Intelligent electric energy meter data analysis method based on big data

Abstract

The invention relates to the field of electric energy meters, in particular to an intelligent electric energy meter data analysis method based on big data. The method comprises the following steps: acquiring preprocessed current data, voltage data and temperature data according to a preset sampling interval; constructing a target window, calculating voltage correlation according to the target window, calculating a voltage important factor, and similarly obtaining a current important factor according to a calculation method of the voltage important factor; calculating a temperature important factor; lagging characteristics of a target moment are calculated according to the important voltage factor, the important current factor and the important temperature factor, lagging characteristics of each sampling moment are obtained through traversal, the mode of the lagging characteristics is used as the optimal lagging characteristic, the input length is determined according to the optimal lagging characteristic, the input length is input into a preset network, an electricity consumption predicted value is output, and data analysis is completed. According to the technical scheme, the accuracy of the data analysis result of the electric energy meter can be improved.

Description

Technical Field

[0001] The present invention relates to the field of electric energy meters, and more specifically to a data analysis method for smart electric energy meters based on big data. Background Art

[0002] With the rapid development of the information society and the continuous advancement of intelligent technologies, the traditional power industry is also undergoing a continuous digital and intelligent transformation. Driven by smart grids, smart cities, and smart homes, the collection, monitoring, and analysis of power data are becoming increasingly important. As a key component of the intelligent transformation of the power industry, smart energy meters are responsible for recording and transmitting user electricity usage data in real time. They are one of the basic infrastructures for remote monitoring and data collection in smart grids. Using sensors and digital technology, smart energy meters can accurately measure electricity consumption and transmit this data via the network to a centralized processing platform.

[0003] The existing Chinese patent application document with publication number CN110598899A discloses a local electricity meter calibration service planning method and system based on data prediction, including: obtaining local electricity meter calibration historical data; using a prediction algorithm to calculate the local electricity meter calibration historical data to obtain future electricity meter calibration prediction data; and integrating the future electricity meter calibration prediction data to plan the electricity meter calibration service investment for each time period in each divided area.

[0004] However, when predicting electricity consumption based on temperature, the above application document uses a weighted method based on similar components of the same period in history to predict the periodic components according to the principle of large near and small far. It ignores the different importance of trends, seasonality, and residuals in the STL decomposition algorithm, resulting in low accuracy of the electricity meter data analysis results. Summary of the Invention

[0005] In order to solve the problem of low accuracy of electric energy meter data analysis results, the present invention proposes a smart electric energy meter data analysis method based on big data.

[0006] The present invention discloses a data analysis method for a smart electric energy meter based on big data, comprising: obtaining pre-processed current data, voltage data, and temperature data according to a preset sampling interval; taking any sampling moment as a target moment, constructing a target window, calculating voltage correlation according to the target window, calculating voltage importance factors based on voltage correlations of target windows with different window sizes, and obtaining current importance factors in a similar manner according to the calculation method of voltage importance factors; decomposing temperature data using an STL decomposition algorithm to calculate temperature importance factors; calculating hysteresis characteristics at the target moment according to the voltage importance factors, current importance factors, and temperature importance factors, traversing to obtain hysteresis characteristics at each sampling moment, taking the mode of the hysteresis characteristics as the optimal hysteresis characteristics, determining the length of the input according to the optimal hysteresis characteristics, inputting the data into a preset network, outputting a power consumption prediction value, and completing data analysis; the hysteresis characteristics satisfy the relationship: , Indicates target time The hysteresis characteristics of Indicates target time The voltage is an important factor, Indicates target time The important factor of current is Indicates target time Temperature is an important factor, represents the expansion coefficient, Indicates rounding up.

[0007] By analyzing the correlation between voltage and current, the solution can identify abnormal fluctuations and load change trends in the power system, thereby providing an accurate reference for power consumption forecasts. At the same time, using the STL decomposition method to decompose the trend, seasonality, and residuals of temperature data can effectively eliminate the periodic impact of temperature changes, making the temperature influencing factor more accurate, and working together with the voltage and current factors to calculate the hysteresis characteristics, thereby optimizing the timeliness and prediction accuracy of the model. The selection and weighted calculation of hysteresis characteristics not only ensures the model's responsiveness to changes in historical data, but also improves the stability and robustness of the forecast, and can better cope with complex factors such as power demand fluctuations and environmental changes in actual applications, ultimately achieving accurate power consumption forecasts.

[0008] Preferably, the preprocessing includes: using the mean filling method to fill in the missing current data, voltage data and temperature data, using the Z-score method to identify and process the outliers in the current data, voltage data and temperature data, and using the Kalman filter to reduce the noise in the current data, voltage data and temperature data.

[0009] The combination of the three not only improves data quality and eliminates the interference of missing and abnormal data, but also improves the signal-to-noise ratio of the data and the prediction accuracy of the model.

[0010] Preferably, the target window is constructed with the target time as the window end.

[0011] Preferably, the voltage correlation satisfies the relationship: , Indicates the window size is Target time The voltage dependence of Indicates the window size is Target time The voltage value, Indicates the window size is Target time The voltage value, represents the covariance, Indicates the window size is The voltage variance.

[0012] By calculating the covariance between the voltage at the target moment and the voltage at the historical moment and normalizing it to the variance of the voltage, the mutual dependence between the various time points in the voltage series is reflected.

[0013] Preferably, the voltage correlation further satisfies the relationship: , Indicates the window size is Target time The voltage dependence of Indicates the window size is Target time The voltage value, Indicates the window size is Target time The voltage value, represents the covariance, Indicates the window size is The voltage variance, represents the voltage standard deviation, Represents the exponential function.

[0014] Preferably, calculating the voltage importance factor includes: for any window size, calculating the voltage correlation variance and the voltage correlation mean in a sampling day, calculating the absolute difference between the voltage correlation at the target moment and the voltage correlation mean, and calculating the ratio of the absolute difference to the voltage correlation variance; traversing to obtain the ratio corresponding to each window size, normalizing the mean of all ratios, and taking the mean as the voltage importance factor at the target moment.

[0015] Preferably, the calculation of the voltage importance factor also satisfies the relationship: , Indicates target time The voltage is an important factor, Indicates the window size is Target time The voltage dependence of Indicates the window size is The mode of the voltage correlation when Indicates the number of window size types, Indicates the window size is The mean voltage correlation value is Indicates the window size is The voltage correlation variance is Represents the normalization function.

[0016] By introducing the difference between the voltage correlation at the target moment and the mode, it is possible to reflect the relative deviation between the target moment and the overall behavior of the data on a longer timeline, and then identify abnormal fluctuations or unstable factors that may exist in the data at that moment.

[0017] Preferably, the temperature importance factor includes: constructing the temperature data obtained according to the preset sampling interval into a temperature sequence, and using the STL decomposition algorithm to obtain the trend, seasonality and residual of the temperature sequence; weighting the ratio of the trend at the target moment to the maximum trend in the temperature sequence as the trend factor, obtaining the seasonal factor and the residual factor according to the calculation method of the trend factor, and normalizing the sum of the trend factor, seasonal factor and residual factor as the temperature importance factor.

[0018] It can reflect the relative relationship between current temperature changes and overall climate trends, thereby identifying possible abnormal or extreme temperature fluctuations. At the same time, seasonal factors and residual factors take into account the contribution of seasonal changes and short-term fluctuations to temperature changes, further enhancing the model's sensitivity and accuracy to temperature behavior on different time scales.

[0019] Preferably, the preset network is an LSTM network, which uses one or more LSTM layers to capture the timing characteristics of the input data. The input data is current data, voltage data and temperature data of the same length as the optimal lag characteristic. The output of the LSTM layer will be passed to the fully connected layer to generate a power consumption forecast value.

[0020] Beneficial effects of the present invention: The present invention first ensures data quality by filling missing values, identifying outliers, and reducing noise through refined preprocessing steps. In the feature extraction stage, the important factors of voltage and current are analyzed for their correlation and combined with the statistical characteristics of different window sizes to capture subtle changes in the power system; the influence of temperature is extracted through the STL decomposition method to extract trend, seasonality and residual components, further refining the analysis of temperature factors. By calculating various types of lag features and applying LSTM network models, this method can not only mine deep relationships in time series data, but also accurately predict future electricity consumption by learning historical data patterns over long time spans. This method can better cope with complex power load fluctuations and improve the accuracy of power system scheduling and data analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 This is a flow chart of a method for analyzing smart electricity meter data based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0023] It should be understood that when the terms "first," "second," and the like are used in the claims, description, and drawings of the present invention, they are merely used to distinguish between different objects, rather than to describe a specific order. The terms "comprise" and "comprising" used in the description and claims of the present invention indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof.

[0024] The present invention provides a data analysis method for smart electric energy meters based on big data. Figure 1 As shown, the data analysis method of the smart electricity meter based on big data includes steps S1 to S3, which are described in detail below.

[0025] S1, obtaining pre-processed current data, voltage data and temperature data according to a preset sampling interval.

[0026] In one embodiment, the preset sampling interval can be set by those skilled in the art. Exemplarily, the present invention sets the preset sampling interval to 15 minutes.

[0027] The mean filling method is used to fill the missing current data, voltage data and temperature data. The Z-score method is used to identify and process outliers in the current data, voltage data and temperature data. The Kalman filter is used to reduce the noise in the current data, voltage data and temperature data.

[0028] By filling missing data with the mean of each variable, the impact of incomplete data on subsequent analysis and model training is avoided, ensuring data continuity and consistency. Secondly, the Z-score method effectively handles outliers by standardizing the data and identifying data points that deviate significantly from the normal range, preventing extreme data from adversely affecting analysis results and ensuring that the model can make accurate predictions based on reasonable assumptions. Finally, the Kalman filter smoothes the measured data, reducing random noise from sensors or the environment, making current, voltage, and temperature data more accurate and stable. The combination of these three methods not only improves data quality and eliminates the interference of missing and anomaly data, but also enhances the signal-to-noise ratio of the data and the prediction accuracy of the model.

[0029] S2: Build a target window, calculate the voltage correlation based on the target window, calculate the voltage importance factor based on the voltage correlation of target windows of different window sizes, and obtain the current importance factor in the same way as the calculation method of the voltage importance factor.

[0030] In one embodiment, any sampling moment is used as the target moment, and the target moment is used as the window end to construct a target window.

[0031] Calculate the voltage correlation, which satisfies the relationship: , Indicates the window size is Target time The voltage dependence of Indicates the window size is Target time The voltage value, Indicates the window size is Target time The voltage value, represents the covariance, Indicates the window size is The voltage variance.

[0032] It should be noted that by calculating the covariance between the voltage at the target moment and the voltage at the historical moment and normalizing it to the voltage variance, the interdependence between the various time points in the voltage sequence is reflected. Specifically, the voltage correlation Indicates that at a given window size The similarity of voltage change trends between the target moment and its previous historical moments is measured. A higher correlation value indicates a strong similarity between the current voltage value and the past voltage values, indicating that the system voltage changes relatively smoothly or has a strong time dependence at that moment. Conversely, a lower correlation may indicate a more drastic voltage change and possible system instability.

[0033] The voltage importance factor is calculated based on the voltage correlation of target windows with different window sizes. The voltage importance factor includes: for any window size, calculating the voltage correlation variance and voltage correlation mean in a sampling day, calculating the absolute difference between the voltage correlation and the voltage correlation mean at the target time, and calculating the ratio of the absolute difference to the voltage correlation variance; traversing to obtain the ratio corresponding to each window size, and normalizing the mean of all ratios to obtain the voltage importance factor at the target time.

[0034] Expressed as formula: , Indicates target time The voltage is an important factor, Indicates the window size is Target time The voltage dependence of Indicates the number of window size types, Indicates the window size is The mean voltage correlation value is Indicates the window size is The voltage correlation variance is Represents the normalization function.

[0035] It should be noted that by first calculating the voltage correlation variance and mean, the stability and regularity of voltage over different time periods can be assessed. Next, by comparing the deviation between the voltage correlation at the target moment and its mean, the degree of abnormality or volatility in voltage behavior at that moment is reflected. Calculating the ratio of this deviation to the correlation variance quantifies the relationship between the voltage abnormality and overall volatility at the target moment. This ratio reflects the degree of deviation of the voltage at a local moment compared to the overall behavior and serves as a sensitivity indicator to voltage state changes. By traversing the ratios of different window sizes and performing normalization, a comprehensive voltage importance factor can be obtained, reflecting the fluctuation characteristics and degree of abnormality of voltage at multiple scales.

[0036] For example, according to the exemplary preset sampling interval in step S1, 96 voltage data points can be obtained in one sampling day. If the window size is set to 3, a maximum of 32 windows can be obtained in one sampling day; if the window size is set to 6, a maximum of 16 windows can be obtained in one sampling day; if the window size is set to 12, a maximum of 8 windows can be obtained in one sampling day. Similarly, the voltage correlation under different window sizes can be calculated, thereby calculating the voltage importance factor.

[0037] The current importance factor is obtained in the same way as the voltage importance factor.

[0038] In one embodiment, the voltage dependency further satisfies the relationship: , Indicates the window size is Target time The voltage dependence of Indicates the window size is Target time The voltage value, Indicates the window size is Target time The voltage value, represents the covariance, Indicates the window size is The voltage variance, represents the voltage standard deviation, Represents the exponential function.

[0039] Will The hysteresis weights for each window size demonstrate that shorter time intervals result in stronger autocorrelation, as short-term changes tend to be more consistent. As time intervals increase, autocorrelation may weaken. Therefore, using hysteresis weights, data points with shorter time intervals are given greater weight when calculating autocorrelation, while data points with longer time intervals are given smaller weights. This more accurately reflects the correlation characteristics of voltage data at different time intervals.

[0040] In one embodiment, the voltage importance factor also satisfies the relationship: , Indicates target time The voltage is an important factor, Indicates the window size is Target time The voltage dependence of Indicates the window size is The mode of the voltage correlation when Indicates the number of window size types, Indicates the window size is The mean voltage correlation value is Indicates the window size is The voltage correlation variance is Represents the normalization function.

[0041] It should be noted that It can be the mode of multiple sampling days or a year. By introducing the difference between the voltage correlation at the target moment and the mode, it can reflect the relative deviation between the target moment and the overall behavior of the data on a longer time line, and then identify the abnormal fluctuations or unstable factors that may exist in the data at that moment.

[0042] S3, decompose the temperature data using the STL decomposition algorithm to calculate the temperature important factors.

[0043] It should be noted that STL decomposition is a decomposition technique for time series data, capable of breaking down time series into three components: trend, seasonality, and residual. The trend reflects the long-term pattern of temperature changes, ignoring seasonal fluctuations. The residual is the portion remaining after removing the trend and seasonality, which may be due to climate anomalies. Seasonality refers to the regular temperature fluctuations caused by seasonal changes throughout the year, such as hot summers, cold winters, and mild spring and autumn.

[0044] When the trend component, seasonal component and residual component of the temperature series are larger, it means that the seasonal dependence of the temperature data is stronger. The weight of the trend component should be the smallest, the weight of the seasonal component should be the largest, and the weight of the residual component should be moderate.

[0045] In one embodiment, the temperature data obtained according to the preset sampling interval is constructed into a temperature series, and the trend, seasonality and residual of the temperature series are obtained using the STL decomposition algorithm; the ratio of the trend at the target moment to the maximum trend in the temperature series is weighted as the trend factor, and the seasonal factor and residual factor are obtained according to the calculation method of the trend factor, and the sum of the trend factor, seasonal factor and residual factor is normalized as the important temperature factor.

[0046] The weight of the trend is The seasonal weight is expressed as The weight of the residual is expressed as Indicates that the weights satisfy For example, The value is 0.2, Take the value 0.5, The value is 0.3.

[0047] S4, calculates the hysteresis characteristics of the target moment based on the voltage important factor, current important factor and temperature important factor, traverses to obtain the hysteresis characteristics of each sampling moment, takes the mode of the hysteresis characteristics as the optimal hysteresis characteristics, determines the input length based on the optimal hysteresis characteristics, inputs it into the preset network, outputs the power consumption prediction value, and completes the data analysis.

[0048] In one embodiment, the hysteresis characteristic satisfies the relationship: , Indicates target time The hysteresis characteristics of Indicates target time The voltage is an important factor, Indicates target time The important factor of current is Indicates target time Temperature is an important factor, represents the expansion coefficient, Indicates rounding up.

[0049] The expansion factor is used to adjust the maximum length of the lag feature, not exceeding 32. This 32 is obtained by dividing the total number of data points in a sampling day (96) by the sum of the normalized values ​​in the brackets and the maximum value of 3. For example, the expansion factor is 10.

[0050] The lag features at each sampling moment are obtained by traversing, and the mode of the lag features is used as the optimal lag feature. The input length is determined based on the optimal lag feature. It should be noted that the lag feature represents the relationship between the data point at the target moment and the data point at a certain moment in history. The length or number of lag features refers to the selected number of lag time steps. In electricity consumption forecasting, if the lag feature is selected as 1, the electricity meter data at the target moment will be used as input features together with the historical electricity meter data from the previous moment; if the lag feature is selected as 3, the data at the target moment will be used as input features together with the data from the previous three moments.

[0051] The input network is a LSTM network, which uses one or more LSTM layers to capture the temporal characteristics of the input data. The input data consists of current, voltage, and temperature data of the same length as the optimal lag feature. The output of the LSTM layer is passed to the fully connected layer to generate the power consumption forecast. The loss function uses the mean square error loss function. Common optimizers include Adam and RMSProp, which can automatically adjust the learning rate during training to ensure model convergence.

[0052] For example, if the optimal lag characteristic is 9, the length of the input data is the target moment and 9 historical moments starting from the target moment, a total of 10 moments of electricity meter data are used as input to predict the electricity consumption data at the next moment after the target moment.

[0053] While this specification has shown and described several embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

[0054] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A smart energy meter data analysis method based on big data, characterized in that: include: Acquire preprocessed current data, voltage data and temperature data according to a preset sampling interval; Taking any sampling moment as the target moment, constructing the target window, calculating the voltage correlation according to the target window, calculating the voltage importance factor based on the voltage correlation of the target windows with different window sizes, and obtaining the current importance factor in the same way according to the calculation method of the voltage importance factor; Decompose the temperature data using the STL decomposition algorithm to calculate the important temperature factors; The hysteresis characteristics of the target time are calculated according to the voltage important factor, the current important factor and the temperature important factor, and the hysteresis characteristics of each sampling time are obtained by traversing. The mode of the hysteresis characteristics is used as the optimal hysteresis characteristics. The input length is determined according to the optimal hysteresis characteristics, and input into the preset network. The power consumption prediction value is output to complete the data analysis; The hysteresis characteristics satisfy the relationship: , Indicates the target time The hysteresis characteristics of Indicates the target time The voltage is an important factor, Indicates the target time The important factor of current is Indicates the target time Temperature is an important factor. represents the expansion factor, Indicates rounding up.

2. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The pre-processing comprises: The mean filling method is used to fill the missing current data, voltage data and temperature data. The Z-score method is used to identify and process the outliers in the current data, voltage data and temperature data. The Kalman filter is used to reduce the noise in the current data, voltage data and temperature data.

3. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: Build the target window with the target time as the window end.

4. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The voltage correlation satisfies the relationship: , Indicates the window size is Target time The voltage dependence of Indicates the window size is Target time The voltage value, Indicates the window size is Target time The voltage value, represents the covariance, Indicates the window size is The voltage variance.

5. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The voltage dependency also satisfies the relationship: , Indicates the window size is Target time The voltage dependence of Indicates the window size is Target time The voltage value, Indicates the window size is Target time The voltage value, represents the covariance, Indicates the window size is The voltage variance, represents the voltage standard deviation, Represents an exponential function.

6. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: Calculating the voltage importance factor includes: For any window size, the voltage correlation variance and voltage correlation mean in a sampling day are calculated, the absolute difference between the voltage correlation at the target time and the voltage correlation mean is calculated, and the ratio of the absolute difference to the voltage correlation variance is calculated; The ratio corresponding to each window size is obtained by traversing, and the mean of all ratios is normalized as the important voltage factor at the target moment.

7. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The calculation of the voltage importance factor also satisfies the relationship: , Indicates the target time The voltage is an important factor, Indicates the window size is Target time The voltage dependence of Indicates the window size is The mode of the voltage correlation when Indicates the number of window size types, Indicates the window size is The voltage correlation mean value is Indicates the window size is The voltage correlation variance is Represents the normalization function.

8. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The important temperature factors include: The temperature data obtained at the preset sampling intervals are constructed into a temperature series, and the trend, seasonality and residual of the temperature series are obtained using the STL decomposition algorithm; The ratio of the trend at the target moment to the maximum trend in the temperature series is weighted as the trend factor. The seasonal factor and residual factor are obtained according to the calculation method of the trend factor. The sum of the trend factor, seasonal factor and residual factor is normalized as the important temperature factor.

9. The method for analyzing smart electric energy meter data based on big data according to claim 1, characterized in that: The preset network is an LSTM network, which uses one or more LSTM layers to capture the timing characteristics of input data. The input data is current data, voltage data, and temperature data of the same length as the optimal lag characteristics. The output of the LSTM layer will be passed to the fully connected layer to generate a power consumption forecast value.

Images