Data acquisition method and system for electrical cabinet

Abstract

The invention relates to the technical field of data processing, in particular to a data acquisition method and system for an electrical cabinet. The method comprises the following steps: constructing a function reference sequence based on operation data of the electrical cabinet, and obtaining a fitting function of data in the function reference sequence; constructing an adjacent reference sequence of the data collected in real time, and obtaining the error accumulation degree of error accumulation based on the difference between the measured value of the data in the reference sequence and the predicted value of the fitting function; in response to the fact that the error accumulation degree is larger than a preset deviation threshold value, the fitting function is obtained again based on the latest data. The method and the device have the effect of improving the accuracy of electrical cabinet anomaly detection.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to a data acquisition method and system for electrical cabinets. Background Art

[0002] An electrical cabinet is a device that integrates a variety of electrical components. Its main function is to monitor and control the power system, and it is the core control unit of the power system. Therefore, the normal operation of the electrical cabinet plays an important role in the power system. The most primitive detection methods for electrical cabinets mainly rely on manual inspections and threshold alarm methods. The method of manual inspections has poor real-time performance and is not conducive to reducing the labor burden of staff. Although the threshold alarm method replaces manual inspections, reduces the labor burden of staff, and improves the real-time performance of detection to a certain extent. However, this type of method only has high accuracy for data with long-term stable distribution, and has a low accuracy for detecting data with certain fluctuations. The least squares method is a mathematical optimization technique that can find the best matching function for data based on minimizing the sum of the squares of errors. According to the fitted function, the data of the equipment operation can be predicted, and the difference between the predicted data and the actual collected data can be used to monitor the operation state of the electrical cabinet. The curve fitted by the least squares method can reflect the overall change trend of the data, which is more accurate than the traditional threshold detection.

[0003] However, the fitted curve of the least squares method is a static model based on historical data. During the use of the electrical cabinet, with its own aging or the slow change of the external environment, the overall change trend of the electrical cabinet operation data is not fixed. Therefore, over time, the accuracy of the traditional least squares method fitting function detection method will decrease, resulting in misjudgments. At the same time, all the data collected during the detection process needs to be stored in full, which occupies more computing and storage resources, and the resource efficiency is low. Summary of the Invention

[0004] In order to solve the problems of decreasing detection accuracy and low resource efficiency in traditional detection methods, this application provides a data acquisition method and system for electrical cabinets.

[0005] In a first aspect, this application provides a data acquisition method for an electrical cabinet, adopting the following technical solution: A data acquisition method for an electrical cabinet, comprising the steps of: constructing a function reference sequence based on the operation data of the electrical cabinet, and obtaining a fitting function of the data in the function reference sequence; constructing a neighboring reference sequence of the real-time collected data, and obtaining an error accumulation degree of error accumulation based on the difference between the measured value of the data in the reference sequence and the predicted value of the fitting function; in response to the error accumulation degree being greater than a preset deviation threshold, re-obtaining a fitting function based on the latest data, and defining the latest obtained fitting function as a real-time function, and the fitting function before the real-time function as a historical function; Obtaining the abnormal degree of the real-time function; in response to the abnormal degree of the real-time function being greater than a preset abnormal threshold, marking the data used in the fitting process of the real-time function, the data used in the real-time function constitutes a data set, determining that the data set is abnormal, and using the data in the collected abnormal set as the original data for anomaly detection.

[0006] The beneficial effects are as follows: A function reference sequence is constructed according to the operation data of the electrical cabinet, and the function reference sequence is used to obtain a fitting function. After the function fitting is completed, as the real-time data is continuously collected, the error accumulation degree of the fitting function at the current moment is calculated. When the error of this fitting function accumulates to a certain extent, it indicates that the overall operation data of the electrical cabinet has changed to a certain extent due to factors such as the aging of the electrical cabinet and the change of the working environment. This fitting function can no longer accurately predict the operation data of the electrical cabinet, which may lead to a decrease in the accuracy of subsequent anomaly detection. Therefore, in this method, when the error accumulation degree is greater than the preset cheap threshold, the function reference sequence is reconstructed, and the fitting function is obtained to maintain the accuracy of the fitting function prediction, thereby ensuring the accuracy of subsequent anomaly detection. After obtaining a new fitting function, analyze the abnormal degree of the real-time function. If the real-time function is abnormal, it indicates that the data in the function reference sequence used in the fitting process of the real-time function is abnormal. Furthermore, the data in this function reference sequence can be marked as an abnormal data set, and the data in the abnormal data set is used as the original data for anomaly detection, thereby further improving the accuracy of anomaly detection; moreover, after such data is marked, the normal data can be cleaned up to optimize the storage space.

[0007] Optionally, taking any moment as the right endpoint, construct a time window with a length of <Object: to obtain all the data within the time window to form a function reference sequence.

[0008] The beneficial effects are as follows: For any moment of the collected data, set a time window according to the data before this moment, and obtain the data in the time window to complete the construction of the function reference sequence.

[0009] Optionally, the steps of constructing a neighboring reference sequence of the real-time collected data include: determining the reference sequence length , for the real-time collected data, taking the sampling moment corresponding to this data as the benchmark, obtain the data before this moment For a piece of data, the moment and the data points before this moment are used as the adjacent reference sequence.

[0010] The beneficial effect is that for any moment in the real-time collected data, multiple data adjacent to this moment before are used as the adjacent reference sequence, and the error accumulation within a certain time period is reflected by these data.

[0011] Optionally, the least squares method is used to obtain the fitting function corresponding to the data in the function reference sequence.

[0012] Optionally, the calculation steps of the error accumulation degree include: for each moment in the adjacent reference sequence, the local error is calculated according to the difference between the predicted value and the measured value of the fitting function at the same moment, and the ratio of the local error to the mean square error of the fitting function is used as the local error degree; the square root of the mean value of the local error degrees corresponding to each moment is used as the error accumulation degree.

[0013] The beneficial effect is that the difference between the predicted value and the measured value of the fitting function indicates the accuracy of the prediction of the fitting function for the equipment operation data. If this difference is too large, it means that there is a large deviation between the prediction of the fitting function and the measured value, and the change of the fitting function can no longer correctly represent the trend of the actual data change of the electrical cabinet. The local error degrees corresponding to each moment in the adjacent reference sequence are accumulated, so that the overall accumulated error degree can be obtained. In order to control the final value range, a square root operation is performed on the final calculation result.

[0014] Optionally, the steps to obtain the abnormality degree of the real-time function include: obtaining the trend difference between the real-time function and each historical function, obtaining the fitting frequency of the real-time function, and taking the product of the trend difference and the fitting frequency as the abnormality degree of the real-time function.

[0015] The beneficial effect is that after re-obtaining the fitting function based on the new function reference sequence, if the change trend of the real-time function and the historical fitting function is too large, it means that the change of the real-time function is abnormal. At the same time, if the fitting period of the real-time function is too short, that is, re-fitting is required within a short time, then it also means that the data is abnormal, and the irregular change of the data leads to frequent fitting of the real-time function. Based on the combined action of the trend difference and the fitting frequency on the abnormality degree of the real-time function, the accuracy of calculating the abnormality degree of the real-time function is improved.

[0016] Optionally, obtain the real-time fitting period of the real-time function, and take the ratio of the mean value of the historical fitting periods of all historical functions to the real-time fitting period as the fitting frequency.

[0017] The beneficial effects are as follows: During the operation of the electrical cabinet, if the overall data changes relatively smoothly due to the aging of the electrical cabinet or environmental changes, the fitting period of the real-time function should be relatively uniform and long. If the fitting period of the real-time function is significantly shorter than the previous fitting periods, it indicates that there are large fluctuations in the data used for the real-time function fitting. Therefore, the ratio of the average value of the historical fitting periods of all historical functions to the real-time fitting period is used as the fitting frequency. When the fitting period of the real-time function shortens, the fitting frequency increases.

[0018] Optionally, the steps of obtaining the trend difference between the real-time function and each historical function include: obtaining the function difference between each historical function and the real-time function, obtaining the abnormality degree based on each function difference, and the degree of difference is positively correlated with the function difference; performing weighted summation on each degree of difference based on the time interval between the historical function and the real-time function to obtain the total difference, and taking the normalized result of the total difference as the trend difference.

[0019] The beneficial effects are as follows: In some cases, there may be multiple historical functions before the real-time function. Therefore, the function differences between multiple historical functions and the real-time function are used to reflect the trend difference between the real-time function and the historical functions, improving the accuracy of calculating the trend difference of the real-time function.

[0020] Optionally, for any historical function, the steps of obtaining the function difference between it and the real-time function include: obtaining the absolute difference between the function values at the same moment of the historical function and the real-time function, and taking the accumulated result of multiple absolute differences as the function difference.

[0021] The beneficial effects are as follows: For the real-time function and the historical function, the difference between the two functions is determined according to their corresponding function values, thereby reflecting the difference in the change trends between the real-time function and the historical function.

[0022] In a second aspect, the present application provides a data acquisition system for an electrical cabinet, adopting the following technical solution: A data acquisition system for an electrical cabinet includes: a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, the above-mentioned data acquisition method for an electrical cabinet is implemented.

[0023] The beneficial effects are as follows: The above-mentioned data acquisition method for an electrical cabinet is generated into a computer program and stored in the memory to be loaded and executed by the processor. Thus, a system is made according to the memory and the processor, which is convenient to use.

[0024] The present application has the following technical effects: In the present application, a new fitting function, i.e., a real-time function, is dynamically obtained multiple times according to the error accumulation degree of the fitting function during the operation of the electrical cabinet. The real-time function is fitted based on the recent data of the electrical cabinet and is more in line with the current state and aging degree of the electrical cabinet; at the same time, the data is marked and classified for collection based on the abnormality degree of the real-time function. Normal data can be cleaned up, thereby optimizing storage. At the same time, in the present application, abnormality detection is performed based on the marked data to improve the accuracy of monitoring the electrical cabinet and maintain the stability of the accuracy of monitoring the operation of the electrical cabinet. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 FIG. is a flowchart of a data acquisition method for an electrical cabinet according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] An embodiment of the present application discloses a data acquisition method for an electrical cabinet, constructs a function reference sequence, and constructs a fitting function based on the data in the function reference sequence. The fitting function can predict the future data of the electrical cabinet. At the same time, there is a certain error between the predicted value and the measured value. Based on this error, the error accumulation degree of the real-time data is obtained to determine whether the fitting function is inaccurate. When the error accumulation degree of the fitting function is greater than a preset deviation threshold, a fitting function is constructed again, and the abnormality degree of the latest fitting function (real-time function) is obtained. Based on the abnormality degree, it is determined whether the function reference sequence corresponding to the real-time function is abnormal, and the initial classification and collection of data are completed. Subsequently, abnormality detection is performed on the abnormal data to complete the accurate judgment of the equipment operation state. In this method, a fitting function is dynamically constructed according to the error accumulation degree, and the real-time function is dynamically updated based on the update of the data, so as to avoid the situation that the sensitivity of the traditional static model to abnormal data decreases over time, resulting in a decrease in monitoring accuracy. At the same time, the abnormality degree of the real-time function is obtained here, which can classify the data. Abnormal data sets can be focused on, while normal data can be cleaned up to optimize the storage space.

[0027] Refer to Figure 1 , a data acquisition method for an electrical cabinet includes steps S1 - step S5.

[0028] S1: Based on the operation data of the electrical cabinet, construct a function reference sequence, and obtain the fitting function of the data in the function reference sequence.

[0029] During the operation of the electrical cabinet, collect the operation data of the electrical cabinet. The operation data may include: voltage, current, power, temperature, etc. In this embodiment, the data acquisition frequency is 30 min / time.

[0030] When the electrical cabinet starts to run, the operating data of the electrical cabinet is collected at a specific sampling frequency. At the beginning of the operation of the electrical cabinet, the amount of operating data generated is small, so fitting is not performed. As the data is continuously collected, when a certain amount of data is accumulated, a function reference sequence is constructed based on the operating data. In this embodiment, the length of the function reference sequence is 30. Then, when the data accumulation reaches 30, the first fitting can be performed to obtain the fitting function. The method for obtaining the fitting function is the least squares method. In other embodiments, other fitting methods can also be used to obtain the fitting function, such as the maximum likelihood estimation method.

[0031] S2: Construct a neighboring reference sequence for the real-time collected data, and obtain the error accumulation degree of error accumulation based on the difference between the measured value and the predicted value of the fitting function in the reference sequence.

[0032] Collect real-time data, construct a neighboring reference sequence based on the real-time data, and calculate the magnitude of the accumulated error in the current state. When the error is too large, it indicates that there is already a large deviation between the predicted result of the currently used fitting function and the measured value, and the fitting function needs to be obtained again according to the actual data.

[0033] In order to more accurately obtain the deviation degree between the predicted result of the current fitting function and the measured value, a neighboring reference sequence is constructed based on the neighboring data of the real-time data. Refer to the differences between the predicted values and the measured values at multiple moments in the neighboring reference sequence, so as to improve the robustness and accuracy of the final error accumulation degree calculation.

[0034] Specifically, the calculation steps of the error accumulation degree include: for each moment in the neighboring reference sequence, calculate the local error according to the difference between the predicted value and the measured value of the fitting function at the same moment, and take the ratio between the local error and the mean square error of the fitting function as the local error degree; take the square root of the mean of the local error degrees corresponding to each moment as the error accumulation degree.

[0035] For the predicted values and measured values corresponding to each moment in the neighboring reference sequence, obtain the difference between the predicted value and the measured value. In order to prevent the difference from being negative and to unify the dimension, the difference is squared here, and the square of the difference is used as the local error.

[0036] Take the ratio between the local error and the mean square error of the fitting function as the local error degree. The mean square error can be directly obtained during the least squares method fitting process and will not be elaborated here.

[0037] The closer the local error calculated in the neighboring reference sequence is to the mean square error in the function fitting process, the smaller the local error degree at the current moment. Sum the multiple local errors corresponding to the data in the neighboring reference sequence and calculate the average value, so as to reflect the overall error of the data in this neighboring reference sequence, that is, the error accumulation degree of the accumulated error at the current moment. To avoid excessive values, perform a square root operation on the overall error, and use the result of the square root of the overall error as the error accumulation degree.

[0038] Specifically, the calculation formula of the error accumulation degree can be expressed as: ; where represents the error accumulation degree at the moment corresponding to the th data; represents the number of data in the neighboring reference sequence; represents the function value corresponding to the th data in the neighboring reference sequence on the latest fitting function, represents the th data in the neighboring reference sequence; represents the mean square error of the fitting function.

[0039] During the calculation of the error degree represents the error between the predicted value of the real-time function and the measured value at the same moment. Squaring this error is to unify the dimension with the mean square error. Finally, perform a square root operation to control the final value range.

[0040] S3: In response to the error accumulation degree being greater than the preset deviation threshold, re-obtain the fitting function based on the latest data, and define the latest obtained fitting function as the real-time function, and the fitting function before the real-time function as the historical function.

[0041] In this embodiment, the deviation threshold is 1.1. When the finally calculated error accumulation degree is greater than the deviation threshold, reconstruct the function reference sequence based on the current data and obtain the latest fitting function. For the convenience of description, the latest obtained fitting function is defined here as the real-time function, and the fitting function before the real-time function is used as the historical function. It should be noted that during the actual operation of the electrical cabinet, the real-time function changes dynamically as the data accumulates, and the real-time function is not a fixed function.

[0042] S4: Obtain the abnormal degree of the real-time function.

[0043] Obtain the trend difference between the real-time function and each historical function.

[0044] Obtain the absolute difference between the function values at the same moment in the historical function and the real-time function, and use the accumulated result of multiple absolute differences as the function difference.

[0045] For a real-time function, it corresponds to a function reference sequence; for any historical function, it also corresponds to a function reference sequence.

[0046] Specifically, the calculation formula of function difference can be expressed as: Where, Indicates the Function difference between historical function and real-time function; Indicates the The first of the function reference sequence corresponding to the history function The function value corresponding to the data; Indicates the first function in the function reference sequence corresponding to the real-time function. The function value corresponding to the data; Indicates the length of the function reference sequence.

[0047] The abnormality degree is obtained based on the difference of each function, and the function difference is positively correlated with the abnormality degree; the total difference is obtained by weighted summing of the difference degrees based on the time interval between the historical function and the real-time function.

[0048] Specifically, the total difference is calculated as Where, Represents the total difference between the historical function and the real-time function; Indicates the total number of history functions; Indicates the The function difference between the historical function and the real-time function, Indicates the The time interval between the historical function and the real-time function; For An exponential function with base .

[0049] The step of obtaining the time interval in this embodiment includes: obtaining the data of the same sequence in the function reference sequence corresponding to the real-time function and the function reference sequence corresponding to the historical function, obtaining their index values, and using the absolute difference between their index values as the time interval between the historical function and the real-time function, that is, .

[0050] Indicates the degree of difference between the historical function and the real-time function. The greater the function difference, the smaller the denominator, and thus the greater the anomaly, achieving a positive correlation between function difference and anomaly.

[0051] is the weight part of the difference degree; middle The larger the value, the greater the distance between the historical function and the real-time function. Therefore, the mutual influence between the two functions is smaller, which can reduce the weight of the difference between the historical function and the real-time function.

[0052] The normalized result of the total difference is used as the trend difference.

[0053] Specifically, the formula for normalizing the total difference can be expressed as: ; where represents the total difference between the historical function and the real-time function; represents the trend difference between the real-time function and the historical function; represents the total number of historical functions; represents the th time interval between the historical function and the real-time function.

[0054] Obtain the fitting frequency of the real-time function, and take the product of the trend difference and the fitting frequency as the abnormality degree of the real-time function.

[0055] Obtain the real-time fitting period of the real-time function, and take the ratio of the mean value of the historical fitting periods of all historical functions to the real-time fitting period as the fitting frequency.

[0056] The real-time fitting period refers to the time interval for the real-time function to fit with an adjacent historical function. Here, the time interval is also the absolute difference of the index values of the data at the same ordinal position in the function reference sequences corresponding to the two functions.

[0057] The historical fitting period refers to the time interval for two adjacent historical functions to fit. The mean value of the historical fitting periods for all historical functions to fit illustrates the frequency of normally obtaining the fitting function under the condition of the aging of the electrical cabinet. By comparing the real-time fitting period with the historical fitting period, it is possible to determine whether the real-time function fits too fast.

[0058] S5: In response to the abnormality degree of the real-time function being greater than the preset abnormality threshold, mark the data used in the fitting process of the real-time function. The data used in the real-time function constitutes a data set. Determine that this data set is abnormal, and use the data in the collected abnormal data set as the original data for anomaly detection.

[0059] In this embodiment, the abnormality threshold is set to 0.77, and in other embodiments, it can be determined according to the actual situation. When the abnormality degree of the real-time function is greater than the abnormality threshold, it indicates that the data for fitting the real-time function is abnormal. Take the data set used to fit the real-time function as the abnormal data set; use this abnormal data set as the original data for anomaly detection to complete the anomaly detection of the operation of the electrical cabinet. Here, it is mainly used to obtain the data set containing abnormal data. The method for anomaly detection can be the LOF anomaly detection algorithm, etc. This anomaly detection algorithm is a conventional technical means in this field and will not be elaborated here.

[0060] An embodiment of the present application also discloses a data acquisition system for an electrical cabinet, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a data acquisition method for an electrical cabinet according to the present application is implemented.

[0061] The above system further includes other components well known to those skilled in the art, such as a communication bus and a communication interface. Their settings and functions are known in the art, so they will not be elaborated here.

[0062] The above are all the preferred embodiments of the present application. The protection scope of the present application is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A data acquisition method for an electrical cabinet, characterized in that, Including steps: constructing a function reference sequence based on the operation data of the electrical cabinet, and obtaining a fitting function for the data in the function reference sequence; constructing a neighboring reference sequence for the real-time collected data, and obtaining an error accumulation degree of error accumulation based on the difference between the measured value and the predicted value of the fitting function in the reference sequence; in response to the error accumulation degree being greater than a preset deviation threshold, re-obtaining a fitting function based on the latest data, and defining the latest obtained fitting function as the real-time function, and the fitting function before the real-time function as the historical function; obtaining the abnormality degree of the real-time function; in response to the abnormality degree of the real-time function being greater than a preset abnormality threshold, marking the data used in the fitting process of the real-time function, the data used in the real-time function constitutes a data set, determining that the data set is abnormal, and using the data in the collected abnormal data set as the original data for anomaly detection.

2. The data acquisition method for an electrical cabinet according to claim 1, wherein, The steps for constructing the function reference sequence include: constructing a time window with a length of using any moment as the right endpoint, and obtaining all the data within the time window to form the function reference sequence.

3. A data acquisition method for an electrical cabinet according to claim 1, characterized in that, The steps for constructing the adjacent reference sequence of the real-time collected data include: determining the reference sequence length , for the real-time collected data, taking the sampling moment corresponding to the data as a reference to obtain the data before this moment, and using this moment and the data points before this moment as the adjacent reference sequence.

4. A data acquisition method for an electrical cabinet according to claim 1, characterized in that, Using the least squares method to obtain the fitting function corresponding to the data in the function reference sequence.

5. A data acquisition method for an electrical cabinet according to claim 1, characterized in that, The calculation steps of the error accumulation degree include: for each moment in the neighboring reference sequence, calculating the local error according to the difference between the predicted value and the measured value of the fitting function at the same moment, and taking the ratio between the local error and the mean square error of the fitting function as the local error degree; taking the square root of the mean value of the local error degrees corresponding to each moment as the error accumulation degree.

6. The data acquisition method for an electrical cabinet according to claim 1, characterized in that, The steps of obtaining the abnormality degree of the real-time function include: obtaining the trend difference between the real-time function and each historical function, obtaining the fitting frequency of the real-time function, and taking the product of the trend difference and the fitting frequency as the abnormality degree of the real-time function.

7. A data acquisition method for an electrical cabinet according to claim 6, characterized in that, Obtaining the real-time fitting period of the real-time function, and taking the ratio of the mean value of the historical fitting periods of all historical functions to the real-time fitting period as the fitting frequency.

8. A data acquisition method for an electrical cabinet according to claim 6, characterized in that, The steps of obtaining the trend difference between the real-time function and each historical function include: obtaining the function difference between each historical function and the real-time function, obtaining the abnormality degree based on each function difference, and the difference degree is positively correlated with the function difference; performing weighted summation on each difference degree based on the time interval between the historical function and the real-time function to obtain the total difference, and taking the normalized result of the total difference as the trend difference.

9. A data acquisition method for an electrical cabinet according to claim 8, characterized in that, For any historical function, the steps of obtaining the function difference between it and the real-time function include: obtaining the absolute difference between the function values at the same moment in the historical function and the real-time function, and taking the accumulated result of multiple absolute differences as the function difference.

10. A data acquisition system for an electrical cabinet, characterized in that, Including: a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, it implements a data acquisition method for an electrical cabinet according to any one of claims 1-9.

Images