Expressway electromechanical system abnormal state real-time monitoring method

Abstract

The invention specifically relates to a real-time monitoring method for an abnormal state of a highway electromechanical system. The method comprises the following steps: data acquisition: acquiring current data and illumination intensity data; data inspection: inspecting the current data and the illumination intensity data to remove error data; data processing: performing differential processing on the checked data to obtain an optimal differential order; and constructing and predicting a model. According to the invention, the collected current data and illumination intensity data are analyzed and processed to eliminate error data in the data transmission process, so that the monitoring accuracy error caused by the deviation of the predicted value obtained after the ARIMA is used for training is effectively avoided; therefore, the predicted value predicted by the real-time monitoring method for the abnormal state of the electromechanical system can more accurately reflect whether the electromechanical system is abnormal or not.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a real-time monitoring method for abnormal state of expressway electromechanical system. BACKGROUND

[0002] The abnormality of expressway electromechanical system may cause traffic accidents or damage to road facilities, which poses a threat to the safety of drivers and road users. By monitoring the abnormal state in real time, potential safety hazards can be found and handled in a timely manner, reducing the likelihood of accidents. The real-time monitoring of the abnormal state of the expressway electromechanical system is achieved by monitoring the abnormal data of the current data of the expressway electromechanical system. For example, application number CN202410634200.6, a real-time monitoring method for abnormal state of expressway electromechanical system, monitors whether there is an abnormality by comparing the predicted value and the actual value through analysis and comparison of current data and illumination intensity data.

[0003] However, during the collection of current data and illumination intensity data, the collected data is not error-removed. Due to the special environment in the tunnel, the transmission of signals will be affected, resulting in errors in the transmitted data. Once the erroneous data is input into the ARIMA model for training, the result of the training will be biased, affecting the accuracy of the monitoring of the abnormal state of the electromechanical system. Therefore, it is necessary to improve it to solve the above-mentioned problems. SUMMARY

[0004] The purpose of the present application is to solve the above-mentioned problems by providing a real-time monitoring method for abnormal state of expressway electromechanical system.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions: The real-time monitoring method for abnormal state of expressway electromechanical system comprises: Data acquisition: obtaining current data and illumination intensity data; Data verification: verifying the current data and the illumination intensity data to remove erroneous data; Data processing: differentiating the verified data to obtain the optimal differentiation order; Model construction and prediction: constructing an ARIMA prediction model to predict the current data of the expressway electromechanical system, and correcting the obtained predicted value to obtain a corrected predicted value for real-time monitoring of the abnormal state of the expressway electromechanical system.

[0006] Preferably, the data acquisition specifically comprises: A current sensor is installed at a position of a circuit load point in a highway electromechanical system, and current data is collected by the current sensor; A light intensity sensor is installed in a preset light intensity area range in a highway tunnel to collect light intensity data in the tunnel.

[0007] Preferably, the data verification specifically includes: The current data and the light intensity data are simultaneously acquired at preset time intervals, and the simultaneously acquired current data and light intensity data are taken as a verification data set; According to the positions of the current sensor and the light intensity sensor, the tunnel interior environment is regionally divided at a preset area, and a verification data set of each region is acquired; The current data in each verification data set in the region is extracted, the mean value of the current data is calculated, and the current deviation value is obtained by taking the absolute value after the difference between each current value data and the mean value of the current is calculated. The current deviation value that is not within the preset allowable fluctuation range of the current is taken as a marked current deviation value, and the light intensity data in the verification data set where the marked current deviation value is located is marked as a marked light intensity deviation value; The light intensity data in each verification data set in the region is extracted, the mean value of the light intensity data is calculated, and the light intensity deviation value is obtained by taking the absolute value after the difference between each light intensity value data and the mean value of the light intensity is calculated. The light intensity deviation value that is not within the preset allowable fluctuation range of the light intensity is taken as a marked light intensity deviation value, and the current data in the verification data set where the marked light intensity deviation value is located is marked as a marked current deviation value; All the marked current deviation values and the marked light intensity deviation values are arranged in descending order according to the numerical value, and the three largest marked current deviation values and the three largest marked light intensity deviation values are extracted; The positions of the current sensors corresponding to the three largest marked current deviation values are taken as the three vertices of a triangle, and the three vertices are connected by straight lines to obtain a complete triangular model, which is taken as a current pseudo-shape; The positions of the light intensity sensors corresponding to the three largest marked light intensity deviation values are taken as the three vertices of a triangle, and the three vertices are connected by straight lines to obtain a complete triangular model, which is taken as a light intensity pseudo-shape; The overlapping area of the current pseudo-shape and the light intensity pseudo-shape is obtained, which is taken as an abnormal similarity value; The current pseudo-shape and the light intensity pseudo-shape are analyzed to obtain a reference ratio; The reference ratio of each verification data set is multiplied by the abnormal similarity value to obtain a post-processing similarity value of each verification data set; performing Pearson correlation analysis on the current data and the illumination intensity data in the test data set to obtain a correlation value; performing comprehensive processing on the post-processing similarity value and the correlation value to obtain a data evaluation coefficient of the test data set; Specifically, after performing normalization processing on the post-processing similarity value and the correlation value, the post-processing similarity value and the correlation value are respectively taken as the long semi-axis and the short semi-axis of an ellipse to establish an ellipse model, and the area of the ellipse model is calculated and recorded as the data evaluation coefficient. A preset data evaluation coefficient threshold is provided, and the data evaluation coefficient is compared with the data evaluation coefficient threshold. If the data evaluation coefficient is greater than the data evaluation coefficient threshold, the current data and the illumination intensity data in the test data corresponding to the data evaluation coefficient are removed.

[0008] Preferably, the process of obtaining the correlation value comprises: All data test sets corresponding to the marker current deviation value and the marker light intensity deviation value are arranged according to the time sequence; The current data and the illumination intensity data in the data test set are subjected to mean value calculation to obtain current mean value and light intensity mean value respectively; The covariance is calculated through the current mean value and the light intensity mean value, and the standard deviations of the current data and the illumination intensity data are calculated respectively; The calculated covariance is divided by the product of the standard deviation of the current data and the standard deviation of the illumination intensity data to obtain the Pearson correlation coefficient. The Pearson correlation coefficient is added by 1, and the reciprocal of the result is taken to obtain the correlation value.

[0009] Preferably, the test data is subjected to difference processing to obtain an optimal difference order, which specifically comprises: The current data is subjected to multiple difference processing to obtain a plurality of current difference data, and the illumination intensity data is subjected to multiple difference processing to obtain a plurality of illumination intensity difference data. According to the similarity of the fluctuation and stable distribution of each current difference data and the same order of illumination intensity difference data, the current-light correlation of each current difference data is obtained. Under a preset each lag parameter, each current difference data is subjected to lag processing to obtain lag current difference data under each lag parameter. According to the obtaining method of the current-light correlation of each current difference data, the current-light correlation of the lag current difference data of each current difference data under each lag parameter is obtained. According to the difference between the current-light correlation of each current difference data and the current-light correlation of the lag current difference data under each lag parameter, the autocorrelation representation degree of each current difference data under each lag parameter is obtained. The degree of autocorrelation of each lag parameter is analyzed to obtain a differential evaluation parameter of each current differential data; and the differential order corresponding to the maximum current differential data in the differential evaluation parameter is recorded as an optimal differential order.

[0010] Preferably, the process of obtaining the current-illumination correlation of the current differential data comprises: The current differential data and the illumination intensity differential data are respectively averaged and segmented to obtain segmented data of the current differential data and the illumination intensity differential data. The current-illumination correlation of each current differential data is obtained according to the Pearson correlation coefficient of each current differential data and each illumination intensity differential data, the number of segments of each current differential data and each illumination intensity differential data, the standard deviation of each segment of each current differential data, and the standard deviation of the same segment of the same illumination intensity differential data. The degree of autocorrelation of each lag parameter is analyzed to obtain a differential evaluation parameter of each current differential data; and the differential order corresponding to the maximum current differential data in the differential evaluation parameter is recorded as an optimal differential order. The degree of autocorrelation of each lag parameter is analyzed to obtain a differential evaluation parameter of each current differential data; and the differential order corresponding to the maximum current differential data in the differential evaluation parameter is recorded as an optimal differential order. The number of data segments is a preset number. The differential evaluation parameter of each current differential data is obtained according to the absolute value of the mean of the degree of autocorrelation of the end data points, the number of data segments of each current differential data, the mean of the degree of autocorrelation of each segment of each current differential data, and the mean of the slope of each segment of each current differential data.

[0011] Preferably, the process of obtaining the degree of autocorrelation of the end data points comprises the following specific steps: The last a differential data points in the autocorrelation graph of each current differential data are recorded as the end data points in the autocorrelation graph of each current differential data, and a is a preset number of end data points; and the degree of autocorrelation of the end data points in the autocorrelation graph of each current differential data is obtained.

[0012] Preferably, the mean of the slope of each segment comprises the following specific steps: For each segment of the autocorrelation graph of each current differential data, the ratio of the absolute value of the difference between each autocorrelation degree and the autocorrelation degree after it to the difference between the corresponding lag parameter and the lag parameter after it is recorded as the slope of each autocorrelation degree; and the slope of each segment is obtained by accumulating the slopes of all autocorrelation degrees in the segment and dividing by the number of lag parameters in the segment.

[0013] Preferably, the model construction and prediction specifically comprises: The current data and the illumination intensity data after inspection are divided into a training set, a verification set and a test set according to a certain proportion; According to the determined optimal differential order, an ARIMA model is constructed; The ARIMA model is trained by using the training set data, and the parameters of the model are determined by minimizing the prediction error; The trained ARIMA model is used to preliminarily predict the data in the verification set and the test set, and a preliminary prediction result is obtained; the ARIMA model preliminary prediction result is normalized and combined with the illumination intensity data and the time characteristics to form new data, and after the new data is normalized, the new data is input into a deep learning model for training; The trained deep learning model is used to correct the preliminary prediction result of the ARIMA model on the test set, and a final prediction value is obtained.

[0014] A preset current fault warning standard value and an abnormal data threshold value are set; When the difference between the predicted value and the actual value of the current data exceeds the preset current fault warning standard value, the actual value is recorded as abnormal data; when the number of continuous abnormal data in the current data is greater than the preset abnormal data threshold value, the highway electromechanical system is alarmed.

[0015] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present application are: 1. The current data and the illumination intensity data collected are analyzed and processed to eliminate error data in the data transmission process, effectively avoiding the deviation of the prediction value obtained after using ARIMA for training, thereby avoiding errors in monitoring accuracy, so that the prediction value predicted by the electromechanical system abnormal state real-time monitoring method can more accurately reflect whether the electromechanical system is abnormal.

[0016] 2. The present application combines the ARIMA prediction model with the deep learning model to form a hybrid prediction system, the ARIMA model is good at processing short-term trends and periodicity of time series data, and the deep learning model has strong nonlinear feature extraction and complex pattern recognition ability; first, the ARIMA model is used to preliminarily predict the current data, and then the prediction result is integrated with the illumination intensity data and the time characteristics and input into the deep learning model for correction, so as to realize the complementary advantages and significantly improve the prediction accuracy of the current data of the highway electromechanical system. BRIEF DESCRIPTION OF DRAWINGS

[0017] In the following description of exemplary embodiments in conjunction with the drawings, more details, features and advantages of the present application are disclosed, and in the drawings: Figure 1 Flowchart of the present application; DETAILED DESCRIPTION

[0018] Several embodiments of the application will be described in greater detail below, with reference to the figures, to enable one of ordinary skill in the art to implement the application. The application can be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. These embodiments are provided so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art. The embodiments are not intended to limit the application.

[0019] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and / or the present disclosure and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0020] Referring to the drawings Figure 1 As shown, the application provides a technical solution: The real-time monitoring method for abnormal state of expressway electromechanical system comprises: Data acquisition: obtaining current data and illumination intensity data; Specifically comprising: A current sensor is installed at the position of the circuit load point in the expressway electromechanical system, and the current sensor is used to collect current data; A light intensity sensor is installed in the preset illumination intensity area range in the expressway tunnel to collect the illumination intensity data in the tunnel; Data verification: verifying the current data and the illumination intensity data to remove error data; Specifically comprising: The current data and the illumination intensity data are obtained at the same time at the preset time interval, and the current data and the illumination intensity data obtained at the same time are taken as a verification data group; According to the positions of the current sensor and the light intensity sensor, the internal environment of the tunnel is divided into areas at a preset area, and the verification data group of each area is obtained; The current data in each verification data group in the area is extracted, the mean value of the area current is obtained after mean value calculation, each current value data is sequentially subjected to difference calculation with the mean value of the area current, the absolute value of the current deviation value is obtained after taking the absolute value, the current deviation value not in the preset allowable fluctuation range of the current is taken as a marked current deviation value, and the illumination intensity data in the verification data group where the marked current deviation value is located is marked as a marked light intensity deviation value; The illumination data in each test data group in the extraction area is subjected to mean value calculation to obtain an area illumination mean value; each illumination value data is sequentially subjected to difference calculation with the area illumination mean value to obtain illumination deviation values by taking absolute values; a preset illumination allowable fluctuation range is set, illumination deviation values not within the preset illumination allowable fluctuation range are recorded as marked illumination deviation values, and current data in the test data group where the marked illumination deviation values are located are recorded as marked current deviation values; All the marked current deviation values and the marked illumination deviation values are respectively arranged in descending order according to numerical values, and the three largest marked current deviation values and the three largest marked illumination deviation values are extracted; The current sensor positions corresponding to the three largest marked current deviation values are taken as three vertices of a triangle, and the three vertices are connected by straight lines to obtain a complete triangle model, which is recorded as a current pseudo-shape; The illumination sensor positions corresponding to the three largest marked illumination deviation values are taken as three vertices of a triangle, and the three vertices are connected by straight lines to obtain a complete triangle model, which is recorded as an illumination pseudo-shape; An overlapping area of the current pseudo-shape and the illumination pseudo-shape is obtained, which is recorded as an abnormal similarity value; The triangle center points of the current pseudo-shape and the illumination pseudo-shape are respectively obtained, the distance between the triangle center of the current pseudo-shape and any triangle vertex is calculated, and the marked current deviation value corresponding to the triangle vertex is obtained; the distance between the triangle center and the vertex of the illumination pseudo-shape is calculated, and the distance between the triangle vertex and the center is summed to obtain a reference value; All the reference values are sequentially obtained, arranged in ascending order according to the numerical values of the reference values, and the smallest reference value is recorded as a standard reference value; the remaining reference values are sequentially divided by the standard reference value to obtain a reference ratio of each test data group; The reference ratio of each test data group is sequentially multiplied by the abnormal similarity value to obtain a post-processing similarity value of each test data group; Pearson correlation analysis is performed on the current data and the illumination intensity data in the test data group to obtain a correlation value; The correlation value obtaining process includes: All the data test groups corresponding to the marked current deviation values and the marked illumination deviation values are arranged according to time sequence; The current data and the illumination intensity data in the data test group are subjected to mean value calculation to obtain current mean value and illumination mean value respectively; The covariance is calculated through the current mean value and the illumination mean value, and the standard deviations of the current data and the illumination intensity data are calculated respectively; The calculated covariance is divided by the product of the standard deviation of the current data and the standard deviation of the light intensity data to obtain a Pearson correlation coefficient, and the obtained Pearson correlation coefficient is added by 1 and then the reciprocal is obtained to obtain a correlation value; The covariance formula is: ; Wherein, I is the current data, and L is the light intensity data. is the current mean value, is the light intensity mean value; represents the current value at the kth time point, represents the light intensity value at the kth time point; n is the total number of data test groups; The data evaluation coefficient of the test data group is obtained by comprehensively processing the post-processed similarity value and the correlation value; After the post-processed similarity value and the correlation value are normalized, the post-processed similarity value and the correlation value are respectively taken as the long semi-axis and the short semi-axis of the ellipse to establish an elliptical model, the area of the elliptical model is calculated, and is recorded as the data evaluation coefficient; A preset data evaluation coefficient threshold is compared with the data evaluation coefficient, if the data evaluation coefficient is greater than the data evaluation coefficient threshold, the current data and the light intensity data in the test data corresponding to the data evaluation coefficient are removed; The test data groups of each region are analyzed in turn to remove the error data in the test data groups of each region; Data processing: the test data is subjected to difference processing to obtain the optimal difference order; Specifically, it includes: The current data is subjected to multiple difference to obtain a plurality of current difference data, and the light intensity data is subjected to multiple difference to obtain a plurality of light intensity difference data, and the current-light correlation of each current difference data is obtained according to the similarity of the fluctuation and stable distribution of each current difference data and the same light intensity difference data; Under each preset lag parameter, each current difference data is subjected to lag to obtain lag current difference data under each lag parameter, and the current-light correlation of the lag current difference data of each current difference data under each lag parameter is obtained according to the obtaining mode of the current-light correlation of each current difference data; According to the difference between the current-light correlation of each current difference data and the current-light correlation of the lag current difference data under each lag parameter, the autocorrelation representation degree of each current difference data under each lag parameter is obtained. constructing an autocorrelation graph of the current difference data according to the autocorrelation representation degree of each current difference data under each lag parameter, and obtaining a difference optimization parameter of the current difference data according to the decline amplitude of the autocorrelation representation degree in the autocorrelation graph of the current difference data with the growth of the lag parameter; and recording the difference order corresponding to the current difference data with the largest difference optimization parameter in the difference optimization parameter as an optimal difference order; By multiple times of differentiating the current data and the illumination intensity data, and according to the fluctuation stable distribution similarity of each current difference data and the same time illumination intensity difference data, and the autocorrelation representation degree under different lag parameters, an autocorrelation graph is constructed and a difference optimization parameter is calculated, so as to determine the optimal difference order. This method of determining the difference order according to the characteristics of the data itself without relying on empirical values can better adapt to complex input current characteristics and improve the accuracy of data processing. The process of obtaining the current-illumination correlation of the current difference data includes: The current difference data and the illumination intensity difference data are respectively averaged and segmented to obtain segmented data of the current difference data and the illumination intensity difference data. The current-illumination correlation of each current difference data is obtained according to the Pearson correlation coefficient of each current difference data and each illumination intensity difference data, the number of segments of each current difference data and each illumination intensity difference data, the standard deviation of each segment data in each current difference data, and the standard deviation of the same segment data in the same time illumination intensity difference data. The formula used for the current-illumination correlation of the mth current difference data is: ; Wherein represents the current-illumination correlation of the mth current difference data. represents the Pearson correlation coefficient of the mth current difference data and the mth illumination intensity difference data. represents the number of segments of the mth current difference data and the mth illumination intensity difference data. represents the standard deviation of the i-th segment data in the mth current difference data. represents the standard deviation of the i-th segment data in the mth illumination intensity difference data. is a linear normalization function. represents the absolute value function. The method for obtaining the autocorrelation representation degree of each current difference data under each lag parameter is: The last n current data are intercepted from the last current data of each current difference data to delete, and the sequence of the current data remaining after the deletion of each current difference data is recorded as the lag current difference data of each current difference data under the lag parameter n, and n is a preset lag parameter; According to the current light correlation of each current difference data, the current light correlation of the lag current difference data of each current difference data under each lag parameter, and the DTW distance between each current difference data and the lag current difference data of each current difference data under each lag parameter, the autocorrelation representation degree of each current difference data under each lag parameter is obtained; The specific calculation formula of the autocorrelation representation degree of the mth current difference data under the lag parameter n is: ; The autocorrelation representation degree of the mth current difference data under the lag parameter n is obtained ; The current light correlation of the mth current difference data is represented. The current light correlation of the lag current difference data of the mth current difference data under the lag parameter n is represented. The mth current difference data is represented. The lag current difference data of the mth current difference data under the lag parameter n is represented. The DTW distance between the mth current difference data and the lag current difference data of the mth current difference data under the lag parameter n is represented. The exponential function with a natural constant as the base is represented. The acquisition method of the difference evaluation parameter of the current difference data is: The autocorrelation representation degree of each current difference data under each lag parameter is mapped into a two-dimensional image to obtain the autocorrelation graph of each current difference data, and the horizontal coordinate of the autocorrelation graph of each current difference data is the lag parameter, and the vertical coordinate is the autocorrelation representation degree. The autocorrelation representation degree in the autocorrelation graph of each current difference data is divided into segments according to the size of the lag parameter, and the autocorrelation representation degree of each segmented data is obtained, is a preset number of data segments; The absolute value of the mean of the autocorrelation representation degree of the end data point in the autocorrelation graph of each current difference data, the number of data segments of each current difference data, the mean of the autocorrelation representation degree of each segment data in each current difference data, and the mean of the slope of each segment data in each current difference data are used to obtain a difference evaluation parameter of each current difference data; The difference evaluation parameter of the mth current difference data is as follows: The difference evaluation parameter of the mth current difference data is obtained by the formula The absolute value of the mean of the autocorrelation representation degree of the end data point in the autocorrelation graph of the mth current difference data is represented by The number of data segments of the mth current difference data is represented by The mean of the autocorrelation representation degree of the jth segment data in the mth current difference data is represented by The mean of the autocorrelation representation degree of the j+1th segment data in the mth current difference data is represented by The mean of the slope of the jth segment data in the mth current difference data is represented by The mean of the slope of the j+1th segment data in the mth current difference data is represented by is a linear normalization function; is an absolute value function; e is a natural constant; The autocorrelation representation degree of the end data point is obtained by the following specific steps: The last a difference data points in the autocorrelation graph of each current difference data are recorded as the end data points in the autocorrelation graph of each current difference data, and a is a preset number of end data points; the autocorrelation representation degree of the end data point in the autocorrelation graph of each current difference data is obtained. The mean of the slope of each segment data is obtained by the following specific steps: For each segment data in the autocorrelation graph of each current difference data, the ratio of the absolute value of the difference between each autocorrelation representation degree and the autocorrelation representation degree behind it to the difference between the corresponding lag parameter is recorded as the slope of each autocorrelation representation degree; the slopes of all autocorrelation representation degrees in the segment data are accumulated, and the ratio of the accumulated value to the number of lag parameters in the segment data is obtained as the mean of the slope of each segment data. ​​​Model construction and prediction: An ARIMA prediction model is constructed to predict the current data of the expressway electromechanical system, and the obtained prediction value is corrected to obtain a corrected prediction value for real-time monitoring of the abnormal state of the expressway electromechanical system. Specifically, it comprises: The tested current data and the illumination intensity data are divided into a training set, a validation set and a test set according to a certain proportion; For example, according to a proportion of 70%, 15% and 15%, the training set is used to train the ARIMA model and the deep learning model, the validation set is used to adjust the model hyperparameters, and the test set is used to evaluate the performance of the final model; According to the determined optimal difference order, an ARIMA (p, d, q) model is constructed, where p is the autoregressive order, d is the difference order, and q is the moving average order; The training set data is used to train the ARIMA model, and the parameters of the model are determined by minimizing the prediction error (such as mean square error MSE); The trained ARIMA model is used to preliminarily predict the validation set and test set data, and the preliminary prediction results are obtained; the ARIMA model preliminary prediction results are normalized and combined with the illumination intensity data and time characteristics to form new data, and the new data is input into the deep learning model for training after being normalized; The trained deep learning model is used to correct the preliminary prediction results of the ARIMA model on the test set to obtain the final prediction value.

[0021] A preset current fault warning standard value and an abnormal data threshold value are set; When the difference between the current data prediction value and the actual value exceeds the preset current fault warning standard value, the actual value is recorded as abnormal data; when the number of continuous abnormal data in the current data is greater than the preset abnormal data threshold value, a voice alarm is given to the expressway electromechanical system to realize real-time monitoring of the abnormal state of the expressway electromechanical system; The ARIMA prediction model is combined with the deep learning model for prediction, the ARIMA model is used for preliminary prediction, and then the results are combined with the illumination intensity data and time characteristics to form new data, which is input into the deep learning model for training and correction. This combination of traditional time series models and deep learning models fully utilizes the advantages of both and can more accurately predict the current data of the expressway electromechanical system.

[0022] The above formulas are obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value, and the influence weight factor and specific coefficient value in the formula are set by the person skilled in the art according to the actual situation, which can be adjusted and modified later.

[0023] The foregoing description of the embodiments enables one of ordinary skill in the art to make or use the application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for real-time monitoring of abnormal states of a highway electromechanical system, characterized in that, The method comprises the following steps: Data collection: install current sensors at the positions of circuit load points in the highway electromechanical system, and collect current data by using the current sensors; Install light intensity sensors in the preset light intensity area range in the highway tunnel to collect the light intensity data in the tunnel; Data verification: verify the current data and the light intensity data to remove the error data; Data processing: perform differential processing on the verified data to obtain the optimal differential order; Model construction and prediction: construct an ARIMA prediction model to predict the current data of the highway electromechanical system, and correct the obtained prediction value to obtain a corrected prediction value, so as to perform real-time monitoring on the abnormal state of the highway electromechanical system.

2. The method of claim 1, wherein, The data verification specifically comprises the following steps: Simultaneously acquire the current data and the light intensity data at preset time intervals, and take the simultaneously acquired current data and light intensity data as verification data groups; Divide the internal environment of the tunnel into regions according to the positions of the current sensors and the light intensity sensors, and acquire the verification data groups of each region; Extract the current data in each verification data group in the region, perform mean value calculation to obtain the regional current mean value, sequentially perform difference calculation on each current value data and the regional current mean value, take the absolute value to obtain the current deviation value, preset the allowable fluctuation range of the current, take the current deviation value not in the preset allowable fluctuation range of the current as a marked current deviation value, and mark the light intensity data in the verification data group where the marked current deviation value is located as a marked light intensity deviation value; Extract the light intensity data in each verification data group in the region, perform mean value calculation to obtain the regional light intensity mean value, sequentially perform difference calculation on each light intensity value data and the regional light intensity mean value, take the absolute value to obtain the light intensity deviation value, preset the allowable fluctuation range of the light intensity, take the light intensity deviation value not in the preset allowable fluctuation range of the light intensity as a marked light intensity deviation value, and mark the current data in the verification data group where the marked light intensity deviation value is located as a marked current deviation value; Respectively arrange all the marked current deviation values and the marked light intensity deviation values in descending order according to the numerical values, and extract the three largest marked current deviation values and the three largest marked light intensity deviation values; Take the positions of the current sensors corresponding to the three largest marked current deviation values as the three vertices of a triangle, connect the three vertices with straight lines to obtain a complete triangular model, and mark the triangular model as a current pseudo-shape; Take the positions of the light intensity sensors corresponding to the three largest marked light intensity deviation values as the three vertices of a triangle, connect the three vertices with straight lines to obtain a complete triangular model, and mark the triangular model as a light intensity pseudo-shape; Obtain the overlapping area of the current pseudo-shape and the light intensity pseudo-shape, and mark the overlapping area as an abnormal similarity value; Analyze the current pseudo-shape and the light intensity pseudo-shape to obtain a reference ratio; Multiply the reference ratio of each verification data group by the abnormal similarity value to obtain the post-processing similarity value of each verification data group; Perform Pearson correlation analysis on the current data and the light intensity data in the verification data group to obtain a correlation value; Perform comprehensive processing on the post-processing similarity value and the correlation value to obtain a data evaluation coefficient of the verification data group; Specifically comprising: after normalizing the post-processing similarity value and the correlation value, the post-processing similarity value and the correlation value are respectively taken as the long semi-axis and the short semi-axis of an ellipse to establish an ellipse model, and the area of the ellipse model is calculated and recorded as a data evaluation coefficient; A preset data evaluation coefficient threshold is set, and the data evaluation coefficient is compared with the data evaluation coefficient threshold. If the data evaluation coefficient is greater than the data evaluation coefficient threshold, the internal current data and the illumination intensity data corresponding to the data evaluation coefficient are removed.

3. The method of claim 2, wherein the method further comprises: The process of obtaining the correlation value includes: All data test groups corresponding to the marker current deviation value and the marker light intensity deviation value are arranged according to the time sequence; The current data and the illumination intensity data in the data test group are subjected to mean value calculation to obtain current mean value and light intensity mean value respectively; The covariance is calculated through the current mean value and the light intensity mean value, and the standard deviation of the current data and the illumination intensity data is calculated respectively; The calculated covariance is divided by the product of the standard deviation of the current data and the standard deviation of the illumination intensity data to obtain the Pearson correlation coefficient, and the inverse of the Pearson correlation coefficient obtained by adding 1 is taken to obtain the correlation value.

4. The method of claim 3, wherein the method further comprises: The test data is subjected to difference processing to obtain an optimal difference order, specifically comprising: The current data is subjected to multiple difference to obtain a plurality of current difference data, and the illumination intensity data is subjected to multiple difference to obtain a plurality of illumination intensity difference data. According to the similarity of the fluctuation and stable distribution of each current difference data and the same order illumination intensity difference data, the current-light correlation of each current difference data is obtained. Under each preset lag parameter, the lag of each current difference data is obtained to obtain the lag current difference data under each lag parameter. According to the obtaining method of the current-light correlation of each current difference data, the current-light correlation of the lag current difference data of each current difference data under each lag parameter is obtained. According to the difference between the current-light correlation of each current difference data and the current-light correlation of the lag current difference data under each lag parameter, the autocorrelation representation degree of each current difference data under each lag parameter is obtained. The autocorrelation representation degree under each lag parameter is analyzed to obtain the difference optimization parameter of each current difference data, specifically comprising: The autocorrelation representation degree of each current difference data under each lag parameter is mapped into a two-dimensional image to obtain the autocorrelation graph of each current difference data. The horizontal coordinate of the autocorrelation graph of each current difference data is the lag parameter, and the vertical coordinate is the autocorrelation representation degree. The degree of autocorrelation representation in the autocorrelation graph of each current difference data is divided into segments according to the size of the lag parameter, and the degree of autocorrelation representation of each segmented data is obtained, is a preset data segment number; According to the absolute value of the mean value of the autocorrelation representation degree of the end data point in the autocorrelation graph of each current difference data, the number of data segments of each current difference data, the mean value of the autocorrelation representation degree of each segment data in each current difference data, and the mean value of the slope of each segment data in each current difference data, the difference optimization parameter of each current difference data is obtained. The difference order corresponding to the current difference data with the largest difference optimization parameter in the difference optimization parameter is recorded as the optimal difference order.

5. The method of claim 4, wherein the method further comprises: The process of obtaining the current-light correlation of the current difference data includes: The current difference data and the illumination intensity difference data are respectively averaged and segmented to obtain segmented data of the current difference data and the illumination intensity difference data; The current-illumination correlation of each current difference data is obtained according to the Pearson correlation coefficient of each current difference data and each illumination intensity difference data, the number of segments of each current difference data and each illumination intensity difference data, the standard deviation of each segment data in each current difference data, and the standard deviation of the same segment data in the same illumination intensity difference data; The degree of autocorrelation representation in the autocorrelation graph of each current difference data is divided into segments, obtain several segmented data, and obtain the degree of autocorrelation representation of each segmented data, is a preset data segment number; The difference evaluation parameter of each current difference data is obtained according to the absolute value of the average of the autocorrelation representation degrees of the end data points in the autocorrelation graph of each current difference data, the number of data segments of each current difference data, the average of the autocorrelation representation degrees of each segment data in each current difference data, and the average of the slopes of each segment data in each current difference data.

6. The method of claim 5, wherein the method further comprises: The autocorrelation representation degrees of the end data points are obtained, including the following specific steps: The last a difference data points in the autocorrelation graph of each current difference data are recorded as the end data points in the autocorrelation graph of each current difference data, and a is a preset number of end data points; The autocorrelation representation degrees of the end data points in the autocorrelation graph of each current difference data are obtained.

7. The method of claim 5, wherein the method further comprises: The average of the slopes of each segment data includes the following specific steps: For each segment data in the autocorrelation graph of each current difference data, the ratio of the absolute value of the difference between each autocorrelation representation degree and the autocorrelation representation degree of the next segment data to the difference between the corresponding lag parameter is recorded as the slope of each autocorrelation representation degree; the slopes of all autocorrelation representation degrees in the segment data are accumulated, and the average of the slopes of each segment data is obtained by comparing the number of lag parameters in the segment data.

8. The method of claim 7, wherein the method further comprises: Model construction and prediction, specifically including: The current data and the illumination intensity data after the test are divided into a training set, a validation set and a test set according to a certain proportion; An ARIMA model is constructed according to the determined optimal difference order; The ARIMA model is trained using the training set data, and the parameters of the model are determined by minimizing the prediction error; The trained ARIMA model is used to preliminarily predict the validation set and the test set data to obtain a preliminary prediction result; the ARIMA model preliminary prediction result is combined with the illumination intensity data and the time characteristics after normalization processing to form new data, and the new data is input into a deep learning model after normalization processing for training; The trained deep learning model is used to correct the preliminary prediction result of the ARIMA model on the test set to obtain the final prediction value; A preset current fault warning standard value and an abnormal data threshold value are set; When the difference between the predicted value and the actual value of the current data exceeds the preset current fault warning standard value, the actual value is recorded as abnormal data; when the number of continuous abnormal data in the current data is greater than the preset abnormal data threshold value, the highway electromechanical system is alarmed.

Images