A smart park infrastructure intelligent monitoring system

Abstract

The application discloses a kind of intelligent park infrastructure intelligent monitoring system, it is related to the technical field of intelligent monitoring, including fusion module, prediction module and optimization module, set state related parameter, characteristic data is obtained by fusion, state prediction model is obtained by training, state prediction model is optimized.The application fusion module fuses the characteristic value of multiple state related parameters, reduces the complexity and delay of data processing, improves the monitoring efficiency through automated data processing and model optimization, the characteristic value of multiple related parameters is fused by fusion module, avoid the limitation of single parameter, optimize state prediction model, can dynamically adjust model parameters according to actual operation data, further improve the prediction accuracy of model, can early warning potential failure risk, take measures in time to maintain and repair, reduce the probability of failure occurrence, improve the reliability and stability of power supply facilities.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of intelligent monitoring, and in particular to a smart park infrastructure intelligent monitoring system. BACKGROUND

[0002] In recent years, with the continuous progress of technology, the smart park infrastructure intelligent monitoring technology will develop in the direction of more intelligence, and the Internet of Things technology plays an important role in the smart park infrastructure monitoring. By deploying a large number of intelligent sensors, the park environment, equipment and the like are comprehensively perceived and monitored in real time, and the edge computing node can process and analyze the data collected by the front-end equipment in real time, and quickly respond to various demands in the park, such as security monitoring, energy management and the like.

[0003] At present, in the Chinese patent for invention with the publication number CN115457471A, a smart park monitoring method and system are disclosed, which can obtain the changes of the emergency conditions in the park in real time by analyzing the video data, classifying the facility operation data and the environment data of each area in the park environment information, and dividing the fire-fighting grades of the park fire-fighting facilities, thereby bringing a certain risk and improving the safety and intelligence. However, in the related technology, the operation state of the power supply network is not analyzed and predicted according to the historical data, the scientificity of data utilization and the persuasiveness of prediction are lacking, different dimensional data are not fused to obtain feature data, which is not conducive to the simplicity of operation and the rapid response of system prediction, the prediction model is not updated according to the evaluation standard of real-time calculation, which is not conducive to the dynamic response of monitoring prediction. SUMMARY

[0004] The technical problem solved by the application is that in the related technology, the operation state of the power supply network is not analyzed and predicted according to the historical data, the scientificity of data utilization and the persuasiveness of prediction are lacking, different dimensional data are not fused to obtain feature data, which is not conducive to the simplicity of operation and the rapid response of system prediction, the prediction model is not updated according to the evaluation standard of real-time calculation, which is not conducive to the dynamic response of monitoring prediction.

[0005] To solve the above technical problems, the application provides the following technical scheme: a smart park infrastructure intelligent monitoring system, comprising a fusion module, a prediction module and an optimization module;

[0006] The fusion module sets state-related parameters, fuses the characteristic values of the state-related parameters, and obtains feature data;

[0007] The prediction module trains a decision tree according to the feature data, and obtains a state prediction model;

[0008] The optimization module optimizes the state prediction model, and predicts the state of the power supply facility according to the optimized state prediction model.

[0009] As a preferred scheme of the intelligent monitoring system for infrastructure of the smart park, the historical related data of the power supply facility includes historical voltage, historical current, historical power switch tube temperature, historical transformer temperature, historical voltage total harmonic distortion rate, historical current total harmonic distortion rate, and historical power supply facility state.

[0010] The historical power supply facility state includes a normal state and an abnormal state.

[0011] The number of times of exceeding the limit of historical voltage fluctuation is counted according to historical voltage.

[0012] The historical voltage, the historical current, the historical power switch tube temperature, the historical transformer temperature, the historical voltage total harmonic distortion rate, and the historical current total harmonic distortion rate are set as state related parameters.

[0013] As a preferred scheme of the intelligent monitoring system for infrastructure of the smart park, the statistical logic of the number of times of exceeding the limit of historical voltage fluctuation includes:

[0014] A first difference value of any two adjacent historical voltages is calculated, an absolute value of the first difference value is selected, a first value is set as a voltage fluctuation threshold, the absolute value of the first difference value is compared with the first value, when the first value is less than or equal to the first value, the next set of adjacent historical voltages is jumped to, and the steps of difference calculation and comparison are repeated, when the absolute value of the first difference value is greater than the first value, the number of times of exceeding the limit of historical voltage fluctuation is weighted by 1 to obtain a first sum value, the first sum value is used for the value of the absolute value of the first difference value greater than the first value in the next time, and the first sum value is updated according to a new sum value obtained by each time of weighting.

[0015] As a preferred scheme of the intelligent monitoring system for infrastructure of the smart park, the fusion module pre-processes the state related parameters, and the pre-processing includes cleaning processing and abnormal value elimination processing.

[0016] The cleaning processing is processed by a data cleaning program.

[0017] The logic of the abnormal value processing includes:

[0018] Select any kind of state related parameters, calculate the average value and standard deviation of the state related parameters, calculate the first difference value of each state related parameter and the average value of the state related parameters respectively, set the value of three times of the standard deviation as the second value, compare the first difference value with the second value, when the first difference value is greater than the second value, set the state related parameter as an abnormal value, eliminate the state related parameter and jump to the next state related parameter, when the first difference value is less than or equal to the second value, set the state related parameter as a normal value, retain the state related parameter and jump to the next state related parameter.

[0019] The average value of the state related parameters after preprocessing is set as the corresponding feature value.

[0020] As a preferred scheme of the intelligent park infrastructure intelligent monitoring system, the feature values are fused through a fusion function, and the expression of the fusion function is:

[0021] ;

[0022] Wherein, Y is a feature data, x n is a feature value, w n is a fusion feature parameter coefficient, and the sum of each fusion feature parameter coefficient is 1.

[0023] The fusion feature parameter coefficient is optimized and matched according to the particle swarm algorithm.

[0024] As a preferred scheme of the intelligent park infrastructure intelligent monitoring system, the logic of optimizing and matching the fusion feature parameter coefficient includes:

[0025] Set the number of particle swarm, the range of particle motion speed, the number of iterations, the inertia coefficient, randomly set the optimal position, construct the target fitness function, calculate the target fitness function value of each particle, compare the target fitness function value of the particle with the corresponding best position, when the target fitness function value of the particle is less than the corresponding best position, set the target fitness function value of the particle as the new best position, when the target fitness function value of the particle is greater than or equal to the corresponding best position, the best position remains unchanged, until the number of iterations is met, the optimization is ended, and the optimal fusion feature parameter coefficient is obtained.

[0026] As a preferred scheme of the intelligent park infrastructure intelligent monitoring system, the calculation expression of the target fitness function is:

[0027] ;

[0028] Wherein, p is the target fitness function value, And The average value of the characteristic value in the normal state and the average value of the characteristic value in the abnormal state, respectively, And The standard deviation of the characteristic value in the normal state and the standard deviation of the characteristic value in the abnormal state, respectively, the smaller the p value, the higher the sensitivity of the fusion feature parameter to the state judgment.

[0029] As a preferred scheme of the intelligent monitoring system for the infrastructure of the smart park, wherein: the training logic of the state prediction model comprises:

[0030] The normal state is set to 1, the abnormal state is set to 0, the feature data is normalized to make the feature data distributed between 0 and 1, the normalized feature data is divided into a test set and a training set according to a first proportion, a decision tree is selected as a base model, and the decision tree includes a normal state decision tree and an abnormal state decision tree.

[0031] Set the decision tree parameters, including the learning rate, the maximum depth of the tree, the number of trees, and the minimum number of samples for node splitting, train the tree model one by one according to the set number of trees, calculate the residual error in each iteration, the residual error is represented as the negative gradient of the loss function and the current predicted value, train a new decision tree according to the current pseudo residual error, weight the predicted value of the newly trained tree with the initial estimate, update the initial estimate, and repeat the above training process until all decision trees are trained, then stop iteration;

[0032] The superposition combination of the normal state decision tree and the abnormal state decision tree after iteration is set as the state prediction model, and the state of the power supply facility is obtained by inputting the feature data into the state prediction model.

[0033] As a preferred scheme of the intelligent monitoring system for the infrastructure of the smart park, wherein: the optimization logic of the state prediction model comprises:

[0034] Set the case of correct prediction in normal state as TP, set the case of incorrect prediction in normal state as FP, set the case of correct prediction in abnormal state as TN, set the case of correct prediction in abnormal state as FN, calculate the precision, recall and overall accuracy of the state prediction model in normal state and abnormal state respectively, set the third value, the fourth value and the fifth value as the precision threshold, the recall threshold and the overall accuracy threshold respectively, compare the precision, the recall and the overall accuracy with the third value, the fourth value and the fifth value respectively, when the precision is less than or equal to the third value or the recall is less than or equal to the fourth value or the overall accuracy threshold is less than or equal to the fifth value, reset the decision tree parameters and retrain the decision tree to obtain the optimized state prediction model, when the precision is greater than the third value and the recall is greater than the fourth value and the overall accuracy threshold is greater than the fifth value, do not optimize the state prediction model.

[0035] As a preferred scheme of the intelligent park infrastructure intelligent monitoring system, the calculation expression of the precision is:

[0036] Precision=TP / TP+FP;

[0037] The calculation expression of the recall is:

[0038] Recall=TP / TP+FN;

[0039] The calculation expression of the overall accuracy is:

[0040] Accruracy=TP+TN / TP+TN+FP+FN;

[0041] Wherein, Precision, Recall and Accruracy represent the precision, the recall and the overall accuracy respectively.

[0042] The fusion module fuses the characteristic values of multiple state-related parameters, can quickly integrate information from different sensors or data sources, reduces the complexity and delay of data processing, thereby realizing real-time monitoring of the state of the power supply facility, reduces the need for manual intervention through automated data processing and model optimization, improves monitoring efficiency, reduces labor costs, fuses the characteristic values of multiple related parameters through the fusion module, can more comprehensively reflect the state of the power supply facility, avoids the limitations of a single parameter, thereby improving the accuracy of prediction, optimizes the state prediction model, can dynamically adjust model parameters according to actual operation data, further improves the prediction accuracy of the model, through the accurate state prediction model, the system can early warning potential failure risks, take timely measures for maintenance and repair, reduce the probability of failure, and improve the reliability and stability of the power supply facility. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 A basic flowchart of a smart park infrastructure intelligent monitoring system is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0044] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all embodiments.

[0045] Embodiments, with reference to Figure 1 For an embodiment of the present application, a smart park infrastructure intelligent monitoring system is provided, which includes a fusion module, a prediction module and an optimization module.

[0046] The fusion module sets state-related parameters, fuses the characteristic values of the state-related parameters, and obtains characteristic data.

[0047] The prediction module trains the decision tree according to the characteristic data to obtain a state prediction model.

[0048] The optimization module optimizes the state prediction model, and predicts the state of the power supply facility according to the optimized state prediction model.

[0049] The fusion module fuses characteristic values of multiple state-related parameters, can quickly integrate information from different sensors or data sources, reduces the complexity and delay of data processing, thereby realizing real-time monitoring of the state of the power supply facility, reduces the need for manual intervention through automated data processing and model optimization, improves monitoring efficiency, reduces labor costs, fuses the characteristic values of multiple related parameters through the fusion module, can more comprehensively reflect the state of the power supply facility, avoids the limitations of a single parameter, thereby improving the accuracy of prediction, optimizes the state prediction model, can dynamically adjust model parameters according to actual operation data, further improves the prediction accuracy of the model, through the accurate state prediction model, the system can early warning potential failure risks, timely take measures for maintenance and repair, reduce the probability of failure, improve the reliability and stability of the power supply facility.

[0050] The historical related data of the power supply facility includes historical voltage, historical current, historical power switch tube temperature, historical transformer temperature, historical voltage total harmonic distortion rate, historical current total harmonic distortion rate, and historical power supply facility state.

[0051] The historical power supply facility state includes a normal state and an abnormal state.

[0052] The number of times of exceeding the limit of historical voltage fluctuation is counted according to the historical voltage.

[0053] The historical voltage, the historical current, the historical power switch tube temperature, the historical transformer temperature, the historical voltage total harmonic distortion rate, and the historical current total harmonic distortion rate are set as state-related parameters.

[0054] The statistical logic of the number of times of exceeding the limit of historical voltage fluctuation includes:

[0055] The first difference value of any two adjacent historical voltages is calculated, the absolute value of the first difference value is selected, the first value is set as the voltage fluctuation threshold, the absolute value of the first difference value is compared with the first value, when the first value is less than or equal to the first value, the next group of adjacent historical voltages is jumped to, and the steps of difference and comparison are repeated, when the absolute value of the first difference value is greater than the first value, the number of times of exceeding the limit of historical voltage fluctuation is weighted by 1 to obtain a first sum value, the first sum value is used for the value of the absolute value of the first difference value greater than the first value weighted by 1 in the next time, and the first sum value is updated according to the new sum value obtained by each weighting.

[0056] In the embodiment, the change of the voltage can be accurately identified by calculating the first difference value of any two adjacent historical voltages, the first value is set as the voltage fluctuation threshold, the threshold is flexibly set according to actual needs, the absolute value of the first difference value is compared with the first value, when the absolute value of the first difference value is less than or equal to the first value, the next set of adjacent historical voltages is jumped to, the statistical efficiency is improved, the first sum value is updated according to the new sum value obtained by each weighting, the dynamicity and accuracy of the statistical result are ensured, the voltage abnormality can be found and handled in time by effectively monitoring and recording the voltage fluctuation out-of-limit event, the stability and reliability of the system are improved, and the law and trend of the voltage fluctuation are analyzed by the statistics of the historical voltage fluctuation out-of-limit times, data support is provided for the maintenance and optimization of the system.

[0057] The fusion module pre-processes the state-related parameters, and the pre-processing includes cleaning processing and abnormal value processing;

[0058] The cleaning processing is processed by a data cleaning program;

[0059] The logic of the abnormal value processing includes:

[0060] Any kind of state-related parameter is selected, the average value and the standard deviation of the state-related parameter are calculated, the first difference value of each state-related parameter and the average value of the state-related parameter is calculated, the value of three times of the standard deviation is set as a second value, the first difference value is compared with the second value, when the first difference value is greater than the second value, the state-related parameter is set as an abnormal value, the state-related parameter is removed and the next state-related parameter is jumped to, when the first difference value is less than or equal to the second value, the state-related parameter is set as a normal value, the state-related parameter is retained and the next state-related parameter is jumped to;

[0061] The average value of the pre-processed state-related parameter is set as the corresponding feature value.

[0062] The feature values are fused by a fusion function, and the expression of the fusion function is:

[0063] ;

[0064] Wherein, Y is the feature data, x n is the feature value, w n is the fusion feature parameter coefficient, and the sum value of each fusion feature parameter coefficient is 1.

[0065] The fusion feature parameter coefficient is optimized and matched according to the particle swarm algorithm.

[0066] In specific implementation, the state-related parameters are cleaned by a data cleaning program, which can remove noise and redundant information in the data, improve the neatness and consistency of the data, and effectively remove extreme values and error data in the data by calculating the average value and standard deviation of the state-related parameters and identifying and removing values exceeding 3 times the standard deviation as abnormal values, thereby reducing the influence of these data on subsequent analysis and modeling. The preprocessed data is more consistent with the assumption conditions of the model, which can make the model have better generalization ability on new data, reduce the risk of overfitting, and provide more accurate results for system monitoring and early warning, reduce false positives and false negatives, and improve user trust in the system. Through automatic identification and removal of abnormal values, the intelligence level of the system is embodied, which can automatically adapt to different data environments.

[0067] The logic of optimizing and matching the fusion feature parameter coefficient includes:

[0068] The number of particle swarm, the range of particle motion speed, the number of iterations, the inertia coefficient, the optimal position, the target fitness function, the target fitness function value of each particle, the comparison between the target fitness function value of the particle and the corresponding best position, the setting of the target fitness function value of the particle as the new best position when the target fitness function value of the particle is less than the corresponding best position, the unchanged best position when the target fitness function value of the particle is greater than or equal to the corresponding best position, the end of optimization when the number of iterations is met, and the optimal fusion feature parameter coefficient are obtained.

[0069] The calculation expression of the target fitness function is:

[0070] ;

[0071] Wherein, p is the target fitness function value, and are the average value of the characteristic value in the normal state and the average value of the characteristic value in the abnormal state, and are the standard deviation of the characteristic value in the normal state and the standard deviation of the characteristic value in the abnormal state, and the smaller the p value, the higher the sensitivity of the fusion feature parameter to state judgment.

[0072] In specific implementation, by setting appropriate particle group number, exploring multiple possible solution spaces, improving parameter optimization efficiency, limiting particle movement speed, ensuring effective movement of particles in solution space, avoiding too fast or too slow search speed affecting optimization efficiency, setting appropriate iteration number, ensuring algorithm has enough time to converge to optimal solution, randomly initializing optimal position of particles, increasing solution diversity, avoiding algorithm falling into local optimal solution, adjusting inertia coefficient, balancing global search and local search ability of particles, improving convergence speed and accuracy of algorithm, constructing target fitness function, quantifying solution quality of particles, guiding particles to move in optimal solution direction, particle swarm optimization algorithm automatically finds optimal fusion feature parameter coefficient, reducing complexity of manual parameter setting, particle swarm optimization algorithm is an intelligent optimization method, which can effectively find optimal solution by simulating foraging behavior of bird swarm, and reflects intelligent level of system.

[0073] The training logic of the state prediction model comprises:

[0074] The normal state is set to 1, the abnormal state is set to 0, the feature data is normalized to make the feature data distributed between 0 and 1, the normalized feature data is divided into a test set and a training set according to a first proportion, a decision tree is selected as a basic model, and the decision tree comprises a normal state decision tree and an abnormal state decision tree.

[0075] The decision tree parameters are set, the decision tree parameters comprise a learning rate, a maximum depth of the tree, a number of trees and a minimum number of samples for node splitting, the tree model is trained one by one according to the set number of trees, in each iteration, the residual error is calculated, the residual error is represented as a loss function and a negative gradient of a current prediction value, a new decision tree is trained according to the current pseudo residual error, the prediction value of the newly trained tree is weighted with the initial estimate, the initial estimate is updated, the above training process is looped, and when all the decision trees are trained, the iteration is stopped.

[0076] The superimposed combination of the normal state decision tree and the abnormal state decision tree after iteration is set as the state prediction model, and the state of the power supply facility is obtained by inputting the feature data into the state prediction model.

[0077] In specific implementation, the feature data is normalized to distribute between 0 and 1, which can eliminate the dimensional difference between different features, improve the training effect of the model, correct the prediction error of the model gradually by calculating the residual (negative gradient of the loss function and the current predicted value) and training a new decision tree according to the residual, improve the prediction accuracy of the model, capture the complex relationship in the data by superimposing and combining the prediction results of multiple decision trees, improve the generalization ability of the model, control the complexity of the model by reasonably setting the parameters of the decision tree (such as learning rate, maximum depth of the tree, number of trees and minimum number of samples for node splitting), improve the robustness of the model, and real-time monitor the state of the power supply facility by the state prediction model, which can timely discover and handle problems and improve user experience.

[0078] The logic for optimizing the state prediction model includes:

[0079] Set the case of correct prediction in normal state as TP, the case of incorrect prediction in normal state as FP, the case of correct prediction in abnormal state as TN, and the case of incorrect prediction in abnormal state as FN, respectively calculate the precision, recall and overall accuracy of the state prediction model for normal state and abnormal state, set the third value, the fourth value and the fifth value as the precision threshold, the recall threshold and the overall accuracy threshold, respectively compare the precision, recall and overall accuracy with the third value, the fourth value and the fifth value, when the precision is less than or equal to the third value or the recall is less than or equal to the fourth value, or the overall accuracy threshold is less than or equal to the fifth value, reset the parameters of the decision tree and retrain the decision tree to obtain the optimized state prediction model, when the precision is greater than the third value and the recall is greater than the fourth value, and the overall accuracy threshold is greater than the fifth value, do not optimize the state prediction model.

[0080] The calculation expression of the precision is:

[0081] Precision = TP / TP + FP;

[0082] The calculation expression of the recall is:

[0083] Recall = TP / TP + FN;

[0084] The calculation expression of the overall accuracy is:

[0085] Accruracy = TP + TN / TP + TN + FP + FN;

[0086] Wherein, Precision, Recall and Accruracy represent the precision, recall and overall accuracy, respectively.

[0087] In a specific implementation, by comparing the precision, recall and overall accuracy with the preset threshold (third value, fourth value and fifth value), it can be dynamically judged whether the model needs to be optimized. When the performance indicators of the model are lower than the preset threshold, the decision tree parameters (such as learning rate, maximum depth of tree, number of trees and minimum number of samples for node splitting) are reset, and the decision tree is retrained, which can effectively improve the performance of the model. This optimization logic can adaptively adjust the model parameters according to different data characteristics and application scenarios to ensure the best performance of the model under different conditions. By optimizing the model, the precision, recall and overall accuracy are all higher than the preset threshold, which can improve the prediction accuracy of the model for normal and abnormal states, reduce false positives and false negatives, improve the reliability of the system, and optimize the logic independent of other modules for design and optimization, facilitating the expansion and upgrade of the system.

[0088] The fusion module of the present application fuses the characteristic values of multiple state-related parameters, which can quickly integrate information from different sensors or data sources, reduce the complexity and delay of data processing, and thus realize real-time monitoring of the state of power supply facilities. Through automated data processing and model optimization, the need for manual intervention is reduced, the monitoring efficiency is improved, and the labor cost is reduced. By fusing the characteristic values of multiple related parameters through the fusion module, the state of the power supply facility can be more comprehensively reflected, avoiding the limitations of a single parameter, thereby improving the accuracy of the prediction. The state prediction model is optimized, which can dynamically adjust the model parameters according to the actual operation data, further improving the prediction accuracy of the model. Through the accurate state prediction model, the system can early warn potential failure risks, take timely measures for maintenance and repair, reduce the probability of failure, and improve the reliability and stability of the power supply facility.

[0089] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (or computer- readable storage media) having computer-usable program code embodied in the medium. The medium may Figure 1 one or more functions specified in the flow or flows and / or blocks Figure 1 one or more functions specified in the flow or flows and / or blocks

[0090] It should be noted that the above-mentioned embodiments are only used to illustrate but not to limit the technical solutions of the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A smart park infrastructure intelligent monitoring system, characterized in that, The fusion module, the prediction module and the optimization module are included. The fusion module sets state-related parameters, fuses characteristic values of the state-related parameters, and obtains characteristic data. The prediction module trains a decision tree according to the characteristic data, and obtains a state prediction model. The optimization module optimizes the state prediction model, and predicts the state of the power supply facility according to the optimized state prediction model. The historical related data of the power supply facility includes historical voltage, historical current, historical power switch tube temperature, historical transformer temperature, historical voltage total harmonic distortion rate, historical current total harmonic distortion rate and historical power supply facility state. The historical power supply facility state includes a normal state and an abnormal state. The number of times of exceeding the limit of historical voltage fluctuation is counted according to historical voltage. The historical voltage, the historical current, the historical power switch tube temperature, the historical transformer temperature, the historical voltage total harmonic distortion rate and the historical current total harmonic distortion rate are set as state-related parameters. The training logic of the state prediction model includes: The normal state is set as 1, the abnormal state is set as 0, the characteristic data is normalized so that the characteristic data is distributed between 0 and 1, the normalized characteristic data is divided into a test set and a training set according to a first proportion, a decision tree is selected as a basic model, and the decision tree includes a normal state decision tree and an abnormal state decision tree. The decision tree parameters are set, the decision tree parameters include a learning rate, a maximum depth of a tree, a number of trees and a minimum number of samples for node splitting, a tree model is trained one by one according to the set number of trees, in each iteration, a residual error is calculated, the residual error is represented as a loss function and a negative gradient of a current prediction value, a new decision tree is trained according to the current pseudo residual error, a prediction value of the newly trained tree is weighted with an initial estimate, the initial estimate is updated, the above training process is repeated until all decision trees are trained, and the iteration is stopped. The superposition combination of the normal state decision tree and the abnormal state decision tree after iteration is set as the state prediction model, the state of the power supply facility is obtained by inputting the characteristic data into the state prediction model. 2.The smart park infrastructure intelligent monitoring system of claim 1, wherein: The statistical logic of the number of times of exceeding the limit of historical voltage fluctuation includes: A first difference value of any two adjacent historical voltages is calculated, an absolute value of the first difference value is selected, a first value is set as a voltage fluctuation threshold, the absolute value of the first difference value is compared with the first value, when the first value is less than or equal to the first value, the next group of adjacent historical voltages is jumped to, and the steps of difference calculation and comparison are repeated, when the absolute value of the first difference value is greater than the first value, the number of times of exceeding the limit of historical voltage fluctuation is weighted with 1 to obtain a first sum value, the first sum value is used for weighting with 1 in the next time when the absolute value of the first difference value is greater than the first value, and the first sum value is updated according to the new sum value obtained by weighting each time. 3.The smart park infrastructure intelligent monitoring system of claim 1, wherein: The fusion module pre-processes the state-related parameters, and the pre-processing includes cleaning processing and abnormal value elimination processing. The cleaning processing is processed by a data cleaning program. The logic of the abnormal value processing includes: Select any kind of state related parameters, calculate the average value and standard deviation of the state related parameters, calculate the first difference value of each state related parameter and the average value of the state related parameters respectively, set the value of three times of the standard deviation as the second value, compare the first difference value with the second value, when the first difference value is greater than the second value, set the state related parameter as an abnormal value, eliminate the state related parameter and jump to the next state related parameter, when the first difference value is less than or equal to the second value, set the state related parameter as a normal value, retain the state related parameter and jump to the next state related parameter; The average value of the state related parameters after preprocessing is set as the corresponding feature value. 4.The smart park infrastructure intelligent monitoring system of claim 1, wherein: The feature values are fused through a fusion function, and the expression of the fusion function is: ； wherein Y is feature data, x n is a feature value, w n is a fusion feature parameter coefficient, and the sum of each fusion feature parameter coefficient is 1; The particle swarm algorithm is used to optimize and match the fusion feature parameter coefficients. 5.The smart park infrastructure intelligent monitoring system of claim 4, wherein: The logic of optimizing and matching the fusion feature parameter coefficients includes: The number of particle swarms, the range of particle motion speed, the number of iterations, the inertia coefficient, the optimal position is randomly set, the target fitness function is constructed, the target fitness function value of each particle is calculated, the target fitness function value of the particle is compared with the corresponding best position, when the target fitness function value of the particle is less than the corresponding best position, the target fitness function value of the particle is set as the new best position, when the target fitness function value of the particle is greater than or equal to the corresponding best position, the best position is unchanged, until the number of iterations is met, the optimization is ended, and the optimal fusion feature parameter coefficient is obtained. 6.The smart park infrastructure intelligent monitoring system of claim 5, wherein: The calculation expression of the target fitness function is: ； Wherein, p is the target fitness function value, and respectively, the average value of the characteristic value in the normal state and the average value of the characteristic value in the abnormal state, and respectively, the standard deviation of the characteristic value in the normal state and the standard deviation of the characteristic value in the abnormal state, the smaller the p value, the higher the sensitivity of the fusion characteristic parameter to the state judgment. 7.The smart park infrastructure intelligent monitoring system of claim 6, wherein: The logic of optimizing the state prediction model includes: The correct prediction in the normal state is set as TP, the incorrect prediction in the normal state is set as FP, the correct prediction in the abnormal state is set as TN, and the correct prediction in the abnormal state is set as FN, the precision, recall and overall accuracy of the state prediction model for normal state and abnormal state are calculated respectively, the third value, the fourth value and the fifth value are set as the precision threshold, the recall threshold and the overall accuracy threshold respectively, the precision, recall and overall accuracy are compared with the third value, the fourth value and the fifth value respectively, when the precision is less than or equal to the third value or the recall is less than or equal to the fourth value or the overall accuracy threshold is less than or equal to the fifth value, the decision tree parameters are reset, and the decision tree is retrained to obtain the optimized state prediction model, when the precision is greater than the third value and the recall is greater than the fourth value and the overall accuracy threshold is greater than the fifth value, the state prediction model is not optimized. 8.The smart park infrastructure intelligent monitoring system of claim 7, wherein: The calculation expression of the precision is: Precision=TP / (TP+FP); The calculation expression of the recall is: Recall=TP / (TP+FN); The calculation expression of the overall accuracy is: Accruracy=(TP+TN) / (TP+TN+FP+FN); Wherein, Precision, Recall and Accruracy represent the precision, recall and overall accuracy, respectively.

Images