A load event identification method and system

Abstract

The present application discloses a load event identification method and system, which relates to the field of data identification technology. The method comprises: obtaining an active power sequence, calculating a power difference corresponding to each time point; comparing the power difference with a standard power range threshold to obtain an initial transient interval; re-determining the starting point and the ending point according to the trend analysis result to obtain a stable transient interval; generating a trend analysis result by comparing the absolute value of the power difference with a constant power threshold; determining a maximum fluctuation value or a minimum fluctuation value according to the start and stop state of the equipment, and determining the start and stop state of the equipment according to the maximum fluctuation value and the minimum fluctuation value; finding a turning point of power change by comparing the absolute value of the power difference with a constant power threshold, accurately determining the starting point and the ending point of the transient interval, avoiding the problem of reduced accuracy due to inaccurate interval boundaries, and achieving the effect of improving detection accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of data identification, and in particular to a load event identification method and system. Background Art

[0002] Non-intrusive load monitoring (NILM) is a computational technique used to estimate the power demand of a single device from a single meter that measures the combined demand of multiple devices. The ultimate goal may be to help users reduce energy consumption; or to identify faulty appliances; or to investigate device usage behavior. With the increasing diversification of electrical equipment, especially the popularity of variable-frequency equipment, for non-intrusive load identification, in the case of mixed variable-frequency equipment, the current non-intrusive load detection has a large number of false detection events because the active power fluctuations of variable-frequency equipment are relatively large, which is similar to the power fluctuations of some non-variable-frequency equipment when starting and stopping.

[0003] The Chinese invention patent with application number 202010546690.6 discloses a non-invasive load detection method, terminal equipment and storage medium. The power difference at each time point is obtained according to each active power in the active power sequence, and the transient interval with the start and stop state of the equipment is obtained. The preset power fluctuation value, the first preset threshold value and the second preset threshold value are set, and the maximum fluctuation value and the minimum fluctuation value are compared with the preset power fluctuation value to determine whether the start and stop state of the equipment is the start or stop of the non-frequency conversion equipment, thereby reducing false detection events and effectively improving the detection accuracy.

[0004] However, the existing patent technology only obtains the transient interval based on the change of active power and compares it with multiple preset thresholds, and cannot accurately determine the position of the boundary point of the transient interval. In addition, the power change of non-frequency conversion equipment and frequency conversion equipment cannot be effectively identified before they are officially and stably started or stopped, resulting in the identified boundary point not being the original boundary point of the actual start or stop, which reduces the accuracy of the detection result. Summary of the invention

[0005] The present application solves the problem that the start-stop state boundary points cannot be accurately obtained in the prior art by providing a load event identification method and system, thereby achieving the technical effect of accurately obtaining the start-stop state boundary points of the equipment and improving the accuracy of the detection results.

[0006] The present application provides a load event identification method, the method comprising:

[0007] S100: acquiring an active power sequence obtained by arranging multiple active powers in the order of acquisition time, and calculating the power difference corresponding to each time point according to the active power corresponding to each acquisition time;

[0008] S200: Compare the power difference with the standard power range threshold to obtain an initial transient interval corresponding to the start / stop state of the device; obtain detection windows of the starting point and the ending point according to the starting point and the ending point of the initial transient interval, obtain change trends of the starting point and the ending point according to the detection windows, generate trend analysis results, and re-determine the starting point and the ending point according to the trend analysis results to obtain a stable transient interval;

[0009] Step S200 also includes: obtaining an initial transient interval, segmenting the power difference sequence, determining whether each segment of data meets the covariance curve characteristics, and if it is identified as meeting the covariance curve characteristics, marking it as a covariance interval, extracting a covariance index, determining the device state, and generating a detection window according to the device state;

[0010] The change trend refers to the change trend of the power difference between the starting point and the end point preliminarily determined within the detection window, and the change trend includes judging whether the change is significant and continuous and whether the direction is consistent; and generating a trend analysis result by comparing the absolute value of the power difference with the constant power threshold;

[0011] S300: Calculate the power fluctuation value corresponding to each time point in the stable transient interval, determine the maximum fluctuation value or the minimum fluctuation value according to the start / stop state of the device, calculate the endpoint difference at the same time, and determine the stable start / stop state of the device according to the maximum fluctuation value and the minimum fluctuation value;

[0012] S400: Repeat steps S100 and S200 to obtain the stable transient intervals of multiple equipment start and stop events, set statistical indicators according to the starting and ending time points of each stable transient interval and the corresponding power values, and generate an analysis report; establish an associated analysis table based on the analysis report for predicting subsequent transient intervals.

[0013] Furthermore, the specific calculation formula of the power fluctuation value is as follows:

[0014]

[0015] The power average value calculation formula is as follows:

[0016]

[0017] Among them, the power sequence in the stable transient interval is , n is the number of sampling points in the interval, is the power fluctuation value, is the average power value; Represents the active power corresponding to each time point i.

[0018] Furthermore, step S200 also includes: calculating the mean and standard deviation of the power difference in the detection window, setting multiple threshold adjustment factors, and dynamically adjusting the standard power range threshold using the threshold adjustment factors during each iteration.

[0019] Furthermore, the covariance curve feature refers to the curve feature of the power difference change curve that changes from large to small and then to large in a short period of time; the covariance indicators include the rate of change of the power difference, the position of the extreme point of the power fluctuation value, and the turning point of the power change.

[0020] Furthermore, generating a detection window according to the device status specifically includes: obtaining a pre-start time point or a pre-stop time point of the device according to the device status, setting an initial detection window size, determining a data sampling frequency, calculating the time interval between each data point, and obtaining a final detection window according to a state feature analysis result and a pre-set reference window, wherein the reference window is a basic step size for adjusting the initial detection window.

[0021] Furthermore, the analysis report includes equipment start and stop events, statistical indicator data corresponding to each event, change trend analysis between adjacent boundary points, and abnormal event detection results; the association analysis table is used to represent the association rules between different stable transient intervals. When a new stable transient interval appears, its characteristics are matched with the association analysis table to predict the characteristic state of the new stable transient interval.

[0022] Furthermore, the equipment information and characteristic data of different stable transient intervals are identified according to the analysis report and the association analysis table, and a characteristic library is established to represent the mapping relationship between different stable transient intervals and different equipment, and to mark different operating status data information.

[0023] Furthermore, the feature library classifies and organizes the detected stable transient interval data according to equipment and operating status, extracts statistical indicators as features from the stable transient interval data under each category, and stores the extracted features and their corresponding classification information in the feature library.

[0024] A load event identification system, the system comprising:

[0025] The data acquisition and processing module is used to collect active power data in real time, sort it according to time to generate an active power sequence, pre-process the active power sequence data, and obtain the previous sample power average, the next sample power average and the power difference;

[0026] An initial transient interval generation module is used to receive the data acquisition and processing module and generate an initial transient interval;

[0027] A stable transient interval generation module is used to receive the initial transient interval, set the detection window to generate trend analysis results, and output the stable transient interval;

[0028] The state recognition module is used to calculate the power fluctuation value and determine the maximum fluctuation value or the minimum fluctuation value according to the start and stop state of the equipment, and at the same time calculate the endpoint difference, and determine the start and stop state of the equipment according to the maximum fluctuation value and the minimum fluctuation value.

[0029] Furthermore, the stable transient interval generation module also includes: a dynamic update unit, which is used to calculate the mean and standard deviation of the power difference in the detection window, set multiple threshold adjustment factors, and use the threshold adjustment factor to dynamically adjust the standard power range threshold in each iteration process.

[0030] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0031] By setting up detection windows around the starting and ending points of the transient interval that are initially determined, analyzing the changing trend of the power difference around these points, and comparing the absolute value of the power difference with the constant power threshold, combined with the linear regression analysis method, the turning point of the power change is found, thereby accurately determining the starting and ending points of the transient interval, solving the problem of interference from frequency conversion equipment, and avoiding the problem of reduced accuracy due to inaccurate interval boundaries, thereby achieving the effect of improving detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A schematic diagram of a load event identification method according to an embodiment of the present invention;

[0033] Figure 2 A schematic diagram of a load event identification system architecture in an embodiment of the present invention;

[0034] Figure 3 The present invention is a schematic diagram of a stable transient interval generation module of a load event identification system in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0037] Embodiment 1: Figure 1 As shown, a load event identification method comprises:

[0038] S100: acquiring an active power sequence obtained by arranging a plurality of active powers in order of acquisition time, and calculating a power difference corresponding to each time point according to the active power corresponding to each acquisition time.

[0039] In some embodiments, the voltage and current of the main meter inlet are detected in real time, and the real-time sequence of active power is calculated through the voltage and current. The real-time sequence of active power is subjected to median filtering to obtain a filtered power set as an active power sequence. In fact, the active power sequence is a sequence of multiple active powers arranged in the order of collection time. The power difference corresponding to each time point is calculated according to the active power corresponding to each collection time in the active power sequence. The specific steps include:

[0040] Step S101, calculating the average power of the front samples according to the active powers corresponding to N time points before a time point, and calculating the average power of the rear samples according to the active powers corresponding to N time points after the acquisition time;

[0041] Step S102, calculating a difference Y between the average power value of the previous sample and the average power value of the subsequent sample, and taking the difference Y as the power difference corresponding to the time point, until the power difference corresponding to each time point is calculated;

[0042] Specifically, the power difference is calculated using the overlapping sliding window strategy. The window size (for example, N1=3, N2=5, N3=7) determines the number of data points covered in each analysis. To achieve partial overlap between windows and ensure the accuracy of data analysis, the step size is set smaller than the window size. For example, the step size is set to one-fourth of the window size, rounded up, and the minimum step size is 1. Starting from the starting position of the active power sequence, the first window is placed. In each window, the average power of the previous sample and the average power of the next sample are calculated. When calculating, the time point itself is removed to divide the front and back windows into two equal parts. Assuming that the active power sequence is , the window size is N, the step size is S, and for each possible window starting position i (from 1 to n-N+1), the power difference within the window is calculated. The specific formula is as follows:

[0043] When the window size N is an odd number, the previous sample power average Q is considered. data points,

[0044]

[0045] The average power of the last sample B, considering the remaining window data points,

[0046]

[0047] When the window size N is an even number, the previous sample power average Q is considered. data points,

[0048]

[0049] The average power of the last sample B, considering the remaining window data points,

[0050]

[0051] For each window, the power difference is the difference between the average value of the next sample and the average value of the previous sample: power difference = average power value of the next sample - average power value of the previous sample. Slide the window to the right by S positions and repeat the above calculation steps until the entire active power sequence is traversed.

[0052] S200: Compare the power difference with the standard power range threshold to obtain an initial transient interval corresponding to the start and stop state of the equipment; obtain detection windows of the starting point and the ending point according to the starting point and the ending point of the initial transient interval, obtain the change trends of the starting point and the ending point according to the detection windows, generate a trend analysis result, and re-determine the starting point and the ending point according to the trend analysis result to obtain a stable transient interval.

[0053] In some embodiments, the power difference corresponding to the time point is compared with the standard power range threshold in the order of collection time, and the initial transient interval of the collection time corresponding to the start and stop state of the device is obtained according to the comparison result, wherein the start and stop state of the device includes the device start state and the device stop state; the standard power range threshold corresponds to a first power threshold and a second power threshold, the first power threshold is greater than the second power threshold and the first power threshold and the second power threshold are opposite to each other, wherein the first power threshold can be set to 80w, and the second power threshold is -80w. When the power difference corresponding to a certain time point is greater than or equal to the first power threshold, it means that the power has fluctuated significantly at this time point, and the power is in an upward trend. Therefore, the time point corresponds to the startup of the device, so the collection interval corresponding to the power difference is used as the starting point of the device startup state, and the power difference corresponding to each time point after the starting point is judged accordingly. Whether the power difference corresponding to the time point is greater than or equal to the first power threshold, until the power difference corresponding to the first time point after the starting point is greater than or equal to the first power threshold, in fact, the power difference corresponding to the first time point after the starting point is greater than or equal to the first power threshold, continue to judge whether the power difference corresponding to the second time point is greater than or equal to the first power threshold, until it is determined that the power difference at a certain time point is less than the first power threshold, and the power has not shown an upward trend, indicating that the device has been started at this time, and the time point corresponding to the first power difference less than the first power threshold after the starting point of the device is used as the end point of the device startup, and the starting point of the device startup until the end point of the device startup is set as the transient interval corresponding to the device startup, the starting point of the device startup and the end point of the device startup are respectively the two endpoints of the transient interval corresponding to the device startup, and the transient interval corresponding to the device startup can be further determined to determine whether the device startup is a real event of non-frequency conversion device startup.

[0054] In some embodiments, when the power difference corresponding to a certain time point is compared with the standard power range threshold, if it is less than or equal to the second power threshold, that is, less than or equal to -80w, it means that the power has fluctuated significantly at this time point, and the power is showing a downward trend. Therefore, this time point corresponds to the stop of the device. Therefore, the time point corresponding to the power difference is used as the starting point of the device stop state, and it is judged whether the power difference corresponding to each time point after the starting point is less than or equal to the second power threshold, until the power difference corresponding to the time point is greater than the second power threshold. In fact, the power difference corresponding to the first time point after the starting point is less than or equal to the second power threshold, and the judgment continues. Whether the power difference corresponding to the two time points is less than or equal to the second power threshold, until it is determined that the power difference at a certain time point is greater than the second power threshold, and the power has no downward trend, it means that the device has stopped at this time, and the time point corresponding to the first power difference greater than the second power threshold after the starting point of the device stop is taken as the end point of the device stop, and the starting point of the device stop until the end point of the device stop is set as the transient interval corresponding to the device stop, and the starting point of the device stop and the end point of the device stop are respectively the two endpoints of the transient interval corresponding to the device stop. Through the transient interval corresponding to the device stop, it can be further determined whether the device stop is a real event of non-frequency conversion device stop.

[0055] The changing trend refers to the changing trend of the power difference between the starting point and the ending point preliminarily determined within the detection window, and the changing trend includes judging whether the change is significant and continuous and whether the direction is consistent; by comparing the absolute value of the power difference with the constant power threshold, a trend analysis result is generated.

[0056] In some embodiments, according to the starting point and the ending point of the initial transient interval, the starting point and the ending point are further accurately analyzed to accurately determine the initial transient interval; after the starting point is preliminarily determined, the power difference change of several sampling points before and after it is checked to ensure that the power has undergone significant and continuous changes at this point, and detection windows are set to analyze the change trends of the starting point and the ending point respectively, and constant power thresholds are pre-set respectively, and power data within the detection window is extracted from the active power sequence. The detection window is centered on the starting point or the ending point of the transient interval preliminarily determined, and a certain number (such as K, K is a positive integer) of sampling points are extended before and after the time period, which is adjusted according to actual conditions or historical experience; the constant power The threshold is pre-set based on the power difference of historical data of multiple starting points or multiple ending points, and the comparison results of multiple power difference values ​​in the detection window and the constant power threshold are obtained to determine whether the power difference value maintains the same change direction (increase or decrease), and generate trend analysis results. The trend analysis results include the significance analysis of the power difference value corresponding to each time point in the detection window, the continuity analysis of the change trend of the power difference value, and the consistency analysis of the change direction. The trend analysis uses a regression analysis method to analyze the power data in the detection window to find the turning point of the change trend as a more accurate starting point and end point. Preferably, a linear regression analysis method is used, and a basic linear regression model is used for analysis. The specific formula is as follows:

[0057] in is the dependent variable (power difference), is the independent variable (time or time index), and are model parameters, is the error term; the model parameters are estimated using the least squares method (OLS) and The least squares method is used to find the best fitting parameters by minimizing the residual sum of squares; the trend of the power difference is analyzed based on the estimated linear regression model. The slope of the linear regression model Indicates the rate of change of the power difference over time; if If it is significantly not zero, it means that there is a significant changing trend in the power difference; sort the analysis results in the detection window from high to low, select the first sorted power difference, and determine whether the current first-ranked power difference is consistent with the starting point or the ending point. If not, replace the current starting point or the ending point, and redetermine the starting point and the ending point according to the trend analysis results to obtain a stable transient interval.

[0058] S300: Calculate the power fluctuation value corresponding to each time point in the stable transient interval, and determine the maximum fluctuation value or the minimum fluctuation value according to the start / stop state of the equipment, and calculate the endpoint difference at the same time, and determine the stable start / stop state of the equipment according to the maximum fluctuation value and the minimum fluctuation value.

[0059] In some embodiments, the power sequence in the stable transient interval is assumed to be , where n is the number of sampling points in the interval, and the power average is calculated. The specific formula is as follows:

[0060]

[0061] Calculate the power standard deviation, that is, the power fluctuation value. The specific formula is as follows:

[0062]

[0063] During the equipment startup process, the power will gradually increase from a low level to a stable level, so the maximum fluctuation value often occurs in the initial startup; while during the equipment stop process, the power will gradually decrease from a high level to zero or a low level, so the minimum fluctuation value occurs in the stop process; in the stable transient interval corresponding to the equipment startup, find the maximum value of the power fluctuation value as the maximum fluctuation value; in the stable transient interval corresponding to the equipment stop, find the power fluctuation value as the minimum fluctuation value; the endpoint difference refers to the power difference between the starting point and the end point of the stable transient interval.

[0064] In this embodiment, the maximum fluctuation value or the minimum fluctuation value is determined according to the power fluctuation value corresponding to each time point in the transient interval and the start-stop state of the equipment. Specifically, when the start-stop state of the equipment is the equipment startup state, the maximum power fluctuation value needs to be determined according to the power fluctuation value corresponding to the time point in the transient interval, and the maximum fluctuation value is compared with the preset power fluctuation value corresponding to the equipment start-stop state. Since it is the equipment startup state at this time, the preset power fluctuation value is the first power fluctuation value. When the maximum fluctuation value is greater than or equal to the first power fluctuation value, the endpoint difference between the active power corresponding to the starting point of the transient interval and the active power corresponding to the end point of the transient interval is calculated, and the endpoint difference is compared with the first preset threshold value. If the endpoint difference is less than the first preset threshold value, it is determined that the time point corresponding to the equipment startup state is the interference point of the variable frequency equipment; if the endpoint difference is greater than or equal to the first preset threshold value, it is determined that the equipment start-stop state is the start of the non-frequency variable equipment.

[0065] When the device start-stop state is the device stop state, it is necessary to determine the minimum power fluctuation value according to the power fluctuation value corresponding to the time point in the transient interval, and compare the minimum fluctuation value with the preset power fluctuation value corresponding to the device start-stop state. Since the device is in the stop state at this time, the preset power fluctuation value is the second power fluctuation value. When the maximum fluctuation value is less than or equal to the second power fluctuation value, the endpoint difference is compared with the second preset threshold value. If the endpoint difference is greater than the first preset threshold value, the time point corresponding to the device start state is determined to be the interference point of the frequency conversion device; if the endpoint difference is less than or equal to the second preset threshold value, the device start-stop state is determined to be the non-frequency conversion device stop. Through multi-step judgment, it can be finally determined that the device start-stop is a real event of the start or stop of the non-frequency conversion device, which can effectively eliminate the interference of the frequency conversion device, avoid the occurrence of false detection caused by the power of the frequency conversion device and the start and stop power amplitude of the non-frequency conversion device being similar, and improve the accuracy of detection. The first preset threshold and the second preset threshold are both pre-set according to historical data and actual conditions.

[0066] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0067] The present application sets a detection window around the starting point and end point of the transient interval that are preliminarily determined, analyzes the changing trend of the power difference around these points, and finds the turning point of the power change by comparing the absolute value of the power difference with the constant power threshold, combined with the linear regression analysis method, thereby accurately determining the starting point and end point of the transient interval, solving the problem of interference from frequency conversion equipment, and avoiding the problem of reduced accuracy due to inaccurate interval boundaries, thereby achieving the effect of improving detection accuracy.

[0068] Embodiment 2: In embodiment 1, the boundary points of the transient interval are analyzed and adjusted by setting a standard power range threshold. However, under different conditions, the use of the preset standard power range threshold affects the accuracy of the adjustment. This embodiment makes further limitations on the basis of the above embodiment.

[0069] Step S200 also includes: calculating the mean and standard deviation of the power difference in the detection window, setting multiple threshold adjustment factors, and dynamically adjusting the standard power range threshold using the threshold adjustment factors in each iteration process.

[0070] In some embodiments, the initial transient interval boundary (starting point and ending point) and the upper limit of the number of iterations are obtained, and a new time window is defined as a detection window using the current transient interval boundary. The power difference is recalculated within the detection window to generate new starting point and ending point. Each time a new starting point and ending point are obtained is an iterative process. The standard power range is adjusted according to the generated trend analysis result and the pre-set threshold adjustment factor. After the adjustment is completed, the transient interval is reselected using the new standard power range threshold, and the detection window is selected for the boundary point (i.e., the starting point and the ending point) of the new transient interval for the second time, and a second iterative process is performed. This process is repeated to determine in real time whether the newly determined transient interval boundary meets the difference threshold with the result of the previous iteration, until the difference threshold is met and the maximum number of iterations is reached. If the maximum number of iterations is reached, the closest set of selected intermediate values ​​is selected as the final boundary.

[0071] In some embodiments, the average of all power differences within the window is calculated. and standard deviation , set multiple threshold adjustment factors, such as and , using analysis and adjustment factors to dynamically adjust the standard power range threshold, setting the new first power threshold to , the new second power threshold is .here, and The selection of needs to consider the balance between the sensitivity of power changes and the false alarm rate, and should be adjusted according to the actual situation. Assume that in the initial iteration process, a window size of 10 is used, and the mean of the power difference in the window is calculated to be 20W and the standard deviation is 10W. Based on historical data experience, set and As the adjustment factor, the new first power threshold will be adjusted to 20W+2×10W=40W, and the new second power threshold will be adjusted to 20W-1.5×10W=5W.

[0072] In some embodiments, variable frequency equipment and non-variable frequency equipment will produce extremely similar power fluctuations before they are officially started and stopped, especially when the variable frequency equipment is pre-started and the non-variable frequency equipment is pre-stopped, their power change curves will be extremely similar and cross-change. In this case, accurate identification cannot be performed, resulting in misjudgment due to the similarity of power fluctuations, and the actual stable start and stop time of the equipment cannot be accurately determined.

[0073] Step S200 also includes: obtaining the initial transient interval, segmenting the power difference sequence, each segment contains a certain number of data points, judging whether each segment of data conforms to the covariance curve characteristics, if it is identified as conforming to the covariance curve characteristics, it is marked as a covariance interval, extracting the covariance index, determining the device status, and generating a detection window according to the device status.

[0074] In some embodiments, within the preliminarily determined initial transient interval, a pattern recognition technology, such as a support vector machine (SVM), a decision tree, etc., is used to analyze the power difference sequence, and the power difference sequence within the initial transient interval is segmented, each segment contains a certain number of data points, and the pattern recognition technology is applied to each segment of data to determine whether it conforms to the covariant curve feature, wherein the covariant curve feature refers to the curve feature in which the power difference change curve changes from large to small and then to large in a short period of time, and the curve has a minimum value, and there are the same maximum values ​​on both sides of the minimum value. If a segment of data is identified In order to meet the characteristics of the covariation curve, the section is marked as a covariation interval with similar power changes; in the identified covariation interval, multiple overlapping capture windows are set, and the initial size of the capture window is determined, such as containing 10 data points. In the covariation interval, the capture window is slid with half of the window size as the step size to ensure that there is a certain overlap between the windows; in the capture window, the linear regression time series analysis method is used to perform trend analysis on the power difference sequence to extract covariation indicators, which include the rate of change of the power difference, the position of the extreme point of the power fluctuation value, and the turning point of the power change.

[0075] According to the extracted covariance index, especially the turning point of power change, determine whether the power change in the capture window conforms to the covariance curve characteristics, check the number and order of the turning points of power change, if there are two turning points, and the first turning point indicates that the power changes from large to small, and the second turning point indicates that the power changes from small to large, it conforms to the covariance curve characteristics. Analyze the power fluctuation between the turning points to ensure that the fluctuation amplitude meets expectations; divide the capture window into different areas, rising area, falling area and stable area based on the turning points; determine the position of the extreme points in each area, as well as the turning point position between areas, and analyze the power value and time point of each key point; according to the divided areas and the determined reaction time interval, use the machine learning classifier to determine whether it is the pre-start / pre-stop of the variable frequency device or the pre-stop / pre-start of the non-variable frequency device. Pre-start and pre-stop refer to the process in which the power gradually stabilizes after the device receives the start and stop instructions.

[0076] Specifically, it extracts characteristic indicators within each capture window, including area division, key point location, power change rate, etc. These characteristic indicators are input into the machine learning classifier for classification judgment. According to the classification results, the state of the equipment is determined (pre-start / pre-stop of variable frequency equipment or pre-stop / pre-start of non-variable frequency equipment), ensuring accurate identification of similar power changes of variable frequency equipment and non-variable frequency equipment in the pre-start / pre-stop stage, and capturing subtle power changes, and finally determining the state of the equipment.

[0077] In the above embodiment, by cross-analyzing the start and stop of different devices, the reaction time intervals of different devices and the pre-start and pre-stop status identification are accurately obtained, thereby effectively distinguishing the power fluctuation patterns of variable frequency devices and non-variable frequency devices in the pre-start / pre-stop stages. By identifying the pre-start and pre-stop states of the equipment, the boundary point positions of the transient intervals can be further accurately identified.

[0078] Generating a detection window according to the device state specifically includes: obtaining a pre-start time point or a pre-stop time point of the device according to the device state, pre-setting an initial detection window size, the size of the initial detection window is pre-set according to actual conditions or historical experience data, determining a data sampling frequency, the data sampling frequency is the number of data points collected per second, calculating the time interval between each data point by the number of seconds and the data sampling frequency, analyzing the device state data in the time period before and after the pre-start time point, identifying features such as the power level change rate and the vibration frequency, preliminarily setting a window containing 5 data points as a reference window according to the state feature analysis results, monitoring the power change law before and after the pre-start time, if it is found that the power has begun to change significantly (such as slowly rising) in a certain period of time before the pre-start, then advancing the start time of the window to the time point when the change begins, ensuring that the window can capture the complete state change process; and obtaining a final detection window with a pre-set reference window, the reference window is a basic step length for adjusting the initial detection window, and is used in the device state monitoring process to finely adjust the initial detection window based on the state feature analysis results and the reference window as the basic step length to ensure the accuracy and effectiveness of the detection window.

[0079] In some embodiments, based on the reference window, if the power changes significantly and continuously (such as continuously rising or falling), the window is gradually expanded; each time it is expanded, a data point in the reference window is added until the change tends to be stable or reaches the preset maximum window size; when the window is expanded, the start time and end time of the window should be recalculated to ensure the continuity and effectiveness of the window; if the power change is not obvious or there is fluctuation, the window is narrowed and the focus is on the part with the most dramatic change. Similarly, the variance or standard deviation of the power change is calculated to determine the severity of the fluctuation. When narrowing the window, the data points of a reference window should be gradually reduced until the window size is appropriate and can accurately reflect the change in device status.

[0080] This embodiment introduces a dynamic threshold adjustment mechanism. In each iteration, the mean and standard deviation of the power difference in the detection window are calculated as the basis for dynamic adjustment. Multiple threshold adjustment factors are set to balance the sensitivity and false alarm rate of power changes. These adjustment factors and the statistics (mean and standard deviation) in the detection window are used to dynamically adjust the standard power range threshold (first power threshold and second power threshold); the boundary (starting point and ending point) of the initial transient interval and the upper limit of the number of iterations are obtained; in each iteration, a new detection window is defined using the current transient interval boundary, and the power difference in the window is recalculated. According to the trend analysis results and the threshold adjustment factor, the standard power range threshold is adjusted, and the transient interval is reselected. The iteration process is repeated until the difference between the newly determined transient interval boundary and the previous iteration result meets the preset threshold or reaches the maximum number of iterations. Through multiple iterations and dynamic threshold adjustment, the starting point and the end point of the transient interval are gradually and accurately located. When the iteration end condition is reached, the closest set of boundary points is selected as the final transient interval boundary.

[0081] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0082] The present application dynamically adjusts the standard power range threshold so that the threshold is more adapted to the power change characteristics under different states, thereby improving the accuracy of positioning the boundary points of the transient interval, reducing the misjudgment and missed judgment caused by the use of fixed thresholds, and improving the accuracy of overall detection; introducing an iterative optimization mechanism to gradually approach the actual transient interval boundary through multiple iterations, thereby enhancing adaptability and stability; identifying and monitoring the pre-start and pre-stop time points and status conditions between different devices based on the initial transient interval, thereby achieving the effect of accurately identifying the initial start or stop status of different devices.

[0083] Embodiment 3: In the above embodiment, a dynamically adjusted standard power range threshold is set to improve the accuracy of transient interval boundary positioning. However, analysis and judgment need to be repeated in each analysis process and the differentiated characteristics of stable transient intervals between different devices cannot be identified. This embodiment makes further improvements on the basis of the above embodiment.

[0084] The method also includes: S400: repeating steps S100 and S200 to obtain the stable transient intervals of multiple equipment start and stop events, setting statistical indicators according to the starting and ending time points of each stable transient interval and the corresponding power values, and generating an analysis report; establishing an associated analysis table according to the analysis report for predicting subsequent transient intervals.

[0085] In some embodiments, the statistical indicators include time intervals, power changes and characteristic indicators, and the characteristic indicators include data indicators such as mean, median, standard deviation and variance; the time interval is the duration of each transient interval, that is, the difference between the end time and the start time, which measures the duration of the transient interval and is used to understand the speed of the equipment start and stop process, evaluate equipment performance and perform time series analysis; the power change directly reflects the power fluctuation in the transient interval and measures the amplitude of the power change, which is used to determine whether the equipment is successfully started or stopped, evaluate the power response speed and detect abnormal power fluctuations; the characteristic indicators include mean, median and standard deviation, which measure the overall level of power in the transient interval and reflect the fluctuation range of the power value in the interval, and are used to evaluate the stability of active power, identify abnormal values ​​and perform power prediction.

[0086] The analysis report includes equipment start and stop events, statistical indicator data corresponding to each event, change trend analysis between adjacent boundary points, and abnormal event detection results; the association analysis table is used to represent the association rules between different stable transient intervals. When a new stable transient interval appears, its characteristics are matched with the association analysis table to predict the characteristic state of the new stable transient interval.

[0087] In some embodiments, the analysis report needs to perform statistical analysis based on the collected stable transient interval data of multiple equipment start and stop events, clean and organize the collected data to ensure the accuracy and completeness of the data, use comprehensive data for statistical analysis, extract statistical indicator data information, calculate the statistical indicators of each event, analyze the change trend between adjacent transient intervals, compare the power values ​​of the starting point and the ending point for adjacent transient intervals, judge the trend of power change, and calculate the power difference and power change rate between the boundary points of adjacent stable transient intervals; also include abnormal event detection results, the abnormal event detection is based on the significant abnormality of the active power change in the stable transient interval, if there is an abnormal situation, it will be recorded and preliminary solutions will be provided. According to the analysis report, the start and stop performance of the equipment can be evaluated through indicators such as time interval and power change; abnormal events can be discovered and reported in a timely manner, and the control strategy of the equipment can be optimized based on trend analysis and abnormal detection results to improve operation efficiency.

[0088] In some embodiments, the association analysis table establishes the association between the stable transient intervals based on the analysis report, obtains relevant associations based on indicators such as time intervals and power changes, uses association rule mining to extract association rules between adjacent stable transient intervals, and organizes the extracted association rules into an association form of rule antecedents and rule consequents; when a new stable transient interval appears, its characteristics are matched with the rule antecedents in the association analysis table, and if a matching rule is found, the characteristics of the adjacent stable transient intervals are predicted based on the rule consequent; the prediction results are evaluated based on actual situation data to verify the accuracy of the prediction, and the association analysis table is adjusted based on the actual situation, and new features are introduced to improve the prediction performance.

[0089] In this embodiment, by repeatedly executing steps S100 and S200, the stable transient intervals of multiple equipment start-stop events are obtained, statistical indicators are set, and in-depth statistical analysis is performed on the stable transient intervals to generate a detailed analysis report, including equipment start-stop events, statistical indicator data, adjacent boundary point change trend analysis, and abnormal event detection results; based on the data in the analysis report, the association rule mining technology is used to establish an association analysis table between stable transient intervals. The association analysis table records the association rules between different stable transient intervals, including rule antecedents and rule consequents, which are used for subsequent predictions. When a new stable transient interval appears, its features are matched with the rule antecedents in the association analysis table to predict the features of adjacent stable transient intervals; the association analysis table is adjusted according to actual conditions, and new features are introduced to improve the accuracy of the prediction. Based on the prediction results, the control strategy of the equipment is optimized to improve the operating efficiency.

[0090] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0091] This application extracts the association rules between stable transient intervals through association rule mining, and realizes the accurate identification and classification of the differentiated characteristics of stable transient intervals between different devices through statistical analysis and association analysis, thereby improving the applicability and flexibility of detection, and achieving more accurate and rapid identification of power changes between different devices and reducing the effect of false detection events.

[0092] Embodiment 4: This embodiment is a further improvement on the basis of the above embodiment.

[0093] According to the analysis report and the association analysis table, the equipment information and characteristic data of different stable transient intervals are identified, and a characteristic library is established to represent the mapping relationship between different stable transient intervals and different equipment, and to mark different operating status data information.

[0094] The feature library classifies and organizes the detected stable transient interval data according to equipment and operating status, extracts statistical indicators as features from the stable transient interval data under each category, and stores the extracted features and their corresponding classification information in the feature library.

[0095] In some embodiments, the feature library is a data set containing stable transient interval features under different devices and different operating conditions, and each feature includes multiple statistical indicators such as time interval, power change, average value, median, standard deviation, etc. The features in the feature library can be used as a benchmark to compare with the stable transient interval features detected in real time. If significant deviations or anomalies are found, the anomaly detection mechanism is triggered in time; by analyzing and mining the data in the feature library, characteristic patterns of different devices or operating conditions can be discovered, thereby achieving more accurate classification and prediction.

[0096] In some embodiments, when a new stable transient interval is detected, its features are extracted and compared with the baseline features in the feature library, and the features in the feature library are used to identify the device, determine which device the newly detected stable transient interval belongs to and obtain the corresponding device status, and match the newly detected features with the known patterns in the feature library to achieve more accurate classification and prediction.

[0097] The present invention also provides a load event identification system, such as Figure 2 As shown, the system comprises:

[0098] The data acquisition and processing module is used to collect active power data in real time, sort it according to time to generate an active power sequence, pre-process the active power sequence data, and obtain the previous sample power average, the next sample power average and the power difference;

[0099] An initial transient interval generation module is used to receive the data acquisition and processing module and generate an initial transient interval;

[0100] A stable transient interval generation module is used to receive the initial transient interval, set the detection window to generate trend analysis results, and output the stable transient interval;

[0101] The state recognition module is used to calculate the power fluctuation value and determine the maximum fluctuation value or the minimum fluctuation value according to the start and stop state of the equipment, and at the same time calculate the endpoint difference, and determine the start and stop state of the equipment according to the maximum fluctuation value and the minimum fluctuation value.

[0102] like Figure 3 As shown, the stable transient interval generation module also includes: a dynamic update unit, which is used to calculate the mean and standard deviation of the power difference in the detection window, set multiple threshold adjustment factors, and use the threshold adjustment factor to dynamically adjust the standard power range threshold in each iteration process.

[0103] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0104] This application stores the extracted features and their classification information in a feature library. The feature library is a data set containing stable transient interval features of different devices and different operating states. By comparing the features detected in real time with the benchmark features in the feature library, anomalies can be discovered and handled in a timely manner; accurate classification and prediction of devices can be achieved through pattern matching algorithms; anomalies in real-time detection data can be more accurately identified, power change trends of different devices can be distinguished, and the corresponding operating status of the device can be obtained at the same time, achieving the effect of accurate identification of the stable transient interval of the device.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A load event identification method, characterized in that: The method comprises: S100: acquiring an active power sequence obtained by arranging multiple active powers in the order of acquisition time, and calculating the power difference corresponding to each time point according to the active power corresponding to each acquisition time; S200: Compare the power difference with the standard power range threshold to obtain an initial transient interval corresponding to the start / stop state of the device; obtain detection windows of the starting point and the ending point according to the starting point and the ending point of the initial transient interval, obtain change trends of the starting point and the ending point according to the detection windows, generate trend analysis results, and re-determine the starting point and the ending point according to the trend analysis results to obtain a stable transient interval; Step S200 also includes: obtaining an initial transient interval, segmenting the power difference sequence, determining whether each segment of data meets the covariance curve characteristics, and if it is identified as meeting the covariance curve characteristics, marking it as a covariance interval, extracting a covariance index, determining the device state, and generating a detection window according to the device state; The change trend refers to the change trend of the power difference between the starting point and the end point preliminarily determined within the detection window, and the change trend includes judging whether the change is significant and continuous and whether the direction is consistent; and generating a trend analysis result by comparing the absolute value of the power difference with the constant power threshold; S300: Calculate the power fluctuation value corresponding to each time point in the stable transient interval, determine the maximum fluctuation value or the minimum fluctuation value according to the start / stop state of the device, calculate the endpoint difference at the same time, and determine the stable start / stop state of the device according to the maximum fluctuation value and the minimum fluctuation value; S400: Repeat steps S100 and S200 to obtain the stable transient intervals of multiple equipment start and stop events, set statistical indicators according to the starting and ending time points of each stable transient interval and the corresponding power values, and generate an analysis report; establish an associated analysis table based on the analysis report for predicting subsequent transient intervals.

2. A load event identification method according to claim 1, characterized in that: The specific calculation formula of the power fluctuation value is as follows: Among them, the power sequence in the stable transient interval is , n is the number of sampling points in the interval, is the power fluctuation value, is the average power value; represents the active power corresponding to each time point i; the power average value calculation formula is as follows: 。 3. A load event identification method according to claim 1, characterized in that: Step S200 also includes: calculating the mean and standard deviation of the power difference in the detection window, setting multiple threshold adjustment factors, and dynamically adjusting the standard power range threshold using the threshold adjustment factors in each iteration process.

4. A load event identification method according to claim 1, characterized in that: The covariance curve feature refers to the curve feature of the power difference change curve that changes from large to small and then to large in a short period of time; the covariance index includes the rate of change of the power difference, the position of the extreme point of the power fluctuation value, and the turning point of the power change.

5. A load event identification method according to claim 1, characterized in that: Generating a detection window according to the device status specifically includes: obtaining a pre-start time point or a pre-stop time point of the device according to the device status, setting an initial detection window size, determining a data sampling frequency, calculating the time interval between each data point, and obtaining a final detection window according to a state feature analysis result and a pre-set reference window, wherein the reference window is a basic step size for adjusting the initial detection window.

6. A load event identification method according to claim 1, characterized in that: The analysis report includes equipment start and stop events, statistical indicator data corresponding to each event, change trend analysis between adjacent boundary points, and abnormal event detection results; the association analysis table is used to represent the association rules between different stable transient intervals. When a new stable transient interval appears, its characteristics are matched with the association analysis table to predict the characteristic state of the new stable transient interval.

7. A load event identification method according to claim 1, characterized in that: According to the analysis report and the association analysis table, the equipment information and characteristic data of different stable transient intervals are identified, and a characteristic library is established to represent the mapping relationship between different stable transient intervals and different equipment, and to mark different operating status data information.

8. A load event identification method according to claim 7, characterized in that: The feature library classifies and organizes the detected stable transient interval data according to equipment and operating status, extracts statistical indicators as features from the stable transient interval data under each category, and stores the extracted features and their corresponding classification information in the feature library.

9. A load event identification system, applied to a load event identification method according to any one of claims 1 to 8, characterized in that: The system comprises: The data acquisition and processing module is used to collect active power data in real time, sort it according to time to generate an active power sequence, pre-process the active power sequence data, and obtain the previous sample power average, the next sample power average and the power difference; An initial transient interval generation module is used to receive the data acquisition and processing module and generate an initial transient interval; A stable transient interval generation module is used to receive the initial transient interval, set the detection window to generate trend analysis results, and output the stable transient interval; The state recognition module is used to calculate the power fluctuation value and determine the maximum fluctuation value or the minimum fluctuation value according to the start and stop state of the equipment, and at the same time calculate the endpoint difference, and determine the start and stop state of the equipment according to the maximum fluctuation value and the minimum fluctuation value.

10. A load event identification system according to claim 9, characterized in that: The stable transient interval generation module also includes: a dynamic update unit, which is used to calculate the mean and standard deviation of the power difference in the detection window, set multiple threshold adjustment factors, and dynamically adjust the standard power range threshold using the threshold adjustment factor during each iteration.

Images