Looped network box with overload fault monitoring and circuit breaking functions

Abstract

The invention relates to the technical field of power load monitoring, in particular to a ring main unit with an overload fault monitoring and circuit breaking function. Obtaining a load filtering time sequence according to the instantaneous fluctuation characteristics of the load time sequence; constructing a load feature vector according to the load filtering time sequence; constructing a temperature feature vector according to the temperature time sequence; obtaining a temperature difference degree and a date cluster according to the temperature feature vector; obtaining a load difference degree according to the load feature vector; constructing a difference scatter diagram and a load difference prediction model according to the temperature difference degree and the load difference degree corresponding to the date cluster; and obtaining a difference degree weight according to the scattered point distribution characteristics and training a load difference prediction model. Obtaining a load prediction difference degree according to the temperature feature vector of the preset current time period; obtaining a deviation index according to the load prediction difference degree, the actual load difference degree and the model prediction error characteristics; and the load state is monitored according to the change characteristics of the deviation index, so that the load monitoring accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power load monitoring, and particularly relates to a ring main unit with an overload fault monitoring and opening function. Background Art

[0002] In the medium-voltage distribution network, to meet the power supply reliability and safety requirements of power supply areas such as urban high-rise buildings, communities, factories, and commercial centers, the ring main distribution system is widely used as a typical distribution solution. A ring main unit with an overload fault monitoring and opening function is a medium-voltage distribution device integrating advanced sensors and intelligent control technologies. Its main function is to monitor the load status of different circuits in real time in the ring main distribution system. When an overload or other fault occurs, it can quickly judge and disconnect the faulty line to improve power supply reliability.

[0003] The existing power grid overload fault monitoring mainly relies on real-time comparison of the operating status of equipment with the rated reference value. Once the load exceeds the rated value, the ring main distribution system will activate the warning and protection mechanism; however, this method ignores the dynamic changes and laws of the power grid load in the actual scenario. For example, instantaneous load fluctuations such as the startup of large equipment and arc welding work will be misjudged as overload risks, triggering false alarms or unnecessary protection actions; this affects the accuracy of the load monitoring and control of the ring main unit in the power grid. Summary of the Invention

[0004] In order to solve the above technical problem that the existing load monitoring method ignores the dynamic changes and laws of the power grid load in the actual scenario, which affects the accuracy of the load monitoring and control of the ring main unit in the power grid, the purpose of the present invention is to provide a ring main unit with an overload fault monitoring and opening function, and the specific technical solutions adopted are as follows: A data acquisition module, used to acquire the load time series of the circuit and the temperature time series of the environment; A feature analysis module, used to perform adaptive filtering according to the instantaneous fluctuation characteristics of the load time series to obtain a load filtered time series; construct a load feature vector according to the data trend characteristics of a preset current period in the load filtered time series; construct a temperature feature vector according to the data trend characteristics of a preset current period in the temperature time series; A feature processing module, used to obtain different temperature difference degrees and date clusters according to the difference characteristics of the temperature feature vectors in the same historical period; obtain a load difference degree according to the difference characteristics of the load feature vectors in the same historical period; construct a difference scatter plot and a load difference prediction model according to the distribution characteristics of the temperature difference degrees and load difference degrees corresponding to the date clusters; obtain a difference degree weight according to the scatter distribution characteristics in the difference scatter plot and train the load difference prediction model; A load monitoring module, configured to obtain a load prediction difference degree through a load difference prediction model according to a temperature feature vector of a preset current period; obtain a deviation index at the current moment according to the load prediction difference degree, the actual load difference degree of the preset current period, and a model prediction error feature; and monitor the load state at the current moment according to the change feature of the deviation index.

[0005] Further, the step of performing adaptive filtering according to the instantaneous fluctuation feature of the load time series to obtain a load filtered time series includes: Calculating the absolute value of the difference between any moment in the load time series and the previous adjacent moment to obtain a first value; calculating the absolute value of the difference between the any moment and the next adjacent moment to obtain a second value; calculating the average value of the first value and the second value to obtain a fluctuation feature value; calculating the absolute value of the difference between the previous adjacent moment and the next adjacent moment to obtain a third value; calculating the ratio of the fluctuation feature value to the third value and normalizing it to obtain an adjustment value; calculating the product of the adjustment value and a preset window length and taking the nearest odd number to obtain an adaptive window, and performing median filtering according to different adaptive windows at different moments in the load time series to obtain the load filtered time series.

[0006] Further, the step of constructing a load feature vector according to the data trend feature of a preset current period in the load filtered time series includes: Calculating the average value of the load in the preset current period and normalizing it to obtain a first load feature value; calculating the average value of the differences between adjacent moments in the preset current period and normalizing it to obtain a second load feature value; calculating the average value of the absolute values of the differences between adjacent moments in the preset current period and normalizing it to obtain a third load feature value; and constructing the load feature vector according to the first load feature value, the second load feature value, and the third load feature value.

[0007] Further, the step of obtaining different temperature difference degrees and date clusters according to the difference feature of the temperature feature vectors in the same historical period includes: Calculating the cosine similarity between the temperature feature vectors of any two same historical periods and performing a negative correlation mapping to obtain the temperature difference degree between the any two same historical periods; and clustering different historical dates through a mean shift clustering algorithm according to the temperature difference degree to obtain different date clusters.

[0008] Further, the step of constructing a difference scatter plot and a load difference prediction model according to the distribution features of the temperature difference degrees and load difference degrees corresponding to the date clusters includes: Construct a rectangular coordinate system with the horizontal axis being the temperature difference degree and the vertical axis being the load difference degree; generate coordinate points in the rectangular coordinate system according to the temperature difference degree and the load difference degree at the same time period of any two dates in the date cluster to obtain a difference scatter plot; construct a load difference prediction model for predicting the load difference degree by the temperature difference degree through the SVR algorithm in the date cluster.

[0009] Further, the step of obtaining the difference degree weight according to the scatter distribution characteristics in the difference scatter plot and training the load difference prediction model includes: Calculate the ratio of the temperature difference degree and the load difference degree of the scatter points in the difference scatter plot to obtain the ratio coefficient of the scatter points; calculate the absolute value of the difference between the ratio coefficients of the scatter points and other scatter points to obtain the ratio difference; calculate the minimum value of the absolute value of the difference in temperature difference degree and the absolute value of the difference in load difference degree between the scatter points and the other scatter points to obtain the difference characterization value; calculate the ratio of the ratio difference to the difference characterization value to obtain the difference degree between the scatter points and the other scatter points; calculate the average value of the difference degrees between the scatter points and all other scatter points to obtain the average difference degree of the scatter points; calculate the ratio of the average difference degree of the scatter points in all scatter points and map it negatively to obtain the difference degree weight of the scatter points. Calculate the square of the difference between the predicted value obtained by predicting the temperature difference value corresponding to the scatter point through the load difference prediction model and the actual load difference degree to obtain the prediction difference; calculate the average value of the product of the prediction differences of all scatter points and the difference degree weight to obtain the comprehensive prediction difference; use the load difference prediction model trained when the comprehensive prediction difference is the smallest as the final load difference prediction model.

[0010] Further, the step of obtaining the load prediction difference degree according to the temperature feature vector of the preset current time period through the load difference prediction model includes: Determine the date cluster where the temperature feature vector of the preset current time period is located, and use the temperature feature vectors of the two time periods corresponding to the scatter point with the largest difference degree weight in the date cluster where the preset current time period is located as the temperature target vector; calculate the average value of the temperature difference degree between the temperature feature vector of the preset current time period and the temperature target vector to obtain the input value; obtain the load prediction difference degree according to the input value through the final load difference prediction model.

[0011] Further, the step of obtaining the deviation index at the current moment according to the load prediction difference degree, the actual load difference degree of the preset current time period, and the model prediction error feature includes: Take the load feature vectors of two time periods corresponding to the scatter point with the largest difference degree weight in the date cluster where the preset current time period is located as the load target vector; calculate the average value of the load difference degree between the load feature vector of the preset current time period and the load target vector to obtain the actual load difference degree; calculate the square of the difference between the actual load difference degree and the predicted load difference degree to obtain the deviation degree; calculate the ratio of the deviation degree to the minimum value of the comprehensive prediction difference to obtain the deviation index at the current moment.

[0012] Further, the step of monitoring the load status at the current moment according to the change characteristics of the deviation index includes: Calculate the average value of the differences between the deviation indices of each moment and the next moment within the preset adjacent time period at the current moment and normalize it to obtain the deviation change degree; normalize the deviation index at the current moment to obtain the deviation characterization value; calculate the average value of the deviation characterization value and the deviation change degree to obtain the load abnormality degree; when the load abnormality degree exceeds the preset abnormality threshold, the load status at the current moment is abnormal.

[0013] The present invention has the following beneficial effects: In the present invention, obtaining the load filtering time series can adaptively weaken the instantaneous fluctuation characteristics and retain the normal load fluctuation characteristics, improving the accuracy of subsequent load analysis and monitoring. Obtaining the load feature vector can represent various load characteristics within a period of time; obtaining the temperature feature vector can represent various temperature characteristics within a period of time, so as to monitor the load status according to the load feature vector and the temperature feature vector. Obtaining the temperature difference degree can represent the degree of temperature feature difference between the same historical time periods, and obtaining the date cluster can cluster the dates with similar temperature characteristics into a cluster. Since the weather temperature has an obvious impact on the load magnitude, analyzing the relationship between the temperature feature and the load feature within the cluster can improve the accuracy of subsequent prediction and load monitoring. Obtaining the load difference degree can represent the degree of load difference between the same historical time periods, obtaining the difference scatter plot can determine the distribution of the temperature difference degree and the load difference degree between the same time periods within the date cluster, obtaining the load difference prediction model can be used to predict the load difference degree of the preset current time period, and then perform load status monitoring; obtaining the difference degree weight can train and optimize the load difference prediction model to improve the model prediction accuracy. Obtaining the predicted load difference degree can represent the relatively normal load difference degree under the same temperature conditions in the preset current time period and the same historical time periods; obtaining the deviation index can represent the deviation of the load status at the current moment from the same historical time periods. Finally, monitoring the load status at the current moment according to the change characteristics of the deviation index takes into account the dynamic changes and regular characteristics of the load characteristics under the same temperature conditions in the same historical time periods, improving the accuracy of load status monitoring of the ring network cabinet. Description of the Drawings

[0014] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0015] Figure 1 It is a block diagram of a ring main unit module with an overload fault monitoring and opening function provided by an embodiment of the present invention. Specific Embodiments

[0016] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the drawings and preferred embodiments, details the specific embodiments, structures, features, and effects of a ring main unit with an overload fault monitoring and opening function according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0018] The following specifically describes in conjunction with the drawings the specific solution of a ring main unit with an overload fault monitoring and opening function provided by the present invention.

[0019] Please refer to Figure 1 , which shows a block diagram of a ring main unit module with an overload fault monitoring and opening function provided by an embodiment of the present invention. The ring main unit includes the following modules: A data acquisition module S1, configured to acquire the load time series of the circuit and the temperature time series of the environment.

[0020] In the embodiment of the present invention, the implementation scenario is to monitor the load status of the circuit through the ring main unit to improve the accuracy of load monitoring and control; first, acquire the load time series of the circuit and the temperature time series of the environment. The load time series is constructed by the power of the circuit passing through the ring main unit, and the temperature is the outdoor temperature in the area where the circuit is located. In the embodiment of the present invention, the load time series and the temperature time series within one month before the current moment are collected as the analysis range, and the implementer can determine the data collection duration and collection frequency according to the implementation scenario.

[0021] The feature analysis module S2 is used to perform adaptive filtering based on the instantaneous fluctuation characteristics of the load time series to obtain the load filtered time series; construct a load feature vector according to the data trend characteristics of a preset current period in the load filtered time series; and construct a temperature feature vector according to the data trend characteristics of a preset current period in the temperature time series.

[0022] During the monitoring process of circuit overload faults, if the load of the power grid suddenly rises and falls within a short period of time, such as in the case of large equipment startup and arc welding, it is easy for the ring network cabinet monitoring system to misjudge such non-persistent fluctuations as overload situations, triggering false alarms or unnecessary protection mechanisms. To improve the monitoring accuracy and anti-interference ability and enhance the recognition ability of real overload faults, first analyze the transient characteristics of the collected load time series, and while effectively removing them using the median filtering method, retain the normal fluctuation characteristics of the load in the power grid; therefore, perform adaptive filtering according to the instantaneous fluctuation characteristics of the load time series to obtain the load filtered time series.

[0023] Preferably, in an embodiment of the present invention, the step of obtaining the load filtering time sequence includes: calculating the absolute value of the difference between any moment and the previous adjacent moment in the load time sequence to obtain a first value; calculating the absolute value of the difference between the arbitrary moment and the subsequent adjacent moment to obtain a second value; calculating the average of the first value and the second value to obtain a fluctuation characteristic value; the first value and the second value characterize the instantaneous fluctuation degree at the arbitrary moment, and the larger the fluctuation characteristic value, the more likely it is that the load will rise or fall sharply at the arbitrary moment. Calculate the absolute value of the difference between the previous adjacent moment and the subsequent adjacent moment to obtain a third value; the third value characterizes the load difference characteristics before and after the arbitrary moment. Calculate the ratio of the fluctuation characteristic value to the third value and normalize it to obtain an adjustment value; it should be noted that in all calculation processes in the embodiment of the present invention, if the denominator appears to be 0, for example, the third value is 0, then a preset minimum positive number is added to the denominator to avoid the denominator being 0. In the embodiment of the present invention, the preset minimum positive number is 0.01. When the fluctuation characteristic value is larger and the third value is smaller, it means that the instantaneous fluctuation characteristics at any moment are more obvious, the adjustment value is larger, and then the window for median filtering at any moment is larger, and the instantaneous fluctuation characteristics at any moment are more effectively removed. Calculate the product of the adjustment value and the preset window length and take the nearest odd number to obtain an adaptive window. In an embodiment of the present invention, the preset window length is 11 acquisition moments, and the implementer can determine it according to the implementation scenario. When the adaptive window is larger, it means that more adjacent data are needed to perform median filtering at any moment to smooth the instantaneous fluctuation characteristics and improve the accuracy of subsequent load state analysis. Perform median filtering according to different adaptive windows at different moments in the load time series to obtain a load filtering time series; it should be noted that median filtering belongs to the prior art, and the specific steps will not be repeated; the load filtering time series can filter the instantaneous fluctuation characteristics of the load and retain the characteristics of normal load fluctuations, which can enhance the ability to identify real overload conditions.

[0024] Furthermore, in overload monitoring, weather temperature data is an important reference factor, because temperature changes have a significant impact on the fluctuation of power load, especially under weather conditions such as high temperatures in summer or severe cold in winter; for example, in continuous high temperature weather, the extensive use of high-power equipment such as air conditioners will cause a significant increase in power load in residential and commercial areas, which can easily cause the equipment to operate close to or above the rated load for a long time, increasing the risk of overload failure. Introducing temperature as a factor into the analysis of load characteristics can help identify the normal load fluctuation range under different temperature conditions; when the load data at the latest moment deviates significantly from the historical distribution under the current temperature conditions, for example, the load is much higher than the load characteristics at similar temperatures in the past, it can be judged as having a potential overload trend. First, determine the load characteristics, and construct a load characteristic vector based on the data trend characteristics of the current period preset in the load filtering time series.

[0025] Preferably, in the embodiments of the present invention, the step of obtaining the load feature vector includes: calculating the average load of a preset current period and normalizing it to obtain a first load feature value; in the embodiments of the present invention, the preset current moment is within 3 hours before the current moment, which is used to reflect the local load characteristics before the current moment, and the implementer can determine it according to the implementation scenario; the first load feature value represents the average load level of this period, and the larger the value, the more it means long-term continuous high-load operation. Calculating the average value of the load differences between adjacent moments in the preset current period and normalizing it to obtain a second load feature value; the load difference between adjacent moments is the load difference between any moment and the previous moment; the second load feature value represents whether the load shows a continuous upward trend. When the second load feature value is larger, it means that the upward trend of the load is more obvious in this period. Calculating the average value of the absolute values of the load differences between adjacent moments in the preset current period and normalizing it to obtain a third load feature value; the third load feature value represents the degree of load fluctuation in this period. When the third load feature value is larger, it means that the load fluctuation characteristics are more obvious in this period. Constructing a load feature vector according to the first load feature value, the second load feature value and the third load feature value; this load feature vector represents various load characteristics of the preset current period.

[0026] Similarly, a temperature feature vector is constructed according to the data trend characteristics of the preset current period in the temperature time series; it should be noted that the steps for obtaining the temperature feature vector are the same as those for the load feature vector. The first temperature feature value, the second temperature feature value and the third temperature feature value are obtained according to the temperature time series of the preset current period to construct the temperature feature vector, and the specific steps will not be elaborated here; the temperature feature vector represents various temperature characteristics of the preset current moment.

[0027] The feature processing module S3 is used to obtain different temperature difference degrees and date clusters according to the difference characteristics of the temperature feature vectors in the same historical period; obtain the load difference degree according to the difference characteristics of the load feature vectors in the same historical period; construct a difference scatter plot and a load difference prediction model according to the distribution characteristics of the temperature difference degrees and load difference degrees corresponding to the date clusters; obtain the difference degree weights according to the scatter distribution characteristics in the difference scatter plot and train the load difference prediction model.

[0028] Since the weather temperature is an important factor affecting the load difference, it is necessary to analyze the load characteristics under similar weather temperature conditions when analyzing the load characteristic differences. Therefore, different temperature difference degrees and date clusters are obtained according to the difference characteristics of the temperature feature vectors in the same historical time period. Preferably, in the embodiment of the present invention, the steps of obtaining the temperature difference degree and the date cluster include: calculating the cosine similarity of the temperature feature vectors of any two same historical time periods and performing a negative correlation mapping to obtain the temperature difference degree of the any two same historical time periods. It should be noted that the cosine similarity belongs to the prior art, and the specific calculation steps will not be elaborated. When the two temperature feature vectors are more similar, the cosine similarity is greater, which means that the temperature features of the two same time periods are more similar and the temperature difference degree is smaller. Clustering the historical different dates by the mean shift clustering algorithm according to the temperature difference degree to obtain different date clusters. It should be noted that the mean shift clustering algorithm belongs to the prior art, and the specific steps will not be elaborated. When the temperature difference degree between the two temperature feature vectors is smaller, the dates corresponding to the two temperature feature vectors are more likely to be in the same date cluster. The temperature feature vectors within each date cluster are similar, and the weather temperature conditions in the same time period of the corresponding dates are similar.

[0029] Furthermore, after obtaining different date clusters, it is necessary to analyze the differences in the load feature vectors corresponding to different dates within the date clusters, so as to be used for subsequent judgment of whether there is an overload situation at the current moment. Therefore, the load difference degree is obtained according to the difference characteristics of the load feature vectors in the same historical time period. It should be noted that the calculation method of the load difference degree is the same as that of the temperature difference degree, and it is obtained through the cosine similarity between the load feature vectors and negative correlation mapping. When the load difference degree is greater, it means that the load characteristic difference between the two same time periods is more obvious.

[0030] Within the same date cluster, if the temperature characteristics are similar for the same time periods of different dates, the electrical load characteristics should also be similar. Furthermore, the load characteristic differences can be predicted based on the temperature characteristic differences. Therefore, a difference scatter plot and a load difference prediction model are constructed according to the distribution characteristics of the temperature difference degree and the load difference degree corresponding to the date cluster. Preferably, in the embodiment of the present invention, the steps of constructing the difference scatter plot and the load difference prediction model include: constructing a rectangular coordinate system, with the horizontal axis being the temperature difference degree and the vertical axis being the load difference degree; generating coordinate points in the rectangular coordinate system based on the temperature difference degree and the load difference degree of the same time periods of any two dates in the date cluster to obtain the difference scatter plot; the horizontal and vertical coordinate values of each scatter point in the difference scatter plot represent the temperature characteristic difference and the load characteristic difference of the corresponding two dates at the same time period. The positions of the scatter points in the difference scatter plot corresponding to the same date cluster are relatively similar. Furthermore, a load difference prediction model for predicting the load difference degree based on the temperature difference degree can be constructed by the SVR algorithm in the date cluster. It should be noted that the SVR support vector regression algorithm belongs to the prior art. This algorithm can perform data prediction, and the specific steps will not be elaborated. This load difference prediction model can predict the load difference degree of the scatter point ordinate based on the temperature difference degree of the scatter point abscissa.

[0031] Furthermore, in order to improve the accuracy of the prediction model, avoid the situation where the model falls into local optimality or the distribution of some scattered points affects the overall prediction accuracy, and ensure the accuracy of the final overload state monitoring, it is necessary to train and optimize the load difference prediction model. Therefore, the difference degree weight is obtained according to the scattered point distribution characteristics in the difference scatter plot, and the load difference prediction model is trained. Preferably, in the embodiment of the present invention, the steps of obtaining the difference degree weight and training the load difference prediction model include: calculating the ratio of the temperature difference degree and the load difference degree of the scattered points in the difference scatter plot to obtain the ratio coefficient of the scattered point; in the same date cluster, since the temperature feature vector and the load feature vector are similar, the ratio coefficients of most scattered points are relatively close. Calculate the absolute value of the difference between the ratio coefficient of the scattered point and the ratio coefficients of other scattered points to obtain the ratio difference; the larger the ratio difference, the greater the difference between the ratio coefficients corresponding to the two scattered points. Calculate the minimum value of the absolute value of the difference between the temperature difference degrees and the absolute value of the difference between the load difference degrees between the scattered point and other scattered points to obtain the difference characterization value; the smaller the difference characterization value, the more similar the temperature difference degrees and the load difference degrees of the two scattered points. Calculate the ratio of the ratio difference to the difference characterization value to obtain the difference degree between the scattered point and the other scattered point; the difference characterization value is used as the weight for calculating the difference degree between the two scattered points. When the horizontal axis or the vertical axis of the two scattered points is more similar, the result of the ratio difference is more important. Calculate the average value of the difference degrees between the scattered point and all other scattered points to obtain the average difference degree of the scattered point; if the distribution positions of the scattered point and other scattered points are relatively close, the ratio difference is smaller and the average difference degree is smaller; if there are obvious differences in the distribution positions of the scattered point and other scattered points, the ratio difference is larger and the average difference degree is larger. The greater the difference between the temperature difference degree and the load difference degree represented by the scattered point and other scattered points, the lower the data reliability of the scattered point in the prediction process. Calculate the proportion of the average difference degree of the scattered point in all scattered points and perform a negative correlation mapping to obtain the difference degree weight of the scattered point; the larger the difference degree weight, the more common the temperature difference degree and the load difference degree corresponding to the scattered point, and the higher the importance in the prediction model training process.

[0032] After obtaining the difference degree weights of each scatter point in the difference scatter plot corresponding to the date cluster, the temperature difference value corresponding to the scatter point can be calculated as the square of the difference between the predicted value obtained through the load difference prediction model and the actual load difference degree, so as to obtain the prediction difference. When the difference between the predicted value and the actual load difference degree is larger, the prediction difference is larger and the prediction effect is worse. Calculate the average value of the product of the prediction differences of all scatter points and the difference degree weights to obtain the comprehensive prediction difference. The difference degree weights weight the prediction differences to avoid the influence of the temperature difference degree and load difference degree corresponding to the scatter points that are originally abnormally distributed in the scatter plot on the prediction accuracy. Furthermore, for the scatter points with larger difference degree weights, their prediction differences are more important. Take the load difference prediction model trained when the comprehensive prediction difference is the smallest as the final load difference prediction model. At this time, the prediction accuracy of this load difference prediction model is the highest, thereby improving the accuracy of monitoring the load status at the latest moment.

[0033] The load monitoring module S4 is used to obtain the load prediction difference degree through the load difference prediction model according to the temperature feature vector of the preset current time period; obtain the deviation index at the current moment according to the load prediction difference degree, the actual load difference degree of the preset current time period, and the model prediction error feature; monitor the load status at the current moment according to the change feature of the deviation index.

[0034] After obtaining the load difference prediction model of the date cluster, the load prediction difference degree can be obtained through the load difference prediction model according to the temperature feature vector of the preset current time period. Preferably, in the embodiment of the present invention, the steps of obtaining the load prediction difference degree include: determining the date cluster where it is located according to the temperature feature vector of the preset current time period, clustering according to the temperature difference degree between the temperature feature vector of the preset current time period and the temperature feature vector of the same historical time period to determine the date cluster where it is located, and the specific steps will not be elaborated. Take the temperature feature vectors of two time periods corresponding to the scatter point with the largest difference degree weight in the date cluster where the preset current time period is located as the temperature target vectors. The scatter point with the largest difference degree weight means that its temperature difference degree and load difference degree are the closest to those of all other scatter points in this date cluster, and can represent the comprehensive temperature difference degree level and load difference degree level in this date cluster. Furthermore, take the temperature feature vectors of two time periods corresponding to this scatter point as the temperature target vectors, and the temperature target vectors represent the representative temperature features of this date cluster. Calculate the average value of the temperature difference degree between the temperature feature vector of the preset current time period and the temperature target vectors to obtain the input value; obtain the load prediction difference degree through the final load difference prediction model according to the input value.

[0035] Further, after obtaining the load prediction difference degree at the preset current moment, the deviation index at the current moment can be obtained according to the load prediction difference degree, the actual load difference degree in the preset current period, and the model prediction error characteristics. Preferably, in the embodiment of the present invention, the steps of obtaining the deviation index include: using the load feature vectors of two time periods corresponding to the scatter point with the largest difference degree weight in the date cluster where the preset current period is located as the load target vector; the load target vector represents the representative load characteristics in the date cluster, and the two time periods are also the two time periods referred to in the temperature difference degree prediction process. Calculate the average value of the load difference degree between the load feature vector of the preset current period and the load target vector to obtain the actual load difference degree; the actual load difference degree represents the difference size between the load feature vector of the preset current period and the load target vector. Calculate the square of the difference between the actual load difference degree and the load prediction difference degree to obtain the deviation degree; if the load feature vector and the load state of the preset current period are more abnormal, the greater the difference between the actual load difference degree and the load prediction difference degree, and the greater the deviation degree. Calculate the ratio of the deviation degree to the minimum value of the comprehensive prediction difference to obtain the deviation index at the current moment; the comprehensive prediction difference represents the prediction accuracy of the load difference prediction model. When the value of the comprehensive prediction difference is smaller, the prediction accuracy is higher, and the result of the deviation degree is more reliable. If the deviation degree is greater at this time, the deviation index is greater.

[0036] Since the deviation index at the current moment is obtained based on the load characteristics of the preset current time period, if there is an obvious overload situation in the recent time period of the current moment, the value of the deviation index will gradually increase. To improve the accuracy of load status monitoring, the load status at the current moment can be monitored according to the change characteristics of the deviation index; preferably, in the embodiment of the present invention, the steps include: calculating the average value of the differences between the deviation indexes of each moment and the next moment within the preset adjacent time period at the current moment and normalizing it to obtain the deviation change degree; in the embodiment of the present invention, the preset adjacent time period is the time period 30 minutes before the current moment, and the implementer can determine it according to the implementation scenario by himself; the greater the deviation change degree, it means that there is an upward trend in the deviation index in this preset adjacent time period, and the abnormal trend of the load characteristics gradually increases. Normalize the deviation index at the current moment to obtain the deviation characterization value. Calculate the average value of the deviation characterization value and the deviation change degree to obtain the load abnormality degree; the greater the load abnormality degree, it means that the load characteristics have a gradually abnormal change trend before and the current abnormal situation is more obvious. When the load abnormality degree exceeds the preset abnormality threshold, the load status at the current moment is abnormal, and the ring main unit needs to activate the warning and protection mechanism in time; in the embodiment of the present invention, the preset abnormality threshold is 0.4, and the implementer can determine it according to the implementation scenario by himself. The embodiment of the present invention monitors the load status through the load abnormality degree, takes into account the influence of the weather temperature factor on the load characteristics, analyzes the load difference under the condition of similar temperature, and determines whether there is a load abnormality according to the load difference; enables the load status monitoring at the current moment to combine the load changes and rules in the same historical time period, and improves the accuracy of load status monitoring.

[0037] In summary, the embodiment of the present invention provides a ring main unit with an overload fault monitoring and opening function; obtaining a load filtered time series according to the instantaneous fluctuation characteristics of the load time series; constructing a load feature vector according to the load filtered time series; constructing a temperature feature vector according to the temperature time series; obtaining a temperature difference degree and a date cluster according to the temperature feature vector; obtaining a load difference degree according to the load feature vector; constructing a difference scatter plot and a load difference prediction model according to the temperature difference degree and the load difference degree corresponding to the date cluster; obtaining the difference degree weight according to the scatter distribution characteristics and training the load difference prediction model. Obtaining a load prediction difference degree according to the temperature feature vector of the preset current time period; obtaining a deviation index according to the load prediction difference degree, the actual load difference degree, and the model prediction error characteristics; monitoring the load status according to the change characteristics of the deviation index, and improving the accuracy of load monitoring.

[0038] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0039] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the differences between each embodiment and other embodiments are emphasized.

Claims

1. A ring network box with overload fault monitoring and circuit breaking function, characterized in that: The ring network box includes the following modules: A data acquisition module, used to acquire the load timing of the circuit and the temperature timing of the environment; A characteristic analysis module, used for performing adaptive filtering according to the instantaneous fluctuation characteristics of the load time series to obtain the load filtering time series; Constructing a load feature vector according to the data trend characteristics of the current period preset in the load filtering time series; constructing a temperature feature vector according to the data trend characteristics of the current period preset in the temperature time series; A feature processing module is used to obtain different temperature differences and date clusters according to the difference characteristics of temperature feature vectors in the same historical period; obtain load differences according to the difference characteristics of load feature vectors in the same historical period; construct a difference scatter plot and a load difference prediction model according to the distribution characteristics of the temperature difference and load difference corresponding to the date clusters; obtain a difference weight according to the scatter distribution characteristics in the difference scatter plot and train a load difference prediction model; The load monitoring module is used to obtain the load prediction difference through the load difference prediction model according to the temperature characteristic vector of the preset current time period; obtain the deviation index at the current moment according to the load prediction difference and the actual load difference of the preset current time period and the model prediction error characteristics; monitor the load state at the current moment according to the change characteristics of the deviation index.

2. The ring main box with overload fault monitoring and circuit breaking function according to claim 1, characterized in that: The step of performing adaptive filtering according to the instantaneous fluctuation characteristics of the load time series to obtain the load filtering time series comprises: Calculate the absolute value of the difference between any moment and the previous adjacent moment in the load time series to obtain a first value; calculate the absolute value of the difference between the arbitrary moment and the next adjacent moment to obtain a second value; calculate the average value of the first value and the second value to obtain a fluctuation characteristic value; calculate the absolute value of the difference between the previous adjacent moment and the next adjacent moment to obtain a third value; calculate the ratio of the fluctuation characteristic value to the third value and normalize it to obtain an adjustment value; calculate the product of the adjustment value and the preset window length and take the nearest odd number to obtain an adaptive window, perform median filtering according to different adaptive windows at different moments in the load time series, and obtain the load filtering time series.

3. The ring main box with overload fault monitoring and circuit breaking function according to claim 1, characterized in that: The step of constructing a load feature vector according to the data trend characteristics of the current period preset in the load filtering time sequence includes: Calculate the load average value of the preset current period and normalize it to obtain a first load characteristic value; calculate the average value of the load difference between adjacent moments in the preset current period and normalize it to obtain a second load characteristic value; calculate the average value of the absolute value of the load difference between adjacent moments in the preset current period and normalize it to obtain a third load characteristic value; construct the load characteristic vector based on the first load characteristic value, the second load characteristic value and the third load characteristic value.

4. The ring main box with overload fault monitoring and circuit breaking function according to claim 1, characterized in that: The step of obtaining different temperature differences and date clusters according to the difference characteristics of the temperature feature vectors in the same historical period includes: The cosine similarity of the temperature feature vectors of any two identical periods in history is calculated and negatively correlated to obtain the temperature difference between the arbitrary two identical periods; different historical dates are clustered by the mean shift clustering algorithm according to the temperature difference to obtain different date clusters.

5. The ring main box with overload fault monitoring and circuit breaking function according to claim 1, characterized in that: The step of constructing a difference scatter plot and a load difference prediction model according to the distribution characteristics of the temperature difference and the load difference corresponding to the date clusters includes: A rectangular coordinate system is constructed, with the horizontal axis representing the temperature difference and the vertical axis representing the load difference; coordinate points are generated in the rectangular coordinate system according to the temperature difference and load difference of the same period of any two dates in the date cluster to obtain a difference scatter plot; a load difference prediction model for predicting the load difference by using the temperature difference in the date cluster is constructed by using the SVR algorithm.

6. The ring main box with overload fault monitoring and circuit breaking function according to claim 5, characterized in that: The step of obtaining the difference weight according to the scatter distribution characteristics in the difference scatter plot and training the load difference prediction model comprises: Calculate the ratio of the temperature difference and the load difference of the scatter point in the difference scatter diagram to obtain the ratio coefficient of the scatter point; calculate the absolute value of the difference between the ratio coefficients of the scatter point and other scatter points to obtain the ratio difference; calculate the minimum value of the absolute value of the difference between the temperature difference and the absolute value of the difference between the scatter point and other scatter points to obtain a difference characterization value; calculate the ratio of the ratio difference to the difference characterization value to obtain the degree of difference between the scatter point and the other scatter points; calculate the average value of the degree of difference between the scatter point and all other scatter points to obtain the average degree of difference of the scatter point; calculate the proportion of the average degree of difference of the scatter point in all scatter points and negatively correlate them to obtain the degree of difference weight of the scatter point; The temperature difference value corresponding to the calculated scattered point is obtained by calculating the square of the difference between the predicted value obtained by the load difference prediction model and the actual load difference to obtain the predicted difference; the average value of the product of the predicted difference of all scattered points and the difference weight is calculated to obtain the comprehensive predicted difference; the load difference prediction model trained when the comprehensive predicted difference is the smallest is used as the final load difference prediction model.

7. The ring main box with overload fault monitoring and circuit breaking function according to claim 6, characterized in that: The step of obtaining the load prediction difference through the load difference prediction model according to the temperature characteristic vector of the preset current period includes: Determine the date cluster according to the temperature characteristic vector of the preset current time period, and use the temperature characteristic vectors of the two time periods corresponding to the scatter points with the largest difference weight in the date cluster where the preset current time period is located as the temperature target vector; calculate the average value of the temperature difference between the temperature characteristic vector of the preset current time period and the temperature target vector to obtain the input value; and obtain the load prediction difference through the final load difference prediction model according to the input value.

8. The ring main box with overload fault monitoring and circuit breaking function according to claim 7, characterized in that: The step of obtaining the deviation index at the current moment according to the load prediction difference, the actual load difference in the preset current period, and the model prediction error characteristics comprises: The load characteristic vectors of the two time periods corresponding to the scattered points with the largest difference weight in the date cluster where the preset current time period is located are used as load target vectors; the average value of the load difference between the load characteristic vector of the preset current time period and the load target vector is calculated to obtain the actual load difference; the square of the difference between the actual load difference and the load forecast difference is calculated to obtain the degree of deviation; the ratio of the degree of deviation to the minimum value of the comprehensive forecast difference is calculated to obtain the deviation index at the current moment.

9. The ring main box with overload fault monitoring and circuit breaking function according to claim 1, characterized in that: The step of monitoring the load state at the current moment according to the variation characteristics of the deviation index comprises: Calculate the average difference of the deviation index between each moment and the next moment in the preset adjacent time period of the current moment and normalize them to obtain the deviation change degree; normalize the deviation index of the current moment to obtain the deviation characterization value; calculate the average value of the deviation characterization value and the deviation change degree to obtain the load abnormality; when the load abnormality exceeds the preset abnormal threshold, the load state at the current moment is abnormal.

Images