A data acquisition method based on an electric energy meter

Abstract

The present invention relates to the field of data processing, and specifically relates to a data acquisition method based on an electric energy meter. By deploying intelligent electric energy meter acquisition devices in the workshop to collect equipment power consumption data and transmitting the data to the central database, an improved ARIMA model is used to predict future electricity consumption based on the preprocessed data. The model introduces relative fluctuation sensitivity, equipment status correction, and anomaly significance index on the basis of the traditional ARIMA to achieve dynamic suppression of prediction residuals. After comparing the final predicted value with the actual power consumption data, when the deviation exceeds the preset threshold, the system automatically triggers an alarm to notify the maintenance personnel in a timely manner to check and repair equipment anomalies. The present invention realizes real-time prediction and early warning of abnormal power consumption of industrial equipment by collecting data through intelligent electric energy meters, combining an improved ARIMA model and dynamic residual suppression.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and particularly to a data acquisition method based on an electric energy meter. Background Art

[0002] With the continuous improvement of industrial automation level, the real-time monitoring and status assessment of production equipment play a crucial role in ensuring workshop safety and improving production efficiency. In an industrial workshop, various production equipment is equipped with electric energy meters to record power consumption data in real time. Under normal circumstances, the power consumption during equipment operation shows stable and periodic characteristics; however, once the equipment fails or is affected by external interference, its power consumption will experience sudden changes and abnormal surges, which may lead to safety accidents or production interruptions. Therefore, how to accurately judge the equipment operation status through real-time acquisition and predictive analysis of power data has become an urgent technical problem to be solved.

[0003] In the prior art, the ARIMA (Autoregressive Integrated Moving Average) model is often applied to power consumption data prediction because it can better capture the trends and periodicities of time series data. This model first determines the data stationarity through the ADF test, and then uses methods such as autoregression, differencing, and moving average to establish a mathematical model to predict future power consumption. However, the ARIMA model has certain limitations: it assumes that the data series is stationary and linear. When encountering non-linear mutations such as equipment failures, the prediction effect is often not ideal, and the model fitting accuracy is easily affected by abnormal data. In addition, relying solely on linear modeling of historical data is difficult to reflect the instantaneous fluctuations caused by changes in equipment status, thereby affecting the timeliness of fault warning. Summary of the Invention

[0004] Aiming at the problem that the above ARIMA model is easily affected by abnormal data and affects the model fitting accuracy, the present invention proposes a data acquisition method based on an electric energy meter, including: obtaining historical power consumption data of an industrial workshop, and training an ARIMA prediction model based on the historical power consumption data; obtaining a real-time power consumption time series of a set length and inputting it into the trained ARIMA prediction model to obtain a predicted value; calculating the difference between the predicted value and the real-time power consumption, and in response to the difference being greater than a set threshold, the system notifies the maintenance personnel to conduct a fault check; the ARIMA prediction model also includes processing the original residuals to obtain abnormal dynamic suppression residuals , , where represents the original residual at time ; represents the natural exponential function; represents the time Anomaly significance index; the anomaly significance index is positively correlated with the corrected fluctuation sensitivity and the absolute value of the original residual, and negatively correlated with the root mean square of the prediction error of a set number of power data; the corrected fluctuation sensitivity is , , where represents the relative fluctuation sensitivity; represents the moment at which the load level is collected; represents the median of the load levels corresponding to all the collection moments; is the difference between adjacent loads; the relative fluctuation sensitivity is positively correlated with the difference between the power at the corresponding collection moment and the power at the previous collection moment, and negatively correlated with the standard deviation of the power at multiple continuously adjacent collection moments.

[0005] By constructing an improved ARIMA model based on the historical data and real-time data of the electricity meter, the present invention realizes accurate prediction and anomaly detection of the power data in the industrial workshop. Compared with the existing solutions that only adopt the traditional ARIMA model or simple threshold comparison, the present invention can effectively suppress the abnormal residuals generated by sudden load changes, thereby greatly improving the accuracy and real-time performance of fault warning, ensuring more timely equipment maintenance and more effective preventive measures, and significantly reducing the risks of safety accidents and economic losses.

[0006] Further, the method for obtaining the relative fluctuation sensitivity is specifically as follows:

[0007] ;

[0008] where represents the relative fluctuation sensitivity; represents the moment at which the power data is collected; represents the moment at which the power data is collected; represents the set sliding window length; represents the moment at which the power data is collected; represents the previous sliding average of the time points.

[0009] By adopting a sliding window and comparing with the previous mean value, the present invention captures the minute fluctuations of the power data in real time, can detect potential anomalies earlier than the traditional methods, improves the sensitivity of data processing and the robustness of the prediction model, provides more accurate basic data for the subsequent calculation of the anomaly significance index, and significantly improves the problems of slowness and misjudgment in anomaly capture in the prior art.

[0010] Further, the method for obtaining the sliding average is specifically as follows:

[0011] ;

[0012] wherein represents the moving average value of the previous time points; represents the set length of the moving window; represents the time collected power data.

[0013] Furthermore, the calculation method of the anomaly significance index is:

[0014] ;

[0015] wherein represents the anomaly significance index at time ; represents the corrected fluctuation sensitivity at time ; represents the original residual at time ; represents the length of the set error window; represents the original residual at time ; represents the root mean square of the prediction error of the previous time points.

[0016] By calculating the anomaly significance index and comprehensively considering the corrected fluctuation sensitivity, the original residual, and the root mean square of the prediction error, the present invention realizes the quantitative evaluation of the anomaly degree of the power data. Compared with the traditional ARIMA model that only relies on the residual judgment method, the present invention can effectively distinguish normal fluctuations from real anomalies, thereby realizing the precise adjustment of the anomaly dynamic suppression residual and improving the accuracy and reliability of the equipment fault warning in the industrial workshop.

[0017] Furthermore, obtaining the historical power data of the industrial workshop further includes: deploying smart meters in the industrial workshop to collect the power consumption of production equipment and setting the collection frequency range as 1 - 3 times / 10 min; transmitting the power consumption data to the workshop database through the RS485 communication interface; collecting a set number of continuous power from the workshop database to obtain the historical power data.

[0018] By deploying smart meters in the industrial workshop, using the RS485 communication interface to transmit data, and extracting continuous power records from the workshop database, the present invention ensures the frequency and continuity of data collection. Compared with the situation where data collection in the traditional method is not real-time or comprehensive enough, the present invention provides a set of high-precision and high-reliability power data collection solutions, laying a solid data foundation for the training of subsequent prediction models and anomaly detection.

[0019] Further, it also includes obtaining the load level using a power load tester PQ5200.

[0020] Further, the ARIMA prediction model also includes using the ADF test to check the stationarity of the historical electricity data, using the autocorrelation function and partial autocorrelation function graphs to determine the autoregressive order and moving average order, using the maximum likelihood estimation method to fit the model, and using the Ljung-Box test for noise.

[0021] By adopting a series of strict statistical methods such as the ADF test, ACF and PACF graph analysis, maximum likelihood estimation, and Ljung-Box test, the present invention ensures the scientific determination of model parameters and the high precision of model fitting.

[0022] Further, the value range of the set threshold is 15 - 25.

[0023] The technical effects of the present invention are as follows:

[0024] The present invention constructs an intelligent anomaly detection method based on electricity meter data acquisition, and innovatively improves the traditional ARIMA model. Specifically, by collecting real-time and historical electricity data of industrial workshops, while using the ARIMA model for electricity prediction, key indicators such as relative fluctuation sensitivity, equipment status correction, and anomaly significance index are introduced to achieve dynamic suppression of prediction residuals. This method comprehensively considers the periodic characteristics during normal operation of equipment and sudden load changes under abnormal conditions. Through multi-dimensional data processing, it significantly improves the sensitivity and accuracy of anomaly detection, effectively distinguishes normal fluctuations from fault warnings, thereby providing more timely and accurate warning information for equipment maintenance, reducing potential safety risks, and optimizing production scheduling management. Description of the Drawings

[0025] By reading the following detailed description with reference to the drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understandable. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 It is a flowchart of a data acquisition method based on electricity meters in an embodiment of the present invention shown schematically. Detailed Embodiments

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0028] The following will describe in detail the specific implementation manners of the present invention with reference to the accompanying drawings.

[0029] Embodiment of the concentrator operation data management method:

[0030] As Figure 1 shown, the concentrator operation data management method of the present invention includes:

[0031] S1. Obtain the power consumption data of the production equipment in the industrial workshop based on the electric energy meter.

[0032] As the core device for recording power consumption, the data of the electric energy meter can reflect the operation status and energy consumption of the equipment. In an industrial workshop, when the production equipment is operating normally, the power consumption usually remains within a stable range. However, when the equipment fails or is affected by external interference, the power consumption of some equipment may suddenly increase, causing the overall power consumption data to deviate from the normal range. Such anomalies may indicate equipment failures or potential safety hazards. Therefore, when collecting data on production equipment based on the electric energy meter in this embodiment, the predicted power consumption of the production equipment is obtained through data prediction means and compared with the actual power consumption to obtain abnormal data, avoiding greater losses or safety accidents caused by the long-term abnormal operation of the equipment. It not only ensures real-time data monitoring and rapid response to abnormal events in the industrial workshop, but also provides a decision-making basis for production plans for administrators, effectively improving the intelligent level of equipment management in the industrial workshop.

[0033] First, intelligent electric meters can be deployed in the industrial workshop to collect power consumption. In this embodiment, the collection frequency of power consumption can be set to 1 time / 10 min, and then the cumulative power consumption data is regularly transmitted to the workshop database through the communication interface (such as RS485 or Modbus protocol) of the electric energy meter. In addition to being deployed in the industrial workshop, the above intelligent electric meters can also be deployed on large-scale production equipment or key production equipment to monitor the relevant data of a specific equipment. Pull a total of historical power consumption data and denote them as , there are:

[0034] ;

[0035] Among them, represents the power consumption data collected at time ; represents the The power consumption data at a collection moment, where It can be set to the empirical value 1000. It should be noted that: since the collection frequency is set to 1 time / 10 min in this embodiment, then a total of 144 data are collected per day accordingly. Here, a total of 1000 historical data are collected, including data for 7 consecutive days, covering as much data of a week as possible, thus providing sufficient data sources for subsequent data prediction. Implementers can also, according to the shift schedule of the industrial workshop and the working status of production equipment, Set it to empirical values such as 2000, 3000, or 4000.

[0036] S2. Obtain the relative fluctuation sensitivity based on the power consumption data near any historical collection moment; correct the relative fluctuation sensitivity according to the equipment status to obtain the corrected fluctuation sensitivity; obtain the anomaly significance index to quantify the significance of the anomaly, and calculate the final anomaly dynamic suppression residual based on the anomaly significance index to obtain an improved ARIMA model.

[0037] In the industrial workshop scenario of this embodiment, with the orderly progress of the production plan, the fixed running time of each device, and the fixed working time of workers, the power consumption of the industrial workshop on each working day will not vary greatly. That is to say, in the long run, there is a certain periodic regularity in the power consumption data of the industrial workshop. Therefore, considering the time series characteristics of the data collected by the electricity meters in the industrial workshop, the ARIMA (Autoregressive Integrated Moving Average) model can be used in this embodiment for power consumption data prediction.

[0038] The ARIMA model can capture the trends and periodicity of the data. It includes the autoregressive order , the differencing order , and the moving average order . This algorithm first judges the stationarity of the sequence through the ADF (Augmented Dickey-Fuller) test, and the differencing order when the sequence is stationary is ; then determines and through the autocorrelation function (ACF) and partial autocorrelation function (PACF) plots; finally, uses the maximum likelihood estimation method to fit the model. The general form of the ARIMA model is:

[0039] ;

[0040] where is the lag operator; is the autoregressive polynomial; ; is white noise. By way of example, if there is a fitting result of a time series as ARIMA(1,1,0), then this model can be expressed as: .

[0041] S2.1. Obtain the relative fluctuation sensitivity based on the power consumption data near any historical collection moment.

[0042] Although the ARIMA model performs well in predicting the power consumption of industrial workshops, it assumes that the time series is stationary and linear and cannot effectively capture the non - linear mutations caused by equipment failures or external interferences, such as sudden surges in power consumption. However, these mutations are closely related to anomalies. Therefore, in this embodiment, an abnormal dynamic suppression residual is constructed to enhance the model's sensitivity and robustness to anomalies.

[0043] For the historical power consumption data of industrial workshops it has the following characteristics: when each device is operating normally, the power consumption shows a certain periodic pattern on weekdays and rest days; but when a device fails, the power consumption will show a short - duration and isolated surge, and the anomaly is usually closely related to the operating state of the device.

[0044] To capture the mutation characteristics, an index can be designed to quantify the dynamic change trend of power consumption, highlighting the anomaly amplitude and speed during anomalies, as follows:

[0045] ;

[0046] where represents the relative fluctuation sensitivity; represents the power consumption data collected at time ; represents the power consumption data collected at time ; represents the set sliding window length, which can be set to the empirical value of 10 in this embodiment; represents the power consumption data collected at time ; , represents the moving average of the previous time points; represents the local standard deviation of the previous time points, reflecting the normal fluctuation level near time .

[0047] represents the sensitivity of the change in the power consumption data collected at time relative to the historical fluctuation level. When normal, is small and the local standard deviation is stable, then is also small; when abnormal, suddenly increases, then also increases significantly, highlighting the mutation characteristics.

[0048] S2.2. Modify the relative fluctuation sensitivity according to the device state to obtain the modified fluctuation sensitivity.

[0049] The above relative fluctuation sensitivity can capture the power data that may be mutated in the power time series, but does not consider the impact of equipment status on anomalies. That is, when there is a sudden high load or load mutation generated by some equipment in the industrial workshop, it causes fluctuations in the power time series. At this time, based on the fluctuation sensitivity the obtained mutated power data may be caused by equipment anomalies rather than normal fluctuations. Therefore, by introducing equipment status information, the mutation sensitivity can be further corrected as follows:

[0050] ;

[0051] where represents the corrected fluctuation sensitivity; represents the relative fluctuation sensitivity; , which represents the load change; represents the logarithmic function, which is used to non-linearly amplify the load change; represents the moment the collected equipment load level; represents the reference load when the equipment is operating normally. In this embodiment, the median of all historical load levels of the equipment can be used as the reference load.

[0052] It should be noted that: when the electricity meter only collects the power data of a single equipment, the load level here refers to the single equipment; and when the power data collected by the electricity meter is for the entire industrial workshop, the load level here , where represents the th equipment load level in the industrial workshop. Correspondingly, represents the sum of the reference loads of all equipment in the industrial workshop. And the above load level can be obtained through power measurement equipment such as a power load tester (PQ5200). The specific acquisition process will not be elaborated here.

[0053] When the equipment operates normally according to the reference load, and is small. At this time ; while when the load surges ( ) or the load mutates ( is large), increases as a whole. At this time is amplified, indicating that the mutation is more likely to be an anomaly.

[0054] S2.3. Obtain the anomaly significance index to quantify the significance of the anomaly, and calculate the final anomaly dynamic suppression residual based on the anomaly significance index to obtain an improved ARIMA model.

[0055] For the moment The collected power data , the predicted value based on the ARIMA model is , and there are corresponding residuals . To quantify the significance of anomalies to ultimately achieve dynamic suppression of model residuals, an anomaly significance index is designed in this embodiment , and the calculation method is specifically as follows:

[0056] ;

[0057] where represents the anomaly significance index at time ; represents the corrected fluctuation sensitivity at time ; represents the residual at time ; represents the length of the error window, which can be set to the empirical value 5 in this embodiment represents the residual at time ; represents the root mean square error (RMSE) of the prediction errors of the previous time points, which can reflect the stability of the prediction. When the industrial workshop equipment is operating normally, the prediction error is small, the RMSE is stable, and at this time is small; when an anomaly occurs, suddenly increases, and is large. At this time is significantly amplified, highlighting the anomaly characteristics

[0058] Thus, the anomaly significance index at time is obtained, and finally, the anomaly dynamic suppression residual of the ARIMA model is calculated based on the anomaly significance index , specifically as follows:

[0059] ;

[0060] where represents the anomaly dynamic suppression residual at time ; represents the original residual at time in the initial ARIMA model; represents the natural exponential function; represents the anomaly significance index at time ;

[0061] When the power time series is normal, is small. At this time is close to 0, and there is , retain the normal residual information in the time series; when there may be abnormalities in the power time series, the anomaly significance index increases. At this time is close to 0, and the original residual is suppressed, that is , reducing the interference of abnormal phenomena on model training, making the model training more focused on the normal mode, thereby improving the prediction ability for future mutations.

[0062] Exemplary illustration: During the model training process, assume that the total power of the equipment in the industrial workshop at noon every day is 100 kW, that is, the power consumption data per hour at noon increases by 100 kW·h. Then under normal circumstances, the prediction result of the ARIMA model will increase at a rate of 100 per hour, that is, the residual value of the prediction remains around 0 and can pass the Ljung-Box test; but there may be a situation where the total power increases to 110 kW due to equipment failure on a certain day or for several consecutive days, resulting in a gradual increase in the residual (the residual of the power consumption data per hour at noon may be around 10). At this time, due to the small change in power, the model can still identify the periodic pattern, that is, the subsequent model may consider the change in power consumption to be normal, thus adjusting the parameters to adapt to this abnormal change, resulting in a deterioration of the prediction ability of the subsequent model for normal power consumption; and the abnormal dynamic suppression of residuals in this embodiment can suppress the influence of abnormal phenomena. For example, in the above example, the residual 10 can be adjusted to decrease or even approach 0, thereby retaining the normal model training information, making the model training more focused on the normal training data. The training model obtained by abnormal dynamic suppression of residuals can reflect the power consumption data under normal conditions, so that the state of the equipment can be evaluated based on the difference between the predicted value and the actual value of the power meter power consumption.

[0063] S3. When collecting power consumption data based on the power meter, complete the prediction of the power consumption in the industrial workshop according to the improved ARIMA model, and evaluate the equipment status based on the predicted power consumption data and the actual power consumption data.

[0064] The historical power consumption data obtained based on step S1 is input into the ARIMA model improved by abnormal dynamic suppression of residuals in step S2 for training. This model dynamically suppresses the model residuals by introducing relative fluctuation sensitivity, equipment status correction, and anomaly significance index, enhancing the ability to capture sudden anomalies.

[0065] Collect the time series of real-time power consumption data of the electricity meter. In this embodiment, the length of the time series of real-time power consumption data can be set to 150. Implementers can also set the length of the time series of real-time power consumption data according to the production scheduling plan of the industrial workshop and the data acquisition frequency of the electricity meter to ensure that there is enough original data to obtain a more accurate prediction value. Then, input the collected time series of real-time power consumption data into the trained ARIMA model to output the predicted value at the current time. Next, compare the real-time power consumption data with the predicted value. When the error exceeds the preset threshold, it is judged that the device may be abnormal (such as a fault, load mutation, etc.), and then the early warning mechanism is triggered. In this embodiment, the preset threshold can be set to the empirical value of 20. Finally, the early warning signal can be timely notified to the on-site maintenance personnel by means of text messages or mobile APPs, etc., to guide them to conduct on-site inspections and troubleshooting.

Claims

1. A data collection method based on an electric energy meter, characterized in that: The method comprises: Obtain historical power data of the industrial workshop, and train the ARIMA prediction model based on the historical power data; obtain the real-time power time series of a set length and input it into the trained ARIMA prediction model to obtain a predicted value; calculate the difference between the predicted value and the real-time power, and in response to the difference being greater than a set threshold, the system notifies the maintenance personnel to conduct a fault inspection; The ARIMA forecasting model also includes processing the original residual to obtain abnormal dynamic suppression residual , ,in Indicates time The original residual of represents the natural exponential function; Indicates time The abnormal significance index of The abnormal significance index is positively correlated with the modified fluctuation sensitivity and the absolute value of the original residual, and negatively correlated with the root mean square of the prediction error of the set amount of electricity data; the modified fluctuation sensitivity is , ,in Indicates relative volatility sensitivity; Indicates time The load level collected; Represents the median of the load level corresponding to all acquisition moments; is the difference between adjacent loads; The relative fluctuation sensitivity is positively correlated with the difference between the power at the corresponding collection moment and the power at the previous collection moment, and negatively correlated with the standard deviation of the power at multiple consecutive collection moments; The method for obtaining the relative fluctuation sensitivity is specifically as follows: ； in Indicates relative volatility sensitivity; Indicates time Collected power data; Indicates time Collected power data; Indicates the set sliding window length; Indicates time Collected power data; Before The sliding average of time points; The calculation method of the abnormal significance index is: ； in Indicates time The abnormal significance index of Indicates time Modified volatility sensitivity; Indicates time The original residual of Indicates the length of the set error window; Indicates time The original residual of Before The root mean square of the prediction error at each time point.

2. The data collection method based on electric energy meter according to claim 1 is characterized in that: The method for obtaining the sliding average value is specifically as follows: ； in Before The sliding average of time points; Indicates the set sliding window length; Indicates time Collected power data.

3. The data collection method based on electric energy meter according to claim 1 is characterized in that: Obtain historical power data of industrial workshops, including: Deploy smart meters in industrial workshops to collect electricity consumption of production equipment and set the collection frequency range to 1 to 3 times / 10 minutes; Transmit power consumption data to the workshop database via RS485 communication interface; A set number of continuous power quantities are collected from the workshop database to obtain historical power quantity data.

4. The data collection method based on electric energy meter according to claim 1 is characterized in that: It also includes obtaining the load level using the power load tester PQ5200.

5. The data collection method based on electric energy meter according to claim 1 is characterized in that: The ARIMA forecasting model also includes using ADF to test the stationarity of the historical electricity data, using autocorrelation function and partial autocorrelation function graph to determine the autoregressive order and the moving average order, using maximum likelihood estimation method to fit the model, and using Ljung-Box to test noise.

6. The data collection method based on electric energy meter according to claim 1 is characterized in that: The value range of the set threshold is 15~25.

Images