Grounding grid safety evaluation system and method based on fuzzy evaluation and machine learning

Abstract

This invention discloses a grounding grid safety assessment system based on fuzzy evaluation and machine learning. The system fuzzifies all key assessment indicators of the grounding grid to obtain the membership degree of each key indicator at each safety level, constructing a fuzzy relation matrix. Machine learning algorithms are used to analyze and predict the machine learning weights of each key assessment indicator. Based on dynamic fusion coefficients, the predefined expert weights and machine learning weights of each key assessment indicator are dynamically fused to obtain a comprehensive weight vector. Fuzzy synthesis operations are performed using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid. The membership degree of each safety level in the grounding grid is then defuzzified to obtain the grounding grid safety assessment score. This invention improves the accuracy and reliability of grounding grid safety assessment results.

Description

Technical Field

[0001] This invention relates to the field of power system safety monitoring and intelligent assessment technology, specifically to a grounding grid safety assessment system and method based on fuzzy evaluation and machine learning. Background Technology

[0002] The safe operation of the grounding grid in a power system is directly related to the safety of power equipment and personnel, as well as the stability of the power system. Traditional grounding grid safety assessment methods often rely solely on expert experience to set weights, which is highly subjective, or rely solely on machine learning algorithms without the guidance of human experience. Furthermore, the weight settings are fixed and cannot be dynamically adjusted according to changes in the actual operating data of the grounding grid. At the same time, the assessment process does not adequately handle the fuzziness of the indicators, resulting in poor accuracy, adaptability, and dynamism of the assessment results, making it difficult to accurately reflect the actual safety status of the grounding grid.

[0003] Fuzzy comprehensive evaluation, as an effective method for processing fuzzy information, can effectively handle the uncertainties and fuzzinesses in grounding grid condition assessment and has been widely used in qualitative analysis of grounding grid conditions. However, traditional fuzzy comprehensive evaluation methods rely heavily on expert experience for weight determination, lacking a data-driven mechanism and making it difficult to dynamically adjust weights over time and with environmental changes. On the other hand, machine learning algorithms, especially advanced algorithms such as random forests and deep neural networks, can automatically learn the complex mapping relationship between indicators and risk levels using historical data. However, these methods often lack physical interpretability and struggle to incorporate the prior knowledge of domain experts. Summary of the Invention

[0004] The purpose of this invention is to provide a grounding grid safety assessment system and method based on fuzzy evaluation and machine learning. This invention realizes intelligent, dynamic and adaptive optimization of grounding grid safety assessment, and improves the accuracy and reliability of assessment results.

[0005] To achieve this objective, the present invention provides a grounding grid safety assessment system based on fuzzy evaluation and machine learning, comprising: The fuzzy relation matrix construction module is used to fuzzify all key evaluation indicators of the grounding grid, obtain the membership degree of each key evaluation indicator at each safety level, and construct a fuzzy relation matrix based on the membership degree of each key evaluation indicator at each safety level. The machine learning weight prediction module is used to learn, analyze and predict the machine learning weights of each key evaluation index of the grounding grid through machine learning algorithms. The adaptive weight fusion module is used to dynamically fuse the predefined expert weights and machine learning weights of each key evaluation indicator based on the dynamic fusion coefficient, so as to obtain the comprehensive weight of each key evaluation indicator and form a comprehensive weight vector. The fuzzy comprehensive evaluation module is used to perform fuzzy synthesis operation using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid, and to defuzzify the membership degree of each safety level in the grounding grid to obtain the grounding grid safety evaluation score.

[0006] Preferably, the learning update module is used to automatically update the model parameters in the machine learning algorithm and the dynamic fusion coefficients of the adaptive weight fusion module using a partial fitting method after all the key evaluation index data of the grounding grid collected have reached the preset accumulation conditions.

[0007] Preferably, in the fuzzy relation matrix construction module, the step of fuzzifying all key evaluation indicators of the grounding grid to obtain the membership degree of each key evaluation indicator at each safety level, and constructing a fuzzy relation matrix based on the membership degree of each key evaluation indicator at each safety level, is specifically used for: For each key evaluation index of the grounding grid, the membership degree of each key evaluation index at each safety level is calculated using a trapezoidal membership function, expressed as follows: in, x This represents the standardized value of each key evaluation indicator; a j Indicates the first j The first function parameter corresponding to each security level b j Indicates the first j The second function parameter corresponding to each security level c j Indicates the first j The third function parameter corresponding to each security level d j Indicates the first j The fourth function parameter corresponding to each security level u ij Indicates the first i The key evaluation indicators in the first j Membership degree at each security level; The membership degrees of all key assessment indicators at each security level are combined into a fuzzy relation matrix R.

[0008] Preferably, in the machine learning weight prediction module, the step of obtaining the machine learning weights of each key evaluation index by learning, analyzing, and predicting each key evaluation index of the grounding grid using machine learning algorithms is specifically used for: The machine learning algorithm includes random forest and deep neural network, and uses a dual model of random forest and deep neural network in parallel to predict the machine learning weights of each key evaluation index: The random forest model calculates the importance score of each key evaluation metric based on feature importance, and then normalizes it to obtain the weights: in, Indicates the first i Importance scores of key evaluation indicators Indicates the first k Importance scores of key evaluation indicators This represents the summation of the importance scores for 1 to n key evaluation indicators. The first one obtained by the random forest model i Predicted weights of key evaluation indicators; The deep neural network model employs a multilayer perceptron structure. The input layer receives all key evaluation metric data, performs nonlinear transformations through hidden layers, and the output layer uses the softmax function to normalize the weights. in, Represents the deep neural network's first... i The exponential score of each output node. The first term obtained from the deep neural network model i Predicted weights of key evaluation indicators; The first obtained from the random forest model i Predicted weights of key evaluation indicators The first and deep neural network models obtained i Predicted weights of key evaluation indicators The weighted fusion yields the first... i Machine learning weights for key evaluation metrics: Where β is the machine learning fusion coefficient. For the first i Machine learning weights for key evaluation metrics.

[0009] Preferably, in the adaptive weight fusion module, the predefined expert weights and machine learning weights of each key evaluation indicator are dynamically fused based on dynamic fusion coefficients to obtain a comprehensive weight for each key evaluation indicator, forming a comprehensive weight vector, specifically used for: By inputting the expert weights of each key evaluation indicator predefined by the grounding grid evaluation experts, an adaptive mechanism is used to dynamically fuse the predefined expert weights and the machine learning weights of each key evaluation indicator. The fusion formula is as follows: in, Represents the predefined first i Expert weights for key evaluation indicators, For the first i The machine learning weights for the key evaluation metrics, where α is the dynamic fusion coefficient. For the first i The combined weight of each key evaluation indicator; Based on the mean absolute error within the sliding window, α is dynamically updated. The formula for calculating the mean absolute error is: in, for t Predicted value of grounding grid safety assessment at any given time. for t The actual value of the grounding grid safety assessment at any given time. Where is the size of the time window, and MAE is the mean absolute error; Calculate the mean sliding window error based on the mean absolute error (MAE): in, W To adjust the sliding window size, This represents the mean error of the sliding window. Based on the average error of the sliding window Update the dynamic fusion coefficient α: in, This is the adjustment factor for the window step size. The average prediction error threshold, This sets the minimum value for the dynamic fusion coefficient α. This sets the maximum value for the dynamic fusion coefficient α, and clip represents the truncation function. Indicates the first t The dynamic fusion coefficient value for each evaluation period. Indicates the first t+1 The dynamic fusion coefficient value for each evaluation period, i.e., the updated dynamic fusion coefficient; Based on the combined weight of all key evaluation indicators This forms a comprehensive weight vector. .

[0010] Preferably, in the fuzzy comprehensive evaluation module, the step of performing fuzzy synthesis operation using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid, and then defuzzifying the membership degree of each safety level in the grounding grid to obtain the grounding grid safety evaluation score, is specifically used for: Using the fuzzy relation matrix R and the comprehensive weight vector Fuzzy synthesis operations are performed, and a weighted average synthesis operator is used to calculate the membership degree of each safety level in the grounding grid, resulting in the membership degree vector of each safety level. B : in, Let R be the comprehensive weight vector, and let R be the fuzzy relation matrix. b j For the first j Membership degree of each security level, b m For the first m Membership degree of each security level, B This represents the membership vector for each security level. No. j Membership of each security level b j The calculation formula is: in, For the first i The combined weight of the key evaluation indicators, u ij Indicates the first i The first key evaluation indicator corresponds to the first j Membership degree of each security level; Membership vectors of each security level are determined using the centroid method. B Defuzzification is performed to obtain the grounding grid safety assessment score. S The calculation formula is: in, y j For the first j The score for each security level.

[0011] This invention also provides a grounding grid safety assessment method based on fuzzy evaluation and machine learning, comprising: All key evaluation indicators of the grounding grid are fuzzified to obtain the membership degree of each key evaluation indicator at each safety level. A fuzzy relation matrix is ​​constructed based on the membership degree of each key evaluation indicator at each safety level. The machine learning weights of each key evaluation index of the grounding grid are obtained by learning, analyzing and predicting them through machine learning algorithms. Based on the dynamic fusion coefficient, the expert weights of each predefined key evaluation indicator and the machine learning weights of each key evaluation indicator are dynamically fused to obtain the comprehensive weight of each key evaluation indicator, forming a comprehensive weight vector. Fuzzy synthesis is performed using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid. The membership degree of each safety level in the grounding grid is then defuzzified to obtain the grounding grid safety assessment score.

[0012] The present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, implements the steps of a grounding grid safety assessment method based on fuzzy evaluation and machine learning as described above.

[0013] The beneficial effects of this invention are: This invention integrates expert experience with data-driven methods, preserving the reliability of domain knowledge while fully utilizing the data adaptability of machine learning, overcoming the limitations of fixed weights in traditional methods. Through an adaptive fusion coefficient adjustment mechanism, this invention dynamically optimizes the ratio of expert weights to machine learning weights in real time, improving prediction accuracy. This invention employs a dual-model approach of random forest and deep neural network for parallel weight prediction, combining the advantages of feature importance analysis in random forests with the nonlinear learning capabilities of neural networks, further enhancing the accuracy and stability of weight prediction through weighted fusion. This invention significantly improves the accuracy, reliability, and adaptability of grounding grid safety assessment results, providing scientific support for the safe operation of substation grounding grids.

[0014] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the structure of the present invention; Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0016] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0017] Example 1 A grounding grid safety assessment system based on fuzzy evaluation and machine learning, such as Figure 1 As shown, it includes: The fuzzy relation matrix construction module is used to fuzzify all key evaluation indicators of the grounding grid, and obtain the membership degree of each key evaluation indicator (including the corrosion degree of the grounding grid down conductor, grounding resistance, contact voltage, step voltage, grounding grid corrosion, burial depth and discontinuity) at each safety level. Based on the membership degree of each key evaluation indicator at each safety level, a fuzzy relation matrix is ​​constructed. This design objectively reflects the degree of belonging of each key evaluation indicator to each safety level by calculating the membership degree of each key evaluation indicator at each safety level, and lays an objective data foundation for subsequent evaluation. The machine learning weight prediction module is used to learn, analyze and predict the machine learning weights of each key evaluation index of the grounding grid through machine learning algorithms. This design uses machine learning algorithms to learn and analyze the data of each key evaluation index of the grounding grid, and uses the algorithm to explore the patterns behind the key evaluation index data. Based on the data characteristics, the machine learning weights of each key evaluation index are obtained, making the weight allocation more in line with the actual operation of the grounding grid and getting rid of the subjective problem of relying solely on expert experience to set the weights. The adaptive weight fusion module is used to dynamically fuse the predefined expert weights and machine learning weights of each key evaluation indicator based on a dynamic fusion coefficient, thereby obtaining a comprehensive weight of each key evaluation indicator and forming a comprehensive weight vector. This design, based on a dynamic fusion coefficient, retains the professional experience of power field experts in grounding grid evaluation and combines the objective machine learning weights obtained from data by machine learning algorithms. At the same time, the dynamic fusion coefficient allows the fusion ratio to be adjusted, avoiding the limitations of a fixed fusion ratio. The resulting comprehensive weight vector can more scientifically reflect the degree of impact of each key evaluation indicator on the safety of the grounding grid. The fuzzy comprehensive evaluation module is used to perform fuzzy synthesis operations using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid. The membership degree of each safety level in the grounding grid is then defuzzified to obtain the grounding grid safety evaluation score (the grounding grid safety evaluation score is the final safety score). This design uses the fuzzy relation matrix and the comprehensive weight vector to perform fuzzy synthesis operations to obtain the membership degree of each safety level, and then uses defuzzification to transform the fuzzy membership degree information into specific evaluation scores, thereby realizing the quantitative evaluation of the grounding grid safety level and allowing the evaluation results to clearly and intuitively reflect the actual safety status of the grounding grid.

[0018] In the above technical solution, the learning and updating module is used when all the collected key evaluation index data of the grounding grid reach the preset accumulation condition (i.e., the number of collected key evaluation index samples N of the grounding grid). new ≥ Threshold Number of Samples N threshold After a period of time, the model parameters are updated using a partial fitting method. Based on the new key evaluation index sample set, the parameters are recalculated, and the β value that minimizes the evaluation error is selected as the new optimal value. The α value is recalculated as the new dynamic fusion coefficient based on the mean of the sliding window error. After all the key evaluation index data of the grounding grid collected reach the preset accumulation conditions, the above design uses a partial fitting method to automatically update the model parameters of the machine learning algorithm and the dynamic fusion coefficient of the adaptive weight fusion module. The partial fitting method can optimize and adjust the model parameters and fusion coefficients based on the newly accumulated data without retraining the entire model. This allows the model and fusion coefficients to adapt to changes in the grounding grid operation data, always conforming to the actual situation and improving adaptability and evaluation accuracy.

[0019] In the above technical solution, the fuzzification of all key evaluation indicators of the grounding grid to obtain the membership degree of each key evaluation indicator at each safety level, and the construction of a fuzzy relation matrix based on the membership degree of each key evaluation indicator at each safety level, are specifically used for: For each key evaluation index of the grounding grid, the membership degree of each key evaluation index at each safety level is calculated using a trapezoidal membership function, expressed as follows: in, x This represents the standardized value of each key evaluation indicator; a j Indicates the first j The first function parameter corresponding to each security level b j Indicates the first j The second function parameter corresponding to each security level c j Indicates the first j The third function parameter corresponding to each security level d j Indicates the first j The fourth function parameter corresponding to each security level u ij Indicates the first i The key evaluation indicators in the first j Membership degree at each security level; The safety levels include safe, good, caution, and danger. The membership degree of each key assessment indicator at each safety level is calculated using a trapezoidal membership function, which means calculating the membership degree of each key assessment indicator at safe, good, caution, and danger respectively. The membership degrees of all key evaluation indicators at each security level are combined into a fuzzy relation matrix R. The above design calculates the membership degree using a trapezoidal membership function, ensuring that the calculated membership degree values ​​are accurate and reliable, and improving the scientific nature of the constructed fuzzy relation matrix.

[0020] In the above technical solution, the step of using machine learning algorithms to learn, analyze, and predict the key evaluation indicators of the grounding grid to obtain the machine learning weights of each key evaluation indicator is specifically used for: The machine learning algorithm includes random forest and deep neural network, and uses a dual model of random forest and deep neural network in parallel to predict the machine learning weights of each key evaluation index: The random forest model calculates the importance score of each key evaluation metric based on feature importance, and then normalizes it to obtain the weights: in, Indicates the first i Importance scores of key evaluation indicators Indicates the first k Importance scores of key evaluation indicators This represents the summation of the importance scores for 1 to n key evaluation indicators. The first one obtained by the random forest model i Predicted weights of key evaluation indicators; The deep neural network model employs a multilayer perceptron structure. The input layer receives all key evaluation metric data, performs nonlinear transformations through hidden layers, and the output layer uses the softmax function to normalize the weights. in, Represents the deep neural network's first... i The exponential score of each output node. The first term obtained from the deep neural network model i Predicted weights of key evaluation indicators; The first obtained from the random forest model i Predicted weights of key evaluation indicators The first and deep neural network models obtained i Predicted weights of key evaluation indicators The weighted fusion yields the first... i Machine learning weights for key evaluation metrics: Wherein, β is the machine learning fusion coefficient. β is used to balance the weight of the random forest model and the deep neural network model in the final machine learning weight. The fusion weight can be calculated separately using a preset β value set. The safety score is obtained by combining the validation set and fuzzy comprehensive evaluation. Finally, the optimal β value is determined based on the criterion of minimizing the evaluation error. For the first i The above design employs a dual-model approach, combining random forest and deep neural network, to predict machine learning weights for key evaluation metrics. The random forest model accurately mines the importance features of metrics based on feature importance, while the deep neural network model captures the nonlinear characteristics of metric data through a multilayer perceptron structure. The parallel prediction by the two models is followed by weighted fusion to obtain the machine learning weights. This approach overcomes the limitations of a single algorithm and improves the accuracy, stability, and reliability of machine learning weight prediction. By combining the advantages of different machine learning algorithms, the predicted machine learning weights are made to better reflect the actual key evaluation metric characteristics of the grounding grid.

[0021] In the above technical solution, the dynamic fusion coefficient is used to dynamically fuse the predefined expert weights and machine learning weights of each key evaluation indicator to obtain the comprehensive weight of each key evaluation indicator, forming a comprehensive weight vector, which is specifically used for: By inputting the expert weights of each key evaluation indicator predefined by grounding grid evaluation experts (relying on expert experience), an adaptive mechanism is used to dynamically fuse the predefined expert weights and the machine learning weights of each key evaluation indicator. The fusion formula is as follows: in, Represents the predefined first i Expert weights for key evaluation indicators, For the first i The machine learning weights for the key evaluation metrics, where α is the dynamic fusion coefficient (α ranges from 0.3 to 0.9). For the first i The combined weight of each key evaluation indicator; α is dynamically updated based on the mean absolute error (MAE) within the sliding window. The formula for calculating the mean absolute error is: in, for t Predicted value of grounding grid safety assessment at any given time. for t The actual value of the grounding grid safety assessment at any given time. Where is the size of the time window, and MAE is the mean absolute error; Calculate the mean sliding window error based on the mean absolute error (MAE): in, W To adjust the sliding window size, This represents the mean error of the sliding window. Based on the average error of the sliding window Update the dynamic fusion coefficient α: in, The adjustment factor for the window step size ( (Possible value: 0.5) The average prediction error threshold ( (Possible value: 0.05) This sets the minimum value for the dynamic fusion coefficient α. This sets the maximum value for the dynamic fusion coefficient α, and clip represents the truncation function. Indicates the first t The dynamic fusion coefficient value for each evaluation period. Indicates the first t+1 The dynamic fusion coefficient value for each evaluation period, i.e., the updated dynamic fusion coefficient; Based on the combined weight of all key evaluation indicators This forms a comprehensive weight vector. The above design can accurately measure the deviation between the predicted value and the actual value through the mean absolute error. Based on this, the mean value of the sliding window error is calculated and the dynamic fusion coefficient is updated to ensure that the allocation of the comprehensive weight can be adaptively adjusted according to the changes in the actual operation data of the grounding grid, so that the comprehensive weight can always optimally reflect the actual impact of each key evaluation indicator.

[0022] In the above technical solution, the step of performing fuzzy synthesis operation using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid, and then defuzzifying the membership degree of each safety level in the grounding grid to obtain the grounding grid safety assessment score, is specifically used for: Using the fuzzy relation matrix R and the comprehensive weight vector Fuzzy synthesis operations are performed, and a weighted average synthesis operator is used to calculate the membership degree of each safety level in the grounding grid, resulting in the membership degree vector of each safety level. B : in, Let R be the comprehensive weight vector, and let R be the fuzzy relation matrix. b j For the firstj Membership degree of each security level, b m For the first m Membership degree of each security level, B This represents the membership vector for each security level. No. j Membership of each security level b j The calculation formula is: in, For the first i The combined weight of the key evaluation indicators, u ij Indicates the first i The first key evaluation indicator corresponds to the first j Membership degree of each security level; Membership vectors of each security level are determined using the centroid method. B Defuzzification is performed to obtain the grounding grid safety assessment score. S The calculation formula is: in, y j For the first j The design employs a weighted average composite operator for fuzzy composite operations. This effectively integrates comprehensive weights and membership information, accurately calculates the membership of each safety level, and uses the centroid method for defuzzification. The fuzzy membership vector is transformed into a specific, quantified safety assessment score, ensuring that the final grounding grid safety assessment score accurately and intuitively quantifies the safety level of the grounding grid, allowing the assessment results to more objectively reflect the actual safety status of the grounding grid.

[0023] Example 2 like Figure 2 As shown, this embodiment provides a grounding grid safety assessment method based on fuzzy evaluation and machine learning, including the following steps: S1. Fuzzyenize all key evaluation indicators of the grounding grid to obtain the membership degree of each key evaluation indicator at each safety level. Construct a fuzzy relation matrix based on the membership degree of each key evaluation indicator at each safety level, specifically including: For each key evaluation index of the grounding grid, the membership degree of each key evaluation index at each safety level is calculated using a trapezoidal membership function, expressed as follows: in, xThis represents the standardized value of each key evaluation indicator; a j Indicates the first j The first function parameter corresponding to each security level b j Indicates the first j The second function parameter corresponding to each security level c j Indicates the first j The third function parameter corresponding to each security level d j Indicates the first j The fourth function parameter corresponding to each security level u ij Indicates the first i The key evaluation indicators in the first j Membership degree at each security level; The membership degrees of all key assessment indicators at each security level are combined into a fuzzy relation matrix R.

[0024] S2. The machine learning weights of each key evaluation index of the grounding grid are obtained by learning, analyzing, and predicting them using machine learning algorithms. Specifically, these include: The machine learning algorithm includes random forest and deep neural network, and uses a dual model of random forest and deep neural network in parallel to predict the machine learning weights of each key evaluation index: The random forest model calculates the importance score of each key evaluation metric based on feature importance, and then normalizes it to obtain the weights: in, Indicates the first i Importance scores of key evaluation indicators Indicates the first k Importance scores of key evaluation indicators This represents the summation of the importance scores for 1 to n key evaluation indicators. The first one obtained by the random forest model i Predicted weights of key evaluation indicators; The deep neural network model employs a multilayer perceptron structure. The input layer receives all key evaluation metric data, performs nonlinear transformations through hidden layers, and the output layer uses the softmax function to normalize the weights. in, Represents the deep neural network's first... i The exponential score of each output node. The first term obtained from the deep neural network model iPredicted weights of key evaluation indicators; The first obtained from the random forest model i Predicted weights of key evaluation indicators The first and deep neural network models obtained i Predicted weights of key evaluation indicators The weighted fusion yields the first... i Machine learning weights for key evaluation metrics: Where β is the machine learning fusion coefficient. For the first i Machine learning weights for key evaluation metrics.

[0025] S3. Based on the dynamic fusion coefficient, the predefined expert weights and machine learning weights of each key evaluation indicator are dynamically fused to obtain the comprehensive weight of each key evaluation indicator, forming a comprehensive weight vector, specifically including: Using the predefined expert weights of each key evaluation indicator based on the analytic hierarchy process (AHP), an adaptive mechanism is employed to dynamically fuse the predefined expert weights and the machine learning weights of each key evaluation indicator. The fusion formula is as follows: in, Represents the predefined first i Expert weights for key evaluation indicators, For the first i The machine learning weights for the key evaluation metrics, where α is the dynamic fusion coefficient. For the first i The combined weight of each key evaluation indicator; Based on the mean absolute error within the sliding window, α is dynamically updated. The formula for calculating the mean absolute error is: in, for t Predicted value of grounding grid safety assessment at any given time. for t The actual value of the grounding grid safety assessment at any given time. Where is the size of the time window, and MAE is the mean absolute error; Calculate the mean sliding window error based on the mean absolute error (MAE): in, W To adjust the sliding window size, This represents the mean error of the sliding window. Based on the average error of the sliding window Update the dynamic fusion coefficient α: in, This is the adjustment factor for the window step size. The average prediction error threshold, This sets the minimum value for the dynamic fusion coefficient α. This sets the maximum value for the dynamic fusion coefficient α, and clip represents the truncation function. Indicates the first t The dynamic fusion coefficient value for each evaluation period. Indicates the first t+1 The dynamic fusion coefficient value for each evaluation period, i.e., the updated dynamic fusion coefficient; Based on the combined weight of all key evaluation indicators This forms a comprehensive weight vector. .

[0026] S4. Perform fuzzy synthesis operation using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid. Defuzzify the membership degree of each safety level in the grounding grid to obtain the grounding grid safety assessment score, specifically including: Using the fuzzy relation matrix R and the comprehensive weight vector Fuzzy synthesis operations are performed, and a weighted average synthesis operator is used to calculate the membership degree of each safety level in the grounding grid, resulting in the membership degree vector of each safety level. B : in, Let R be the comprehensive weight vector, and let R be the fuzzy relation matrix. b j For the first j Membership degree of each security level, b m For the first m Membership degree of each security level, B This represents the membership vector for each security level. No. j Membership of each security level b j The calculation formula is: in, For the first i The combined weight of the key evaluation indicators, u ij Indicates the first i The first key evaluation indicator corresponds to the first j Membership degree of each security level; Membership vectors of each security level are determined using the centroid method. B Defuzzification is performed to obtain the grounding grid safety assessment score. S The calculation formula is: in, y j For the first j The score for each security level.

[0027] Example 3 A computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in Embodiment 2.

[0028] Example 4 This embodiment provides a computer storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in Embodiment 2.

[0029] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0030] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0031] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0032] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0033] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading the present invention, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

[0034] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. A ground grid safety assessment system based on fuzzy evaluation and machine learning, characterized by, It includes: The fuzzy relation matrix construction module is used to fuzzify all key evaluation indicators of the grounding grid, obtain the membership degree of each key evaluation indicator at each safety level, and construct a fuzzy relation matrix based on the membership degree of each key evaluation indicator at each safety level. The machine learning weight prediction module is used to learn, analyze and predict the machine learning weights of each key evaluation index of the grounding grid through machine learning algorithms. The adaptive weight fusion module is used to dynamically fuse the predefined expert weights and machine learning weights of each key evaluation indicator based on the dynamic fusion coefficient, so as to obtain the comprehensive weight of each key evaluation indicator and form a comprehensive weight vector. The fuzzy comprehensive evaluation module is used to perform fuzzy synthesis operation using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid, and to defuzzify the membership degree of each safety level in the grounding grid to obtain the grounding grid safety evaluation score.

2. The grounding grid safety assessment system based on fuzzy evaluation and machine learning according to claim 1, characterized in that, It also includes: The learning update module is used to automatically update the model parameters in the machine learning algorithm and the dynamic fusion coefficients of the adaptive weight fusion module using a partial fitting method after all the key evaluation index data of the grounding grid collected have reached the preset accumulation conditions.

3. The grounding grid safety assessment system based on fuzzy evaluation and machine learning according to claim 1, characterized in that, In the fuzzy relation matrix construction module, the fuzzification of all key evaluation indicators of the grounding grid is performed to obtain the membership degree of each key evaluation indicator at each safety level. A fuzzy relation matrix is ​​then constructed based on the membership degrees of each key evaluation indicator at each safety level. Specifically, this is used for: For each key evaluation index of the grounding grid, the membership degree of each key evaluation index at each safety level is calculated using a trapezoidal membership function, expressed as follows: in, x This represents the standardized value of each key evaluation indicator; a j Indicates the first j The first function parameter corresponding to each security level b j Indicates the first j The second function parameter corresponding to each security level c j Indicates the first j The third function parameter corresponding to each security level d j Indicates the first j The fourth function parameter corresponding to each security level u ij Indicates the first i The key evaluation indicators in the first j Membership degree at each security level; The membership degrees of all key assessment indicators at each security level are combined into a fuzzy relation matrix R.

4. The grounding grid safety assessment system based on fuzzy evaluation and machine learning according to claim 1, characterized in that, In the machine learning weight prediction module, the process of learning, analyzing, and predicting the key evaluation indicators of the grounding grid using machine learning algorithms to obtain the machine learning weights of each key evaluation indicator is specifically used for: The machine learning algorithm includes random forest and deep neural network, and uses a dual model of random forest and deep neural network in parallel to predict the machine learning weights of each key evaluation index: The random forest model calculates the importance score of each key evaluation metric based on feature importance, and then normalizes it to obtain the weights: in, Indicates the first i Importance scores of key evaluation indicators Indicates the first k Importance scores of key evaluation indicators This represents the summation of the importance scores for 1 to n key evaluation indicators. The first one obtained by the random forest model i Predicted weights of key evaluation indicators; The deep neural network model employs a multilayer perceptron structure. The input layer receives all key evaluation metric data, performs nonlinear transformations through hidden layers, and the output layer uses the softmax function to normalize the weights. in, Represents the deep neural network's first... i The exponential score of each output node. The first term obtained from the deep neural network model i Predicted weights of key evaluation indicators; The first obtained from the random forest model i Predicted weights of key evaluation indicators The first and deep neural network models obtained i Predicted weights of key evaluation indicators The weighted fusion yields the first... i Machine learning weights for key evaluation metrics: Where β is the machine learning fusion coefficient. For the first i Machine learning weights for key evaluation metrics.

5. The grounding grid safety assessment system based on fuzzy evaluation and machine learning according to claim 1 or 2, characterized in that, In the adaptive weight fusion module, based on dynamic fusion coefficients, the predefined expert weights and machine learning weights of each key evaluation indicator are dynamically fused to obtain the comprehensive weight of each key evaluation indicator, forming a comprehensive weight vector, which is specifically used for: By inputting the expert weights of each key evaluation indicator predefined by the grounding grid evaluation experts, an adaptive mechanism is used to dynamically fuse the predefined expert weights and the machine learning weights of each key evaluation indicator. The fusion formula is as follows: in, Represents the predefined first i Expert weights for key evaluation indicators, For the first i The machine learning weights for the key evaluation metrics, where α is the dynamic fusion coefficient. For the first i The combined weight of each key evaluation indicator; Based on the mean absolute error within the sliding window, α is dynamically updated. The formula for calculating the mean absolute error is: in, for t Predicted value of grounding grid safety assessment at any given time. for t The actual value of the grounding grid safety assessment at any given time. Where is the size of the time window, and MAE is the mean absolute error; Calculate the mean sliding window error based on the mean absolute error (MAE): in, W To adjust the sliding window size, This represents the mean error of the sliding window. Based on the average error of the sliding window Update the dynamic fusion coefficient α: in, This is the adjustment factor for the window step size. The average prediction error threshold, This sets the minimum value for the dynamic fusion coefficient α. This sets the maximum value for the dynamic fusion coefficient α, and clip represents the truncation function. Indicates the first t The dynamic fusion coefficient value for each evaluation period. Indicates the first t+1 The dynamic fusion coefficient value for each evaluation period, i.e., the updated dynamic fusion coefficient; Based on the combined weight of all key evaluation indicators This forms a comprehensive weight vector. .

6. The grounding grid safety assessment system based on fuzzy evaluation and machine learning according to claim 3 or 5, characterized in that, In the fuzzy comprehensive evaluation module, the fuzzy synthesis operation is performed using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid. The membership degree of each safety level in the grounding grid is then defuzzified to obtain the grounding grid safety evaluation score. Specifically, this is used for: Using the fuzzy relation matrix R and the comprehensive weight vector Fuzzy synthesis operations are performed, and a weighted average synthesis operator is used to calculate the membership degree of each safety level in the grounding grid, resulting in the membership degree vector of each safety level. B : in, Let R be the comprehensive weight vector, and let R be the fuzzy relation matrix. b j For the first j Membership degree of each security level, b m For the first m Membership degree of each security level, B This represents the membership vector for each security level. No. j Membership of each security level b j The calculation formula is: in, For the first i The combined weight of the key evaluation indicators, u ij Indicates the first i The first key evaluation indicator corresponds to the first j Membership degree of each security level; Membership vectors of each security level are determined using the centroid method. B Defuzzification is performed to obtain the grounding grid safety assessment score. S The calculation formula is: in, y j For the first j The score for each security level.

7. A grounding grid safety assessment method based on fuzzy evaluation and machine learning, characterized in that, Includes the following steps: All key evaluation indicators of the grounding grid are fuzzified to obtain the membership degree of each key evaluation indicator at each safety level. A fuzzy relation matrix is ​​constructed based on the membership degree of each key evaluation indicator at each safety level. The machine learning weights of each key evaluation index of the grounding grid are obtained by learning, analyzing and predicting them through machine learning algorithms. Based on the dynamic fusion coefficient, the expert weights of each predefined key evaluation indicator and the machine learning weights of each key evaluation indicator are dynamically fused to obtain the comprehensive weight of each key evaluation indicator, forming a comprehensive weight vector. Fuzzy synthesis is performed using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid. The membership degree of each safety level in the grounding grid is then defuzzified to obtain the grounding grid safety assessment score.

8. The grounding grid safety assessment method based on fuzzy evaluation and machine learning according to claim 7, characterized in that, The process involves fuzzifying all key evaluation indicators of the grounding grid to obtain the membership degree of each key evaluation indicator at each safety level. A fuzzy relation matrix is ​​then constructed based on the membership degrees of each key evaluation indicator at each safety level, including: For each key evaluation index of the grounding grid, the membership degree of each key evaluation index at each safety level is calculated using a trapezoidal membership function, expressed as follows: in, x This represents the standardized value of each key evaluation indicator; a j Indicates the first j The first function parameter corresponding to each security level b j Indicates the first j The second function parameter corresponding to each security level c j Indicates the first j The third function parameter corresponding to each security level d j Indicates the first j The fourth function parameter corresponding to each security level u ij Indicates the first i The key evaluation indicators in the first j Membership degree at each security level; The membership degrees of all key assessment indicators at each security level are combined into a fuzzy relation matrix R.

9. The grounding grid safety assessment method based on fuzzy evaluation and machine learning according to claim 7, characterized in that, The process of learning, analyzing, and predicting the key evaluation indicators of the grounding grid using machine learning algorithms to obtain the machine learning weights for each key evaluation indicator includes: The machine learning algorithm includes random forest and deep neural network, and uses a dual model of random forest and deep neural network in parallel to predict the machine learning weights of each key evaluation index: The random forest model calculates the importance score of each key evaluation metric based on feature importance, and then normalizes it to obtain the weights: in, Indicates the first i Importance scores of key evaluation indicators Indicates the first k Importance scores of key evaluation indicators This represents the summation of the importance scores for 1 to n key evaluation indicators. The first one obtained by the random forest model i Predicted weights of key evaluation indicators; The deep neural network model employs a multilayer perceptron structure. The input layer receives all key evaluation metric data, performs nonlinear transformations through hidden layers, and the output layer uses the softmax function to normalize the weights. in, Represents the deep neural network's first... i The exponential score of each output node. The first term obtained from the deep neural network model i Predicted weights of key evaluation indicators; The first obtained from the random forest model i Predicted weights of key evaluation indicators The first and deep neural network models obtained i Predicted weights of key evaluation indicators The weighted fusion yields the first... i Machine learning weights for key evaluation metrics: Where β is the machine learning fusion coefficient. For the first i Machine learning weights for key evaluation metrics.

10. The grounding grid safety assessment method based on fuzzy evaluation and machine learning according to claim 7, characterized in that, The method based on dynamic fusion coefficients dynamically fuses the predefined expert weights and machine learning weights of each key evaluation indicator to obtain a comprehensive weight for each key evaluation indicator, forming a comprehensive weight vector, including: By inputting the expert weights of each key evaluation indicator predefined by the grounding grid evaluation experts, an adaptive mechanism is used to dynamically fuse the predefined expert weights and the machine learning weights of each key evaluation indicator. The fusion formula is as follows: in, Represents the predefined first i Expert weights for key evaluation indicators, For the first i The machine learning weights for the key evaluation metrics, where α is the dynamic fusion coefficient. For the first i The combined weight of each key evaluation indicator; Based on the mean absolute error within the sliding window, α is dynamically updated. The formula for calculating the mean absolute error is: in, for t Predicted value of grounding grid safety assessment at any given time. for t The actual value of the grounding grid safety assessment at any given time. Where is the size of the time window, and MAE is the mean absolute error; Calculate the mean sliding window error based on the mean absolute error (MAE): in, W To adjust the sliding window size, This represents the mean error of the sliding window. Based on the average error of the sliding window Update the dynamic fusion coefficient α: in, This is the adjustment factor for the window step size. The average prediction error threshold, This sets the minimum value for the dynamic fusion coefficient α. This sets the maximum value for the dynamic fusion coefficient α, and clip represents the truncation function. Indicates the first t The dynamic fusion coefficient value for each evaluation period. Indicates the first t+1 The dynamic fusion coefficient value for each evaluation period, i.e., the updated dynamic fusion coefficient; Based on the combined weight of all key evaluation indicators This forms a comprehensive weight vector. .

11. The grounding grid safety assessment method based on fuzzy evaluation and machine learning according to claim 8 or 10, characterized in that, The process involves performing fuzzy synthesis operations using the fuzzy relation matrix and the comprehensive weight vector to obtain the membership degree of each safety level in the grounding grid. Then, the membership degree of each safety level in the grounding grid is defuzzified to obtain the grounding grid safety assessment score, including: Using the fuzzy relation matrix R and the comprehensive weight vector Fuzzy synthesis operations are performed, and a weighted average synthesis operator is used to calculate the membership degree of each safety level in the grounding grid, resulting in the membership degree vector of each safety level. B : in, Let R be the comprehensive weight vector, and let R be the fuzzy relation matrix. b j For the first j Membership degree of each security level, b m For the first m Membership degree of each security level, B This represents the membership vector for each security level. No. j Membership of each security level b j The calculation formula is: in, For the first i The combined weight of the key evaluation indicators, u ij Indicates the first i The first key evaluation indicator corresponds to the first j Membership degree of each security level; Membership vectors of each security level are determined using the centroid method. B Defuzzification is performed to obtain the grounding grid safety assessment score. S The calculation formula is: in, y j For the first j The score for each security level.

12. A computer storage medium, wherein the computer-readable storage medium stores a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the grounding grid safety assessment method based on fuzzy evaluation and machine learning as described in any one of claims 7-12.

Images