Electromagnetic compatibility evaluation and optimization system based on intelligent algorithm

Abstract

The invention discloses an electromagnetic compatibility evaluation and optimization system based on an intelligent algorithm, and relates to the technical field of electromagnetic compatibility evaluation, and the system comprises a data preparation module, a feature extraction and model training module, an evaluation integration module, a parameter optimization module and an optimization adjustment module which are operated in sequence. The technical key points are as follows: the scheme not only can predict whether the electromagnetic compatibility reaches the standard, but also provides a specific electromagnetic interference level prediction value through a neural network model, which is helpful for power plant managers to understand the condition of the electromagnetic compatibility deeper so as to take targeted optimization measures; by improving the accuracy of electromagnetic compatibility evaluation, realizing targeted optimization measures, improving the flexibility and efficiency of system adjustment and enhancing decision support of electromagnetic compatibility management, the problems of accurate evaluation and optimization in the electromagnetic compatibility problem are solved, and the technical effect of improving the electromagnetic compatibility management level of a power plant control room is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic compatibility evaluation, and in particular to an electromagnetic compatibility evaluation and optimization system based on an intelligent algorithm. Background Art

[0002] Electromagnetic compatibility assessment is the process of measuring and evaluating the ability of electronic and electrical equipment and systems to work properly in their intended electromagnetic environment and not cause unbearable electromagnetic disturbance to anything else in the environment. In simple terms, electromagnetic compatibility assessment is to ensure that the equipment can work properly in a complex electromagnetic environment without causing interference to other equipment or systems. Electromagnetic compatibility assessment is a very important part of the product design, development and production process, especially today when the number of electronic devices is increasing and the electromagnetic environment is becoming increasingly complex. Through electromagnetic compatibility assessment, the quality and reliability of the product can be ensured while meeting the requirements of relevant standards and regulations.

[0003] The existing application publication number is CN115935794A, and its name is a system-level electromagnetic compatibility evaluation method under small sample constraints. The present invention constructs a multi-level electromagnetic compatibility evaluation model, quantifies the electromagnetic compatibility test data of equipment, subsystems and systems by levels, determines the index weights, and performs bottom-up reasoning and integration of test sample data, and finally realizes the electromagnetic compatibility evaluation of the system as a whole. The present invention is based on the electromagnetic compatibility test sample data of a small sample, and uses the emission limit and equipment sensitivity limit specified in the GJB151B-2013 electromagnetic compatibility standard as the standard, and uses mathematical methods to calculate the electromagnetic compatibility of the system. By constructing a multi-level electromagnetic compatibility evaluation model, the electromagnetic compatibility test data of equipment, subsystems and systems are quantified by levels, the index weights are determined, and the bottom-up reasoning and integration of test sample data are performed, and finally the electromagnetic compatibility evaluation of the system as a whole is realized; but the above evaluation process cannot further improve the accuracy of the evaluation results.

[0004] In combination with the above documents and prior art:

[0005] In traditional scenarios, electromagnetic compatibility management within power stations often faces many problems and challenges; traditional electromagnetic interference detection and identification methods may be insufficient. Due to the lack of advanced data collection and analysis technology, the identification of potential electromagnetic interference sources and interfered equipment may not be accurate and timely, which may lead to electromagnetic compatibility issues being ignored or delayed in handling; in traditional systems, the discovery, reporting and handling of problems may involve multiple departments and complex processes, resulting in a slow response to problems and increasing the risk of equipment failure and downtime; in addition, the optimization suggestions provided by traditional methods may not be specific or effective enough. Due to the lack of precise data support and intelligent analysis, the suggestions given may be too general or unable to optimize for specific problems, thus failing to effectively solve electromagnetic compatibility problems. Summary of the invention

[0006] 1. Technical issues to be resolved

[0007] In view of the shortcomings of the prior art, the present invention provides an electromagnetic compatibility evaluation and optimization system based on an intelligent algorithm, which improves the accuracy of electromagnetic compatibility evaluation, implements targeted optimization measures, enhances the flexibility and efficiency of system adjustment, and enhances decision support for electromagnetic compatibility management, thereby solving the problems raised in the background technology.

[0008] (II) Technical solution

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0010] An electromagnetic compatibility evaluation and optimization system based on intelligent algorithm, comprising:

[0011] The data preparation module obtains the label data and feature data corresponding to the control room in the selected power plant as a data set, preprocesses the data set, and divides the data set to obtain a training set, a validation set, and a test set with a ratio of 3:1:1; it is characterized by:

[0012] The feature extraction and model training module receives the preprocessed data set, extracts relevant features, trains the SVM model using the training set, adjusts the SVM model using the validation set, and selects the model with the best performance; trains the neural network using the training set, adjusts the neural network using the validation set, and selects the model with the best performance;

[0013] The evaluation integration module integrates the trained SVM model and neural network model to construct a comprehensive evaluation model. Any control compartment to be evaluated is used as a sample and the comprehensive evaluation model is used for prediction.

[0014] If the two prediction results are consistent, the judgment mechanism is triggered. If the consistent prediction results are: the standard is met, no response action is taken; if the consistent prediction results are: the standard is not met, an early warning signal is issued;

[0015] If the two prediction results are inconsistent, an optimization instruction is issued until the two prediction results are consistent;

[0016] The parameter optimization module, upon receiving the optimization instruction, optimizes the parameters of the comprehensive evaluation model using an intelligent algorithm and continues to predict the same sample to be evaluated until the judgment mechanism is triggered;

[0017] The optimization and adjustment module, under the condition of receiving any warning signal, marks each control sub-chamber with a warning signal, and simultaneously executes the interference judgment strategy. The neural network model is used for the second time. After model training and evaluation, the relevant features of the corresponding marked control sub-chamber are input. The neural network model outputs the corresponding electromagnetic interference level prediction value according to the learned mapping relationship, and sorts the control sub-chambers corresponding to each electromagnetic interference level prediction value from large to small to generate a sorting table A. The ranking of the marked control sub-chamber is positively correlated with the priority of executing the adjustment mechanism.

[0018] Further, label data: setting a data set with labels, wherein the labels indicate whether the electromagnetic compatibility meets the standards; characteristic data: including but not limited to electromagnetic radiation measurement values, equipment working status parameters, environmental data, and equipment layout information;

[0019] Among them, the electromagnetic radiation measurement values ​​include electric field strength and magnetic field strength;

[0020] Equipment operating status parameters include voltage and current;

[0021] Environmental data includes temperature and humidity;

[0022] The device layout information includes the device spacing.

[0023] Furthermore, the relevant features include electromagnetic intensity features, electrical parameter features, environmental features, and layout features.

[0024] Further, about the SVM model:

[0025] Model training: Use the training set to train the SVM model, with the goal of finding a hyperplane that maximizes the separation of qualified and unqualified data points; select kernel functions and parameters; kernel functions include linear kernels, polynomial kernels, and RBFs; parameters include penalty coefficients C and kernel function parameters;

[0026] Model evaluation: Use the validation set to adjust the parameters of the SVM model and select the model with the best performance; use accuracy, recall, and F1 score as evaluation indicators, and use weighted average to obtain a comprehensive indicator of comprehensive accuracy, recall, and F1 score to evaluate the performance of the SVM model;

[0027] About the neural network model:

[0028] Model construction: Design the classification neural network structure, including input layer, hidden layer and output layer; determine the activation function as softmax activation function;

[0029] Model training: Use the training set to train the neural network, and adjust the network weights through the back-propagation algorithm to minimize the gap between the predicted value and the true value; use the validation set for early stopping;

[0030] Model evaluation: The principle is the same as that of the SVM model. The validation set is used to adjust the parameters and structure of the neural network, and the model with the best performance is selected. The same evaluation indicators are used to evaluate the performance of the neural network model.

[0031] Furthermore, intelligent algorithms include, but are not limited to, genetic algorithms and particle swarm optimization algorithms;

[0032] The optimization process is as follows:

[0033] During the training process of the SVM model and the neural network model, a genetic algorithm is used to optimize the parameters of the SVM model and the neural network model, and a fitness function is set. The fitness function calculates the fitness value according to the accuracy of the model on the validation set; and the parameters of the model are continuously optimized through N rounds of adjustment operations;

[0034] Wherein, N = {1, 2, ..., n}, n is a positive integer, and N represents the number of cycles;

[0035] Adjustment operations include selection, crossover, and mutation operations.

[0036] Furthermore, the neural network model is used for the second time to output the predicted value of the electromagnetic interference level of the control compartment with the corresponding mark. The process is as follows: on the basis of the original neural network model, the output layer is changed to use a linear activation function; and the evaluation indicators are changed to mean square error and root mean square error.

[0037] Furthermore, the process of executing the adjustment mechanism is as follows:

[0038] According to the predicted value of the electromagnetic interference level of the corresponding control chamber marked, a piecewise function is used to calculate the estimated value of the increased shielding material thickness, and according to the estimated value of the shielding material thickness, the thickness of the shielding material originally laid in the control chamber is increased.

[0039] Furthermore, a piecewise function is used to calculate the estimated value of the added shielding material thickness, based on the following formula:

[0040]

[0041] Wherein, H represents the estimated value of the added shielding material thickness, k1 and k2 are the first proportional constant and the second proportional constant, respectively, and 0<k1<1, 0.01<k2<0.1, I pred ,I thresh They represent the predicted value of the electromagnetic interference level and the acceptable electromagnetic interference threshold respectively, ΔI represents the threshold increment, and the value of ΔI is 10.

[0042] (III) Beneficial effects

[0043] The present invention provides an electromagnetic compatibility evaluation and optimization system based on an intelligent algorithm, which has the following beneficial effects:

[0044] (1) This solution can not only predict whether the electromagnetic compatibility meets the standard, but also provide specific electromagnetic interference level prediction values ​​through the neural network model, which helps power plant managers to have a deeper understanding of the electromagnetic compatibility status and take targeted optimization measures. Combining multiple machine learning models and rich feature data, it improves the accuracy and efficiency of electromagnetic compatibility assessment in the control room of the power plant, providing strong support for the safe and stable operation of the power plant.

[0045] (2) This solution improves the accuracy and robustness of electromagnetic compatibility assessment, reduces the uncertainty of model selection, improves the system's adaptability and optimization capabilities, and enhances the system's fault tolerance and decision support capabilities by integrating the SVM model and the neural network model and combining intelligent algorithms for parameter optimization operations.

[0046] (3) This solution solves the problem of accurate assessment and optimization of electromagnetic compatibility issues by improving the accuracy of electromagnetic compatibility assessment, implementing targeted optimization measures, enhancing the flexibility and efficiency of system adjustment, and strengthening decision support for electromagnetic compatibility management, thereby achieving the technical effect of improving the level of electromagnetic compatibility management in the power plant control room. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a system modular structure diagram of the present invention;

[0048] Figure 2 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] 1. Background Analysis

[0051] With the increasing number of electronic devices, the electromagnetic compatibility (EMC) issue has become increasingly important. Electromagnetic compatibility refers to the ability of a device or system to work normally in its electromagnetic environment and not cause unbearable electromagnetic disturbance to anything in the environment. In order to effectively evaluate and optimize electromagnetic compatibility, this scheme proposes an electromagnetic compatibility evaluation and optimization system based on intelligent algorithms to judge whether the electromagnetic compatibility of each sub-room in the power plant meets the standards;

[0052] In traditional scenarios, electromagnetic compatibility management within power stations often faces many problems and challenges;

[0053] First, traditional EMI detection and identification methods may be inadequate. Due to the lack of advanced data collection and analysis technology, the identification of potential EMI sources and affected equipment may not be accurate and timely, which may lead to EMC issues being ignored or delayed in processing. Second, the response to EMC issues is usually slow. In traditional systems, the discovery, reporting and processing of problems may involve multiple departments and complex processes, resulting in insufficient response to problems and increased risks of equipment failure and downtime. In addition, the optimization suggestions provided by traditional methods may not be specific or effective. Due to the lack of precise data support and intelligent analysis, the suggestions given may be too general or unable to optimize specific problems, thus failing to effectively solve EMC issues. Finally, traditional scenarios usually lack a unified platform to monitor and manage EMC issues, which may lead to information dispersion and poor communication, making it difficult for power station managers to fully understand and grasp the EMC status of the power station.

[0054] These problems not only affect the safety and efficiency of power station operation, but may also lead to equipment failures and increased downtime, posing a threat to the stability and reliability of power supply. Therefore, the researchers thought that by introducing an electromagnetic compatibility assessment and optimization system based on intelligent algorithms, these problems can be effectively solved and the electromagnetic compatibility management level of the power station can be improved.

[0055] 2. Brief description of the overall system structure

[0056] Data collection:

[0057] Responsible for collecting electromagnetic radiation data and working environment data of the corresponding indoor equipment; using IoT sensor technology to monitor and record electromagnetic environment data in real time, and clean, denoise and standardize the collected raw data; using data fusion technology to integrate data from different sensors;

[0058] Electromagnetic compatibility assessment:

[0059] Construct assessment models based on machine learning, such as support vector machines (SVMs) and neural networks, to conduct preliminary assessments of electromagnetic compatibility, and use intelligent algorithms (such as genetic algorithms and particle swarm optimization algorithms) to optimize the parameters of the assessment models to improve assessment accuracy;

[0060] Model integration and tuning optimization:

[0061] Combine multiple machine learning algorithms to build an integrated evaluation model; this can be achieved through voting, stacking or hybrid methods; based on the evaluation results, provide optimization suggestions to power plant managers, such as adjusting equipment layout, adding shielding measures, etc., to improve electromagnetic compatibility.

[0062] 3. Application of Intelligent Algorithms and Creative Design

[0063] Creatively combine multiple machine learning algorithms, such as SVM, neural network and decision tree, to build an integrated evaluation model. This model can combine the advantages of different algorithms to improve the accuracy and robustness of the evaluation; use genetic algorithm or particle swarm optimization algorithm to automatically adjust and optimize the parameters of the evaluation model to adapt to the changing electromagnetic environment; and based on; At the same time, the neural network model can also provide specific electromagnetic interference level prediction values, which helps to have a deeper understanding of the electromagnetic compatibility status and make specific and accurate improvement measures based on the prediction value.

[0064] Embodiment 1:

[0065] See also Figure 1 , This embodiment provides an electromagnetic compatibility evaluation and optimization system based on intelligent algorithms, the system includes a data preparation module, a feature extraction and model training module, an evaluation integration module, a parameter optimization module and an optimization adjustment module that are run in sequence, and the entire system acts as a control end to evaluate and optimize the electromagnetic compatibility of control sub-rooms of various built-in electrical equipment in a power plant;

[0066] Data preparation module:

[0067] Obtain the label data and feature data corresponding to the control room in the selected power plant as the data set, preprocess the data set, and divide the data set into a training set, a validation set, and a test set with a ratio of 3:1:1;

[0068] Among them, the power plant in this scheme has a main control room, which is the location where the entire system is installed, and several control sub-rooms. Each control sub-room is equipped with corresponding electrical equipment, such as generators, transformers and switch cabinets. The equipment types and distribution in each control sub-room are the same, and there are corresponding feature data. All control sub-rooms are composed to obtain the required label data;

[0069] Label data:

[0070] A labeled dataset is required, where the labels indicate whether the electromagnetic compatibility meets the standard. These data are obtained from historical records, experimental tests, or expert evaluations, which will not be elaborated here. For example, the electromagnetic compatibility evaluation data of 100 control rooms are obtained from historical records, of which 50 are marked as "meeting the standard" and 50 are marked as "not meeting the standard".

[0071] Feature data:

[0072] Including but not limited to electromagnetic radiation measurement values, equipment working status parameters, environmental data and equipment layout information; wherein, characteristic data represents characteristic data corresponding to each control sub-chamber, electromagnetic radiation measurement values ​​include electric field strength and magnetic field strength, these data can reflect the electromagnetic radiation generated by each device in the power plant; equipment working status parameters include voltage and current, which are used to analyze voltage fluctuations and current stability, these data can help analyze the electromagnetic characteristics of the equipment during normal operation; environmental data includes temperature and humidity in the corresponding control sub-chamber, these environmental factors may also affect electromagnetic compatibility; equipment layout information includes equipment spacing, equipment type, etc., and equipment layout information is crucial for evaluating electromagnetic interference between devices;

[0073] The process of characteristic data is directly monitored by arranging corresponding sensors in the corresponding control chambers. For example, temperature and humidity can be obtained by installing temperature and humidity sensors. For other types of data, corresponding types of sensors are required for real-time acquisition or collection.

[0074] Take label data and feature data as data sets;

[0075] The process of preprocessing the dataset is as follows:

[0076] Data cleaning: remove duplicate, invalid or abnormal data points; data denoising: reduce random noise in the data set through filtering algorithms; data standardization: convert data to a unified measurement range to facilitate subsequent analysis; data fusion: integrate data from different sensors to form a complete description of the electromagnetic environment;

[0077] Divide the dataset into training, validation and test sets;

[0078] Among them, the training set is used to train the model, the validation set is used to adjust the model parameters and select the best model, and the test set is used to evaluate the final performance of the model; for example: training set: 60 samples (30 meet the standard, 30 do not meet the standard); validation set: 20 samples (10 meet the standard, 10 do not meet the standard); test set: 20 samples (10 meet the standard, 10 do not meet the standard).

[0079] Feature extraction and model training module:

[0080] Receive the preprocessed data set, extract relevant features, including electromagnetic intensity features, electrical parameter features, environmental features, and layout features, use the training set to train the SVM model with the goal of obtaining a hyperplane, use the validation set to adjust the SVM model, and select the model with the best performance; use the training set to train the neural network, adjust the network weights through the back-propagation algorithm, use the validation set to adjust the neural network, and select the model with the best performance;

[0081] Electromagnetic strength characteristics: electric field strength and magnetic field strength within a set frequency range;

[0082] Electrical parameter characteristics: voltage waveform and current waveform to reflect the parameters of the electrical performance of the equipment;

[0083] Environmental characteristics: changes in temperature and humidity;

[0084] Layout features: Equipment spacing is converted into numerical features;

[0085] It should be noted that the set frequency range represents a specific frequency range, that is, after analyzing previous data or based on experience, the electric field strength and magnetic field strength data within a certain range can be obtained to ensure the effectiveness of subsequent training;

[0086] Among them, about the SVM model:

[0087] Model training: Use the training set to train the SVM model. The goal is to find a hyperplane that can maximize the separation of qualified and unqualified data points; select appropriate kernel functions (such as linear kernels, polynomial kernels, RBF kernels, etc.) and parameters (such as penalty coefficient C and kernel function parameters);

[0088] Model evaluation: Use the validation set to adjust the parameters of the SVM model and select the model with the best performance; evaluate the performance of the model through indicators such as accuracy, recall, and F1 score. Usually, a weighted average method is used to obtain a comprehensive index after comprehensive accuracy, recall, and F1 score, which is used for refined evaluation of model performance. The comprehensive index is positively correlated with model performance. The specific evaluation process is not described in detail here; the weight coefficients of each evaluation indicator are the same;

[0089] About the neural network model:

[0090] Model construction: Design a classification neural network structure, including input layer, hidden layer and output layer; determine the activation function as softmax activation function to output the probability of each category;

[0091] Model training: Use the training set to train the neural network, and adjust the network weights through the back-propagation algorithm to minimize the gap between the predicted value and the true value; use the validation set for early stopping or other regularization techniques to prevent overfitting;

[0092] Model evaluation: Similar to the SVM model, the validation set is used to adjust the parameters and structure of the neural network, and the model with the best performance is selected to evaluate the performance of the neural network through the same evaluation indicators (such as accuracy, recall rate, F1 score, etc.). As it has been explained above, the process of model evaluation will not be elaborated here.

[0093] In the above feature extraction and model training module, both the SVM model and the neural network model are used to predict whether the electromagnetic compatibility of the corresponding new control room meets the standard. For the new electric control room sample, each data in its data set is input into the trained model; the model will output the probability of the sample being "meets the standard" or "does not meet the standard", and select the category with a higher probability as the final prediction result.

[0094] Improved forecast accuracy:

[0095] By using the SVM model and the neural network model in combination with multiple feature data (including electromagnetic intensity, electrical parameters, environmental data and layout information), the electromagnetic compatibility of the control room can be evaluated more comprehensively. This multi-model and multi-feature approach improves the accuracy and reliability of the prediction.

[0096] Automated evaluation:

[0097] Traditional electromagnetic compatibility assessment may rely on manual inspection and expert evaluation, which is inefficient and susceptible to subjective factors. This technical solution realizes rapid and automatic evaluation of electromagnetic compatibility through automated data collection, preprocessing, feature extraction and model prediction.

[0098] Improved evaluation efficiency:

[0099] By dividing the model into training sets, validation sets, and test sets, the model can be efficiently trained and adjusted to ensure the performance of the model in practical applications. In addition, once the model training is completed, the electromagnetic compatibility prediction of the new control room can be quickly performed, which greatly improves the evaluation efficiency.

[0100] By adopting the above technical solutions, not only can it be predicted whether the electromagnetic compatibility meets the standards, but also the neural network model can provide specific electromagnetic interference level prediction values, which helps power plant managers to have a deeper understanding of the electromagnetic compatibility status and take targeted optimization measures; combining multiple machine learning models and rich feature data, it improves the accuracy and efficiency of electromagnetic compatibility assessment in the control room of the power plant, providing strong support for the safe and stable operation of the power plant.

[0101] Evaluation Integration Module:

[0102] The trained SVM model and neural network model are integrated to construct a comprehensive evaluation model. For any control compartment to be evaluated as a sample, the comprehensive evaluation model is used for prediction, that is, the SVM model and the neural network model are used for prediction respectively.

[0103] If the two prediction results are consistent, the judgment mechanism is triggered;

[0104] When the consistent prediction result is: reaching the target, no response action is taken;

[0105] When the consistent prediction result is: not reaching the target, a warning signal is issued;

[0106] If the two prediction results are inconsistent, an optimization instruction is issued until the two prediction results are consistent;

[0107] Among them, the example process of using the comprehensive evaluation model for prediction is as follows:

[0108] Example: Electromagnetic compatibility assessment of the corresponding control room in the power plant:

[0109] 1. Data preparation

[0110] Dataset:

[0111] Label data: The electromagnetic compatibility assessment data of 100 control rooms were obtained from historical records, of which 50 were marked as "compliant" and 50 were marked as "not compliant";

[0112] Characteristic data: The data of each control room includes electromagnetic radiation measurement values ​​(such as electric field strength, magnetic field strength), equipment working status parameters (such as voltage fluctuation, current stability), environmental data (average temperature and humidity in the control room), and equipment layout information (equipment spacing, equipment type, etc.);

[0113] Data partitioning:

[0114] Training set: 60 samples (30 met the standard, 30 did not meet the standard)

[0115] Validation set: 20 samples (10 met the criteria, 10 did not)

[0116] Test set: 20 samples (10 met the standard, 10 did not meet the standard)

[0117] 2. SVM model

[0118] Model training: Use 60 samples in the training set to train an SVM model, select the RBF kernel function, and initialize the penalty coefficient C and kernel function parameters; the training goal is to find an optimal hyperplane to maximize the distinction between control rooms that meet the standards and those that do not;

[0119] Model evaluation: Use the validation set to adjust the parameters of the SVM, such as C and kernel function parameters, to achieve the best performance; evaluate the accuracy, recall, and F1 score of the model through the validation set, assuming that the best model has an accuracy of 85%;

[0120] 3. Neural Network Model

[0121] Model construction: Design a neural network with an input layer, two hidden layers, and an output layer; choose softmax as the activation function.

[0122] Model training: Use the training set to train the neural network and adjust the network weights by minimizing the gap between the predicted value and the true value; apply early stopping techniques to prevent overfitting and stop training when the loss on the validation set is no longer significantly reduced.

[0123] Model evaluation: Use the validation set to adjust the parameters and structure of the neural network and select the model with the best performance; assume that the best model has an accuracy of 80% on the validation set.

[0124] 4. Model integration and prediction (corresponding to the above evaluation integration module)

[0125] Model integration: Integrate the trained SVM model and neural network model;

[0126] Judgment mechanism and optimization instructions: For each sample in the test set, the SVM model and the neural network model are used for prediction respectively; if both models predict that a control room meets the standard or both predict that it does not meet the standard, the final evaluation result is consistent with the prediction; if the prediction results are inconsistent, parameter optimization can be used.

[0127] In addition, unlike the optimization instructions in this embodiment;

[0128] If the predictions of SVM and neural network on a certain sample are inconsistent, it is also possible to decide which model to use based on the performance of the two models on the validation set.

[0129] Parameter optimization module:

[0130] Under the condition of receiving the optimization instruction, the intelligent algorithm is used to optimize the parameters of the comprehensive evaluation model, and the prediction of the same sample to be evaluated is continued until the judgment mechanism is triggered;

[0131] Among them, intelligent algorithms include but are not limited to genetic algorithms and particle swarm optimization algorithms;

[0132] The optimization process is as follows:

[0133] During the training process of the SVM model and the neural network model, a genetic algorithm is used to optimize the parameters of the SVM model and the neural network model, and a fitness function is set. The fitness function calculates the fitness value according to the accuracy of the model on the validation set; and the parameters of the model are continuously optimized through N rounds of adjustment operations;

[0134] N = {1, 2, ..., n}, n is a positive integer, and N represents the number of cycles, for example: starting from the first round, after optimizing the parameters of the model, a cycle is performed to predict the same sample to be evaluated. If the judgment mechanism is not triggered, the second round is continued, and after optimizing the parameters of the model, a second cycle is performed to predict the same sample to be evaluated, and this cycle is repeated until the judgment mechanism is triggered;

[0135] Among them, a rule engine can also be introduced. When the duration of the cycle exceeds the set duration T, whether the judgment mechanism is triggered or not, it is necessary to decide which model to use for prediction based on the performance of the SVM model and the neural network model on the validation set. The performance here refers to the accuracy rate.

[0136] Adjustment operations include selection, crossover, and mutation operations;

[0137] Specifically, during the model training process, a genetic algorithm is used to optimize the parameters of the SVM and neural network; for example, a fitness function is set that calculates the fitness value based on the accuracy of the model on the validation set; then, the model parameters are continuously optimized through the selection, crossover, and mutation operations of the genetic algorithm to improve the evaluation accuracy of the model.

[0138] Improved forecast accuracy and robustness:

[0139] By integrating the SVM model and the neural network model, this technical solution combines the advantages of two different models. The SVM model performs well when processing linearly separable or approximately linearly separable data, while the neural network model can handle more complex nonlinear relationships. The combination of the two can more comprehensively capture the characteristics of the data, thereby improving the accuracy of the prediction; at the same time, when the prediction results of the two models are consistent, this consistency enhances the reliability of the prediction results;

[0140] Reduced uncertainty in model selection:

[0141] In practical applications, choosing a suitable prediction model is often a challenge. Different models may have different performances on the same data set. By integrating two different types of models and determining the final result based on their prediction consistency, this technical solution reduces the reliance on a single model, thereby reducing the uncertainty of model selection.

[0142] Improved the system's adaptability and optimization capabilities:

[0143] By introducing a parameter optimization module and using intelligent algorithms to optimize model parameters, this technical solution enables the system to automatically adjust model parameters according to actual conditions to improve prediction performance. This adaptive optimization capability makes the system more robust in the face of complex and changing data environments.

[0144] Enhanced system fault tolerance and decision support capabilities:

[0145] When the prediction results of the SVM model and the neural network model are inconsistent, the system will issue optimization instructions to optimize the parameters until the prediction results of the two models are consistent. This mechanism enhances the fault tolerance of the system and ensures that reliable predictions can still be made when faced with complex or ambiguous data. At the same time, this also provides stronger support for decision makers and reduces the decision-making risks caused by inconsistent model predictions.

[0146] In summary, this technical solution improves the accuracy and robustness of electromagnetic compatibility assessment, reduces the uncertainty of model selection, improves the system's adaptability and optimization capabilities, and enhances the system's fault tolerance and decision support capabilities by integrating the SVM model and the neural network model and combining it with a parameter optimization module.

[0147] Optimization and adjustment module:

[0148] Under the condition of receiving any early warning signal, each control sub-chamber with an early warning signal is marked, and the interference judgment strategy is executed synchronously. The neural network model is used for the second time. After model training and evaluation, the relevant features of the corresponding marked control sub-chamber are input. The neural network model outputs the corresponding electromagnetic interference level prediction value according to the learned mapping relationship, and the control sub-chambers corresponding to each electromagnetic interference level prediction value are sorted from large to small to generate a sorting table A. The front ranking of the marked control sub-chamber is positively correlated with the priority of executing the adjustment mechanism. The adjustment mechanism is preferentially executed on the control sub-chambers at the front in the sorting table A.

[0149] The purpose of implementing the adjustment mechanism is to: increase the thickness of the shielding material in the corresponding control compartment to improve electromagnetic compatibility;

[0150] The process of implementing the adjustment mechanism is as follows:

[0151] According to the predicted value of the electromagnetic interference level of the corresponding control chamber marked, a piecewise function is used to calculate the estimated value of the added shielding material thickness, and according to the estimated value of the shielding material thickness, the thickness of the shielding material originally laid in the control chamber is increased; wherein;

[0152] The process of using the neural network model to output the predicted value of the electromagnetic interference level of the corresponding marked control compartment is as follows:

[0153] Based on the original neural network model, the output layer was changed to use a linear activation function; and the evaluation indicators were changed to mean square error (MSE) and root mean square error (RMSE);

[0154] It should be noted that the original neural network model was used to predict whether the control compartment met the standards;

[0155] The difference between the secondary neural network model and the original neural network model is:

[0156] Different goals: Predicting whether the standard is met is a classification problem, the purpose is to determine whether the electric control room meets the electromagnetic compatibility standard; predicting the electromagnetic interference level is a regression problem, the purpose is to give a specific electromagnetic interference value;

[0157] The output layer design is different: the output layer of the classification problem usually uses the softmax activation function to output the probability of each category; the output layer of the regression problem may use a linear activation function to directly output the predicted electromagnetic interference level value.

[0158] Different evaluation indicators: The evaluation indicators of classification problems include accuracy, recall, F1 score, etc.; the evaluation indicators of regression problems include mean square error (MSE), root mean square error (RMSE), etc.

[0159] Prediction order:

[0160] In practical applications, the neural network model is usually used first to predict whether the electric control room meets the standards. If the prediction result is "meets the standards", there may be no need to further predict the electromagnetic interference level. However, if the prediction result is "does not meet the standards", it may be necessary to further predict the specific electromagnetic interference level in order to understand the problem more accurately and take corresponding improvement measures. This order can improve efficiency and avoid more complex electromagnetic interference level predictions when it is not necessary.

[0161] The application of the neural network model to predict EMI levels involves the following steps:

[0162] Data collection and preprocessing:

[0163] First, a large amount of data related to the electromagnetic interference level needs to be collected; this data may include environmental parameters of the electric control room (such as temperature, humidity), equipment configuration, intensity and frequency of electromagnetic radiation sources, etc.; these data are preprocessed, including data cleaning, standardization and normalization, so that the neural network can better learn and process these data;

[0164] Neural network model design:

[0165] Design a multi-layer feedforward neural network, which includes an input layer, a hidden layer, and an output layer; the input layer is responsible for receiving pre-processed data, such as environmental parameters, equipment configuration, etc.; the hidden layer is responsible for learning and extracting the features of the input data, and mapping these features to a higher level representation through a nonlinear activation function; the output layer is responsible for predicting the electromagnetic interference level based on the features extracted by the hidden layer;

[0166] Training the Neural Network:

[0167] Use the collected data set to train the neural network. This usually involves selecting a suitable loss function (such as mean square error MSE) and using an optimization algorithm (such as gradient descent) to minimize the loss function; through multiple iterations of training, the neural network gradually learns the mapping relationship from input data to electromagnetic interference levels;

[0168] Model evaluation and optimization:

[0169] Use the validation set to evaluate the performance of the neural network to help understand whether the model is overfitting or underfitting; optimize the model based on the evaluation results, such as adjusting the network structure, increasing or decreasing the number of hidden layer nodes, changing the activation function, etc.;

[0170] Predicted EMI levels:

[0171] Once the neural network model is trained and passes the evaluation, it can be used to predict the electromagnetic interference level of the new electric control sub-room. The relevant data of the new electric control sub-room is input into the neural network model, and the model will output a predicted value of the electromagnetic interference level based on the learned mapping relationship.

[0172] Example:

[0173] Suppose we have a trained neural network model to predict the electromagnetic interference level of an electric control room. Now, we have a new electric control room with environmental parameters of 25°C temperature, 50% humidity, and one type A device and one type B device. We input these data into the neural network model, and after calculation, the model outputs a predicted value of the electromagnetic interference level, such as 45dBμV / m. This predicted value can help us understand the electromagnetic interference status of the electric control room, so that we can take corresponding measures to optimize electromagnetic compatibility.

[0174] A piecewise function is used to calculate the estimated added shielding material thickness based on the following formula:

[0175]

[0176] Wherein, H represents the estimated value of the added shielding material thickness, k1 and k2 are the first proportional constant and the second proportional constant, respectively, and 0<k1<1, 0.01<k2<0.1, I pred ,I thresh They represent the predicted value of electromagnetic interference level and the acceptable electromagnetic interference threshold respectively, ΔI represents the threshold increment, and the value of ΔI is 10;

[0177] The above formula uses a piecewise function to express the relationship between the thickness of the shielding material and the electromagnetic interference level exceeding the threshold. The first proportional constant is also the proportional constant of the linear growth stage, indicating that when the electromagnetic interference level prediction value is within the linear range, the amount by which the shielding material thickness needs to be increased for each unit increase in the electromagnetic interference level prediction value; the second proportional constant is also the proportional constant of the exponential growth stage, which is related to the shielding effectiveness of the material and the electromagnetic wave frequency factors; the threshold increment indicates how much the electromagnetic interference level exceeds the acceptable threshold, and the increase in the thickness of the shielding material changes from linear to exponential growth;

[0178] Logic description:

[0179] Linear growth stage: When the predicted value of the EMI level exceeds the acceptable EMI threshold, but the excess does not exceed the threshold increment, the thickness of the shielding material increases in a linear relationship. This is because within this range, increasing the thickness of the shielding material can more effectively reduce EMI.

[0180] Exponential growth stage: When the predicted value of the EMI level exceeds the sum of the acceptable EMI threshold and the threshold increment, the thickness of the shielding material will increase exponentially. This is because at high EMI levels, thicker shielding materials may be required to achieve the same shielding effect.

[0181] Example:

[0182] Assume that in a power plant control room, the predicted value of the electromagnetic interference level predicted by the neural network model = 120 (based on an arbitrary unit), and the acceptable electromagnetic interference threshold = 100; set the threshold increment = 10, the first proportional constant and the second proportional constant are 0.1 and 0.05 respectively;

[0183] If the predicted value of the electromagnetic interference level = 105, it is in the linear growth stage. According to the formula, the thickness of the shielding material that needs to be increased is T = 0.1 × (105-100) = 0.5 unit thickness, usually mm;

[0184] If the predicted value of the electromagnetic interference level = 130, it is in the exponential growth stage. According to the formula, the thickness of the shielding material that needs to be increased is T = 0.05 × e∧ (130-100-10) + 0.1 × 10 = 2.72 unit thickness (approximately equal to, the specific value may vary depending on the calculation accuracy), usually mm;

[0185] This formula is a corresponding model that explains how to calculate the thickness of the shielding material that needs to be increased based on the predicted value of the electromagnetic interference level; in practical applications, it may be necessary to adjust the parameters in the formula according to the specific situation and verify its effectiveness through experiments.

[0186] Improved accuracy of EMC assessments:

[0187] By reusing the neural network model to predict specific EMI levels, the technical solution provides more detailed assessment information than simple classification (meeting or not meeting the standard), which helps to more accurately identify and solve EMI problems;

[0188] Targeted optimization measures have been implemented:

[0189] According to the predicted value of the electromagnetic interference level, the added shielding material thickness is calculated using a piecewise function, which can ensure that shielding materials of corresponding thickness are used for different levels of electromagnetic interference. This targeted optimization measure improves the efficiency of resource utilization and avoids unnecessary waste.

[0190] Improved flexibility and efficiency of system adjustment:

[0191] By marking control rooms with early warning signals and sorting them according to the level of electromagnetic interference, the technical solution can prioritize the areas with the most serious interference. This prioritization ensures flexibility and efficiency in system adjustments, allowing resources to be invested in the areas that need improvement the most first;

[0192] Enhanced decision support for EMC management:

[0193] By providing detailed electromagnetic interference level predictions and optimized shielding material thickness recommendations, this technical solution provides strong decision support for electromagnetic compatibility management. Decision makers can develop more effective electromagnetic compatibility improvement plans based on this information.

[0194] In summary, this technical solution solves the difficulties of accurate evaluation and optimization in electromagnetic compatibility issues by improving the accuracy of electromagnetic compatibility assessment, implementing targeted optimization measures, improving the flexibility and efficiency of system adjustment, and enhancing decision support for electromagnetic compatibility management, thereby achieving the technical effect of improving the level of electromagnetic compatibility management in power plant control rooms.

[0195] Embodiment 2:

[0196] See also Figure 2 Based on Example 1, this embodiment further provides an electromagnetic compatibility evaluation and optimization method based on an intelligent algorithm, comprising the following specific steps:

[0197] S1. Obtain label data and feature data corresponding to the control room in the selected power plant as a data set, preprocess the data set, and divide the data set to obtain a training set, a validation set, and a test set with a ratio of 3:1:1;

[0198] S2, receiving the preprocessed data set, extracting relevant features, using the training set to train the SVM model, and using the validation set to adjust the SVM model, and selecting the model with the best performance; using the training set to train the neural network, and using the validation set to adjust the neural network, and selecting the model with the best performance;

[0199] S3, integrating the trained SVM model and the neural network model to construct a comprehensive evaluation model, taking any control compartment to be evaluated as a sample, and using the comprehensive evaluation model for prediction;

[0200] If the two prediction results are consistent, the judgment mechanism is triggered. If the consistent prediction results are: the standard is met, no response action is taken; if the consistent prediction results are: the standard is not met, an early warning signal is issued;

[0201] If obtained, continue until the two prediction results are consistent;

[0202] S4. Under the condition of receiving the optimization instruction, the parameters of the comprehensive evaluation model are optimized using the intelligent algorithm, and the same sample to be evaluated is predicted continuously until the judgment mechanism is triggered;

[0203] S5. Under the condition of receiving any warning signal, each control sub-chamber with a warning signal is marked, and the interference judgment strategy is executed synchronously. The neural network model is used for the second time. After model training and evaluation, the relevant features of the corresponding marked control chamber are input. The neural network model outputs the corresponding electromagnetic interference level prediction value according to the learned mapping relationship, and the control sub-chambers corresponding to each electromagnetic interference level prediction value are sorted from large to small to generate a sorting table A. The ranking of the marked control chamber is positively correlated with the priority of executing the adjustment mechanism.

[0204] In the application, the several formulas involved are all calculated by taking their numerical values ​​after removing the dimensions, and the formulas are established by collecting a large amount of data and performing software simulation to obtain a formula for the most recent actual situation. Some coefficients or weights in the formulas are set by technical personnel in this field according to actual conditions, so they will not be elaborated here.

[0205] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0206] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0207] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An electromagnetic compatibility evaluation and optimization system based on intelligent algorithm, the system comprising: The data preparation module obtains the label data and feature data corresponding to the control room in the selected power plant as a data set, preprocesses the data set, and divides the data set to obtain a training set, a validation set, and a test set with a ratio of 3:1:1; it is characterized by: The feature extraction and model training module receives the preprocessed data set, extracts relevant features, trains the SVM model using the training set, adjusts the SVM model using the validation set, and selects the model with the best performance; trains the neural network using the training set, adjusts the neural network using the validation set, and selects the model with the best performance; The evaluation integration module integrates the trained SVM model and neural network model to construct a comprehensive evaluation model. Any control compartment to be evaluated is used as a sample and the comprehensive evaluation model is used for prediction. If the two prediction results are consistent, the judgment mechanism is triggered. If the consistent prediction results are: the standard is met, no response action is taken; if the consistent prediction results are: the standard is not met, an early warning signal is issued; If the two prediction results are inconsistent, an optimization instruction is issued until the two prediction results are consistent; The parameter optimization module, upon receiving the optimization instruction, optimizes the parameters of the comprehensive evaluation model using an intelligent algorithm and continues to predict the same sample to be evaluated until the judgment mechanism is triggered; The optimization and adjustment module, under the condition of receiving any warning signal, marks each control sub-chamber with a warning signal, and simultaneously executes the interference judgment strategy. The neural network model is used for the second time. After model training and evaluation, the relevant features of the corresponding marked control sub-chamber are input. The neural network model outputs the corresponding electromagnetic interference level prediction value according to the learned mapping relationship, and sorts the control sub-chambers corresponding to each electromagnetic interference level prediction value from large to small to generate a sorting table A. The ranking of the marked control sub-chamber is positively correlated with the priority of executing the adjustment mechanism.

2. The electromagnetic compatibility evaluation and optimization system based on intelligent algorithm according to claim 1, characterized in that: Label data: set a data set with labels, where the labels indicate whether the electromagnetic compatibility meets the standards; characteristic data: including but not limited to electromagnetic radiation measurement values, equipment working status parameters, environmental data, and equipment layout information; Among them, the electromagnetic radiation measurement values ​​include electric field strength and magnetic field strength; Equipment operating status parameters include voltage and current; Environmental data includes temperature and humidity; The device layout information includes the device spacing.

3. The electromagnetic compatibility evaluation and optimization system based on intelligent algorithm according to claim 2 is characterized in that: The relevant features include electromagnetic strength features, electrical parameter features, environmental features, and layout features.

4. The electromagnetic compatibility evaluation and optimization system based on intelligent algorithm according to claim 3 is characterized in that: About SVM model: Model training: Use the training set to train the SVM model, with the goal of finding a hyperplane that maximizes the separation of qualified and unqualified data points; select kernel functions and parameters; kernel functions include linear kernels, polynomial kernels, and RBFs; parameters include penalty coefficients C and kernel function parameters; Model evaluation: Use the validation set to adjust the parameters of the SVM model and select the model with the best performance; use accuracy, recall, and F1 score as evaluation indicators, and use weighted average to obtain a comprehensive indicator of comprehensive accuracy, recall, and F1 score to evaluate the performance of the SVM model; About the neural network model: Model construction: Design the classification neural network structure, including input layer, hidden layer and output layer; determine the activation function as softmax activation function; Model training: Use the training set to train the neural network and adjust the network weights through the back-propagation algorithm to minimize the gap between the predicted value and the true value; Use the validation set for early stopping; Model evaluation: The principle is the same as that of the SVM model. The validation set is used to adjust the parameters and structure of the neural network, and the model with the best performance is selected. The same evaluation indicators are used to evaluate the performance of the neural network model.

5. The electromagnetic compatibility evaluation and optimization system based on intelligent algorithm according to claim 4 is characterized in that: Intelligent algorithms include but are not limited to genetic algorithms and particle swarm optimization algorithms; The optimization process is as follows: During the training process of the SVM model and the neural network model, a genetic algorithm is used to optimize the parameters of the SVM model and the neural network model, and a fitness function is set. The fitness function calculates the fitness value according to the accuracy of the model on the validation set; and the parameters of the model are continuously optimized through N rounds of adjustment operations; Wherein, N = {1, 2, ..., n}, n is a positive integer, and N represents the number of cycles; Adjustment operations include selection, crossover, and mutation operations.

6. The electromagnetic compatibility evaluation and optimization system based on intelligent algorithm according to claim 5, characterized in that: The process of reusing the neural network model to output the predicted value of the electromagnetic interference level of the corresponding marked control room is as follows: based on the original neural network model, the output layer is changed to use a linear activation function; and the evaluation indicators are changed to mean square error and root mean square error.

7. The electromagnetic compatibility evaluation and optimization system based on intelligent algorithm according to claim 6, characterized in that: The process of implementing the adjustment mechanism is as follows: According to the predicted value of the electromagnetic interference level of the corresponding control chamber marked, a piecewise function is used to calculate the estimated value of the increased shielding material thickness, and according to the estimated value of the shielding material thickness, the thickness of the shielding material originally laid in the control chamber is increased.

8. The electromagnetic compatibility evaluation and optimization system based on intelligent algorithm according to claim 7, characterized in that: A piecewise function is used to calculate the estimated added shielding material thickness based on the following formula: Wherein, H represents the estimated value of the added shielding material thickness, k1 and k2 are the first proportional constant and the second proportional constant, respectively, and 0<k1<1, 0.01<k2<0.1, I pred ,I thresh They represent the predicted value of the electromagnetic interference level and the acceptable electromagnetic interference threshold respectively, ΔI represents the threshold increment, and the value of ΔI is 10.

Images