the three indicator layers that have the greatest annual impact on Fujian's energy security. It is important to emphasize that in the measurement of each indicator layer, the proportion of coal represents the energy mix, the energy production diversity index represents energy types, per capita energy consumption represents energy demand, SO2 emissions represent air pollution, CO2 emissions represent carbon emissions, and gasoline prices and per capita electricity consumption represent energy prices and availability, respectively.

[0245] Example 2

[0246] Please refer to Figure 2A system for assessing energy security risk by considering multi-dimensional indicators is provided, for implementing a method for assessing energy security risk by considering multi-dimensional indicators of the first embodiment, comprising:

[0247] The data input module is used to transmit the data of each dimension and each indicator and other parameters to the storage module;

[0248] A storage module is used to build a multi-dimensional energy security indicator system based on the four security dimensions of supply security, usage security, environmental security, and economic security;

[0249] An energy security evaluation index calculation module is used to construct an energy security evaluation model of autoencoder-BP neural network-entropy weight TOPSIS, and calculate the energy sub-item evaluation index based on the multi-dimensional energy security indicator system;

[0250] The energy security risk assessment module is used to improve the maximum entropy model and evaluate energy security risks, and evaluate the risk level of the area to be tested based on the improved maximum entropy model.

[0251] In summary, the present invention provides an energy security risk assessment method and system that considers multi-dimensional indicators. By constructing an autoencoder-BP neural network-entropy weight TOPSIS energy security index evaluation model, a multi-dimensional and comprehensive energy security indicator system is formed from the four dimensions of supply security, use safety, environmental safety and economic security, thereby realizing a multi-dimensional and comprehensive analysis and evaluation of energy security.

[0252] It should be noted that for the aforementioned method embodiments, for ease of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0253] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0254] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for energy security risk assessment considering multi-dimensional indicators, characterized in that: The following steps are involved: S1. Construct a multi-dimensional energy security indicator system based on the dimensions of supply security, use security, environmental security, and economic security; S2. Construct an energy security evaluation model based on autoencoder-BP neural network-entropy weight TOPSIS; S3. Based on the multi-dimensional energy security indicator system, calculate the energy sub-item evaluation index respectively; S4. Improve the maximum entropy model and evaluate energy security risks; S5. Based on the multi-dimensional energy security indicator system, determine the grading standards for each evaluation indicator, divide the corresponding grade boundaries, and divide the risk levels; S6. Evaluate the risk level of the area to be tested based on the improved maximum entropy model; Step S2 is specifically as follows: Using a BP neural network model to obtain the weights of the edges in the autoencoder neural network, and using the weights to evaluate the energy security index; The energy security index obtained by entropy-weighted TOPSIS evaluation is used as the input layer to supervise the BP neural network learning and construct the autoencoder-BP neural network-entropy-weighted TOPSIS energy security index evaluation model; Step S3 is specifically as follows: The indexes of the four security dimensions of supply security, consumption security, environmental security and economic security are used as the number of neurons in the four-layer input layer of the BP neural network model, and the energy security index is set as the output neuron, and the energy sub-item evaluation index is calculated respectively; Step S4 is specifically as follows: With m evaluation indicators n Energy security samples constitute the reference series: n} With m evaluation indicators t Energy security evaluation standards form a comparative series: The difference between two series is expressed as grey correlation coefficient Expressed as: ， Where, represents the minimum difference between the two levels, Indicates the maximum difference between the two levels, ρ represents the resolution coefficient, Indicates the difference between two levels; Calculate the correlation degree by combining the weights of each indicator: ； Where, Z i Indicates the weight of each indicator; uncertainty H j It can be expressed as: ， Where, u hj Indicates the j The samples belong to h The probability of level 1 energy security is expressed as follows: ； Construct the Lagrangian function: ； By calculating the uncertainty when the information entropy is maximum H j Derivative, solve the final evaluation result is: ； Where, w i It represents the generalized distance between the ith indicator and the evaluation criteria at all levels. B is a constant, which represents the number of information entropies.

2. The energy security risk assessment method considering multi-dimensional indicators according to claim 1 is characterized in that: Step S1 is specifically as follows: S101. Energy structure indicators in the supply security dimension include energy composition, energy types, and energy uses; The energy composition is measured by fossil energy consumption, with the proportion of coal as a specific indicator; The energy types are measured by the diversity index, and the energy production diversity index SWI For specific indicators: , where t represents the amount of different energy types, P i Indicates the i the share of this energy in the total; The energy use is measured by the power generation structure, with the proportion of thermal power generation as a specific indicator; S102. Energy structure indicators using the security dimension include energy demand, consumption structure, and consumption sensitivity; The energy demand is measured by energy usage, with per capita energy usage as a specific indicator; The consumption structure is measured by structural stability and structural differences. Structural stability is measured by the energy consumption diversity index. Structural differences are measured by the energy use carbon emission coefficient. CE For specific indicators: ， In the formula, t represents the amount of different energy types, P i Indicates the i The share of energy use in the total Coef i Indicates the i Carbon emission coefficient of the energy source; The consumption sensitivity is measured by elasticity, and the specific indicator is energy consumption elasticity index = energy growth rate / economic growth rate; S103. Energy structure indicators in the environmental safety dimension include carbon emission factors, air pollution factors, and greenhouse gas impact factors; The specific indicator of the carbon emission factor is CO2 emissions, the specific indicator of the atmospheric pollution factor is SO2 emissions; the specific indicator of the greenhouse effect factor is CO2 emissions / total energy consumption; S104. Energy structure indicators in the economic security dimension include supply economy and energy inclusiveness; The supply economy is measured by the production-sales ratio and external dependence. The specific indicator of the production-sales ratio is energy production / energy consumption. The specific indicator of external dependence is the proportion of fossil energy imports to GDP. The energy accessibility is measured by energy price and availability. The specific indicator of energy price is gasoline price, and the specific indicator of availability is per capita household electricity consumption.

3. The energy security risk assessment method considering multi-dimensional indicators according to claim 1 is characterized in that: The BP neural network model in step S2 is specifically: Determine the number of nodes and connection weights of each layer of the BP neural network, initialize the hidden layer threshold and output layer threshold, and set the learning rate and neuron activation function; Calculating a hidden layer output using the input vector, the connection weights, and the hidden layer threshold; Calculating the output layer output using the connection weights, the output layer threshold, and the hidden layer output; The output layer output and the expected output are used to perform error calculation, weight update and threshold update in sequence, and it is determined whether the iteration is completed. If so, the calculation is completed; otherwise, the step of calculating the hidden layer output is returned to be executed.

4. The energy security risk assessment method considering multi-dimensional indicators according to claim 1 is characterized in that: The corresponding level boundaries described in step S5 are specifically: Based on the average value of each subsystem, calculate P 75 、 P 50 、 P 25 The quantile value serves as a reference benchmark for energy security severity grading; described P 75 75% of energy security incidents in the region have a severity higher than this value, while 25% of energy security incidents have a severity lower than this value, indicating that the risk level is at a high level; described P 50 50% of energy security incidents in the region are more severe than this value, and the other 50% are less severe than this value, indicating that the risk level is at a medium level; described P 25 This means that 25% of energy security incidents in the region have a severity higher than this value, and the other 75% of energy security incidents have a severity lower than this value, indicating that the risk level is at a low level.

5. The energy security risk assessment method considering multi-dimensional indicators according to claim 1 is characterized in that: Step S6 further includes: Construct an obstacle factor model to identify and modify the obstacle factors of various evaluation indicators; The barrier factor model is constructed as follows: Measure the j In the sample i The degree of influence of each indicator on the index O ij Expressed as: ， Where, m is the number of index evaluation indicators; indicator weight ω i Indicates factor contribution F i Measure the i The degree of contribution of each indicator to the index; I ij Indicates the j In the sample i The difference between the actual value of an indicator and the optimal value is the indicator deviation. I ij Expressed as (1- r ij ),but r ij The value after the indicator is standardized; set the factor contribution F i , indicator deviation I ij , Obstacle O uj Three basic variables; No. j Sample No. u The barrier of the system O uj for: ; Where, u i For the u The number of indicators per system.

6. An energy security risk assessment system considering multi-dimensional indicators, used to implement the energy security risk assessment method considering multi-dimensional indicators according to any one of claims 1 to 5, characterized in that: include: The data input module is used to transmit the data of each dimension and each indicator and other parameters to the storage module; A storage module is used to build a multi-dimensional energy security indicator system based on the four security dimensions of supply security, usage security, environmental security, and economic security; An energy security evaluation index calculation module is used to construct an energy security evaluation model of autoencoder-BP neural network-entropy weight TOPSIS, and calculate the energy sub-item evaluation index based on the multi-dimensional energy security indicator system; The energy security risk assessment module is used to improve the maximum entropy model and evaluate energy security risks, and evaluate the risk level of the area to be tested based on the improved maximum entropy model.

Images