Intelligent power grid line loss intelligent detection method based on internet of things big data

Abstract

The application discloses an intelligent power grid line loss intelligent detection method based on Internet of Things big data, relates to the technical field of intelligent power grid electric energy metering, and comprises the following steps: S1, extracting voltage signals and current signals of each electric energy metering node within a set time window, performing frequency domain transformation to obtain a frequency spectrum, calculating an energy aggregation degree index of a high-frequency frequency band, and identifying a high-frequency harmonic interference time section according to the energy aggregation degree change. The application constructs a closed-loop mechanism of high-frequency harmonic interference identification, error tracing, cause discrimination and dynamic correction, realizes error compensation at the segment level and the node level, combines electric quantity consistency verification and trend model updating, improves the accuracy and stability of intelligent power grid line loss anomaly identification, and significantly enhances the reliability and intelligent level of the system in a complex interference environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent power grid electric energy metering, and particularly relates to an intelligent power grid line loss intelligent detection method based on Internet of Things big data. BACKGROUND

[0002] The "intelligent power grid line loss intelligent detection based on Internet of Things big data" refers to deploying Internet of Things sensing terminals at key nodes of the power grid, collecting multi-dimensional data including current, voltage, power factor, power consumption behavior and the like in real time, gathering, cleaning and deeply mining a large amount of spatiotemporal distributed power grid operation information by means of a big data platform, dynamically identifying the power loss position, loss type (technical or management) and change trend in the power transmission process by using intelligent algorithms such as machine learning and pattern recognition. This method breaks through the limitations of traditional line loss calculation relying on manual meter reading and static statistics, realizes real-time perception, accurate evaluation and dynamic early warning of the line loss of the power grid, and provides technical support for energy efficiency optimization and abnormal power consumption monitoring of the intelligent power grid.

[0003] The prior art has the following disadvantages:

[0004] In the process of line loss monitoring and energy consumption evaluation of the intelligent power grid, the smart meter as a key data acquisition terminal directly affects the identification accuracy of the line loss data and the reliability of subsequent decision-making of the system. However, in the actual operating environment, the electric energy metering chip inside part of the smart meter is prone to attenuation of anti-interference ability or misjudgment of electric signal characteristics under long-term exposure to high-frequency harmonic interference environment, especially in the area where high-power nonlinear loads (such as electric welding equipment, electric vehicle fast charging devices, frequency converters, etc.) frequently operate, the metering chip may be affected by non-power frequency current components, resulting in misjudgment of effective power. Such interference has the characteristics of suddenness, locality and intermittence, etc., resulting in random measurement deviation of the metering chip in a short period of time. Even if the single error amplitude is small, the long-term superposition will also cause systematic deviation to the line loss calculation result, thereby interfering with the error compensation mechanism of the line loss model, and even causing false alarm, misidentification of stealing electricity behavior and the like consequences.

[0005] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The application aims to provide an intelligent power grid line loss intelligent detection method based on Internet of Things big data, construct a closed-loop mechanism of high-frequency harmonic interference identification, error tracing, cause discrimination and dynamic correction, realize error compensation at the segment level and node level, combine power consistency verification and trend model updating, improve the accuracy and stability of intelligent power grid line loss anomaly identification, and significantly enhance the reliability and intelligent level of the system in a complex interference environment, so as to solve the problems in the background technology.

[0007] In order to achieve the above-mentioned purpose, the application provides the following technical scheme: an intelligent power grid line loss intelligent detection method based on Internet of Things big data, comprising the following steps:

[0008] S1, extracting the voltage signal and current signal of each electric energy metering node within a set time window, performing frequency domain transformation to obtain a frequency spectrum, calculating an energy aggregation degree index of a high-frequency band, and identifying a high-frequency harmonic interference time section according to the energy aggregation degree change;

[0009] S2, based on the identified high-frequency harmonic interference time section, calling a historical standard active power curve template of the corresponding electric energy metering node, fitting the error between the current active power curve and the template, and extracting an active power deviation vector;

[0010] S3, inputting the active power deviation vector, the terminal device type, the instantaneous load change characteristic and the high-frequency harmonic spectrum characteristic into an abnormal cause classification model, outputting whether the power deviation is caused by high-frequency harmonic interference on the electric energy metering chip, and marking an abnormal electric energy error segment;

[0011] S4, for the abnormal electric energy error segment, generating a high-frequency harmonic interference correction factor according to the time duration interval, the numerical offset direction and the amplitude, and constructing an electric energy metering node-level short-period data correction weight sequence;

[0012] S5, based on the correction weight sequence, performing segment weighting correction on the original electric energy metering data, calculating an electric quantity consistency verification matrix between a plurality of electric energy metering nodes, and overall correcting systematic errors;

[0013] S6, inputting the corrected data into a line loss analysis module, updating the early warning judgment threshold and error tolerance boundary in the line loss trend model, realizing closed-loop correction of electric energy metering interference data, and realizing real-time stable identification and response of line loss anomalies.

[0014] Preferably, step S1 comprises:

[0015] Within the set time window, the voltage signal and the current signal of the electric energy metering node are continuously collected, and window function weighting processing is performed to eliminate the influence of spectral leakage;

[0016] Performing fast Fourier transform on the weighted signal, obtaining voltage spectrum and current spectrum, and calculating frequency domain energy accumulation value of high frequency band;

[0017] Calculating the frequency domain energy aggregation index to form time series data;

[0018] Based on the dynamic threshold and the duration of the disturbance, the time section of the high-frequency harmonic interference is identified, and the interference event log is generated for subsequent error compensation.

[0019] Preferably, step S2 comprises:

[0020] Based on the identified high-frequency harmonic interference time section, the active power measurement value recorded continuously by the corresponding electric energy metering node between the start time and the end time of the high-frequency harmonic interference is extracted, the current active power curve is constructed, and the active power curve is normalized;

[0021] Calling the standard active power curve template under the historical operating condition of the corresponding node;

[0022] Time sequence fitting is performed on the current active power curve and the standard template, dynamic time warping algorithm or window alignment strategy is used for shape alignment, and point-by-point power difference value is calculated;

[0023] The statistical features of the power difference value sequence are extracted, and the active power deviation vector is constructed to quantify the influence of high-frequency harmonic interference on the metering node.

[0024] Preferably, step S3 comprises:

[0025] Extracting the active power deviation vector, the terminal device type, the instantaneous load change feature and the high-frequency harmonic spectrum feature;

[0026] The obtained features are uniformly input into the trained electric energy anomaly cause classification model, and the model is constructed based on a supervised learning algorithm;

[0027] Outputting the classification result of whether the power deviation is caused by high-frequency harmonic interference on the electric energy metering chip;

[0028] Marking the time section corresponding to the classification result as an abnormal electric energy error segment, and taking it as the basis for generating subsequent correction parameters.

[0029] Preferably, step S4 comprises:

[0030] Extracting the time duration interval, the numerical offset direction and the numerical offset amplitude of the abnormal electric energy error segment;

[0031] Based on the offset direction and the offset amplitude, the corresponding correction factor is generated, and the correction gradient is set according to the time distribution in the segment;

[0032] Map the correction factor to the abnormal segment and divide it into short period time segments;

[0033] Assign a correction weight in each time segment to build a short period data correction weight sequence at the node level of the electric energy metering.

[0034] Preferably, step S5 includes:

[0035] Based on the short period data correction weight sequence, the abnormal electric energy error segment in the original electric energy metering data is corrected at the segment level, and a corrected power sequence is generated;

[0036] The corrected power data of each electric energy metering node is converted into equal period electric quantity data, and an electric quantity consistency check matrix between multiple metering nodes is built;

[0037] Based on the deviation degree of the electric quantity difference between nodes in the electric quantity consistency check matrix, the correction weight of the corresponding node is adjusted to generate new corrected electric quantity data;

[0038] The checking and correction process is repeated until the overall consistency error of the system converges to within the tolerance threshold range.

[0039] Preferably, step S6 includes:

[0040] The corrected electric energy metering data is collected according to the power grid topology, and the basic line loss data sequence of each power supply unit is recalculated;

[0041] The corrected basic line loss data sequence is input into the line loss trend model to replace the original data of the corresponding time period and update the model parameters;

[0042] Based on the updated line loss trend model, the line loss anomaly early warning judgment threshold and the error tolerance boundary are adjusted;

[0043] The updated judgment result is integrated with the line loss analysis process to form a closed loop correction and real-time response regulation mechanism for electric energy metering data.

[0044] In the above technical solution, the technical effects and advantages provided by the present application are:

[0045] The application can realize a closed-loop processing mechanism from period identification of high-frequency harmonic interference, error tracing, cause discrimination to refined correction and consistency verification of whole network data, effectively solving the problems of power misjudgment, line loss misidentification and false alarm caused by interference in the traditional electric energy metering system under complex electromagnetic environment. The method not only has multi-dimensional feature fusion interference perception ability, but also introduces dynamic correction factors and weight distribution strategies to realize error compensation at the segment level and node level, and through the electric quantity consistency matrix and trend model updating mechanism, the accuracy and stability of the corrected data in the whole network analysis are comprehensively guaranteed. Finally, the smart grid can still maintain high sensitivity, high accuracy and sustainable response ability to line loss abnormalities under complex operation scenes such as high-frequency interference and nonlinear load concentration, improving the reliability, precision and intelligent level of the line loss analysis system. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0047] Figure 1 The method flowchart of the smart grid line loss intelligent detection method based on Internet of Things big data of the present application. DETAILED DESCRIPTION

[0048] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive gist to those skilled in the art.

[0049] The present application provides a smart grid line loss intelligent detection method based on Internet of Things big data as shown in Figure 1 The method flowchart of the smart grid line loss intelligent detection method based on Internet of Things big data of the present application.

[0050] S1, extract the voltage signal and current signal of each electric energy metering node in a set time window, perform frequency domain transformation to obtain voltage spectrum and current spectrum, calculate the frequency domain energy aggregation degree index in the high frequency band, and identify the time section with high frequency harmonic interference according to the frequency domain energy aggregation degree change;

[0051] In view of the adverse effects of high-frequency harmonic interference in the smart grid on electric energy metering accuracy, a high-frequency harmonic disturbance identification method based on frequency domain energy aggregation degree analysis is proposed to realize time positioning and dynamic marking of high-frequency electromagnetic disturbance, which specifically includes the following steps:

[0052] For each electric energy metering node, the original waveform data of the voltage signal and the current signal in the corresponding time period is continuously extracted within a set time window (e.g. 1 second, 5 seconds or 10 seconds). The above waveform data should be obtained from the electric energy metering terminal through synchronous sampling to ensure the alignment accuracy of the voltage signal and the current signal in the time dimension. All data should meet the minimum sampling frequency of not less than 10 kHz to ensure sufficient high-frequency resolution in the frequency domain analysis. In order to avoid the spectral leakage problem caused by edge effect, the original waveform data needs to be preprocessed by a window function (such as Hamming window, Blackman window, etc.) before entering the frequency domain transformation process to enhance the stability of the spectral calculation.

[0053] In order to avoid the spectral leakage phenomenon caused by signal truncation when performing frequency domain transformation on the original waveform data, i.e. the diffusion of spectral energy at non-real frequencies, which affects the accurate identification of high-frequency interference components, the original waveform data should be weighted by a window function before frequency domain transformation. In specific implementation, the original voltage signal and the current signal in the selected time window are first sampled to obtain data segments with fixed length. Then, a window function with edge attenuation characteristics (such as Hamming window or Blackman window) is selected, and the window function is multiplied point by point with each segment of sampled data in an equal length manner. This operation can be regarded as smoothing the transition at the beginning and end of the waveform while retaining the main energy of the signal, thereby suppressing the spectral spreading effect introduced by non-periodic truncation in the subsequent frequency domain conversion process. Through the introduction of the window function, the energy positioning accuracy of each frequency component in the spectrum can be effectively improved, and the stability and reliability of the frequency energy accumulation analysis can be enhanced.

[0054] The processed voltage signal and the current signal are subjected to Fast Fourier Transform (FFT) respectively to obtain the corresponding voltage spectrum and current spectrum, and the amplitude information of each frequency component is extracted. In the spectral data, a specific high-frequency frequency band is defined as the harmonic interference evaluation interval, for example, the frequency interval of 150 Hz to 2 kHz, which covers the range of high-order harmonics generated by typical power electronic devices. Based on this high-frequency interval, the voltage spectrum and the current spectrum are subjected to integral processing respectively to calculate the frequency energy accumulation value in the frequency band. In order to further improve the discrimination of interference positioning, the normalized energy ratio (Normalized High-Frequency Energy Ratio, N-HFER) is introduced as the frequency energy accumulation index, which is defined as the ratio of the energy integral value in the high-frequency interval to the total energy value in the full frequency band. This index can objectively reflect the concentration degree of high-frequency energy in the overall signal, and then identify the sudden high-frequency harmonic disturbance.

[0055] The calculated frequency domain energy concentration index is constructed into sequence data over time, and the time window boundary is combined to identify the period when the frequency domain energy concentration is significantly higher than the normal benchmark. In order to enhance the robustness of the identification, the dynamic mean and standard deviation reference curve of the energy concentration is constructed based on the sliding window method, and it is judged whether the concentration index of a certain time window exceeds the dynamic threshold (such as the mean plus twice the standard deviation) as the trigger condition of the existence of high-frequency harmonic interference. In addition, in order to avoid misjudgment of accidental disturbance, a minimum disturbance duration threshold (such as two consecutive time windows triggering) is set to ensure that the identified disturbance time period has sufficient time domain continuity, thereby forming a stable interference segment marker.

[0056] Based on the above trigger criterion, the time section of the occurrence of high-frequency harmonic disturbance is marked and archived to form the interference event log on the time sequence. Each interference event log should include its start and end time, frequency domain energy concentration peak, interference intensity level, and the number and sampling identification of the corresponding electric energy metering node. The interference event log can be used for subsequent electric energy metering deviation analysis and error correction model calling, and can be used as an important basis for power deviation judgment and error attribution in the interference environment. In addition,

[0057] Through the above frequency domain energy concentration analysis and interference period positioning mechanism, the automatic identification and time period positioning of high-frequency harmonic disturbance in a complex electromagnetic environment are realized, which significantly improves the accuracy of abnormal segment identification in electric energy metering data, and provides stable and reliable basic data support for the subsequent smart grid line loss error compensation mechanism.

[0058] The purpose of this step is to establish a high-sensitivity, high-resolution high-frequency interference detection mechanism for subsequent electric energy metering error identification and correction. The core purpose is to accurately identify whether there is high-frequency harmonic disturbance in the environment of the electric energy metering node in the smart grid from the frequency domain perspective. Since high-frequency harmonic interference is usually caused by nonlinear loads (such as frequency converters, electric vehicle fast charging devices, etc.), its disturbance characteristics are sudden, local and aperiodic, and it is difficult to identify directly from the time domain signal. Therefore, the voltage signal and current signal are obtained within a set time window, the frequency spectrum characteristics are extracted through frequency domain transformation, and the energy distribution of the high-frequency band is focused on. By introducing the frequency domain energy concentration index, the proportion of high-frequency energy in the overall signal is quantitatively analyzed, which can effectively measure the interference intensity and its trend. When the high-frequency energy concentration in a certain period is significantly higher than the normal operating level, it can be preliminarily determined that there is external harmonic interference in that period. This step not only provides a clear time positioning basis for subsequent power deviation calculation and error attribution, but also significantly enhances the system's perception ability of abnormal interference events in a complex electromagnetic environment, and is the basic link and prerequisite for realizing dynamic error correction and accurate line loss analysis.

[0059] S2, based on the identified time section in which high-frequency harmonic interference exists, calling the standard active power curve template corresponding to the current electric energy metering node under the historical operating condition, fitting the error between the current active power curve and the standard active power curve, and extracting the active power deviation vector of the current electric energy metering node under the interference;

[0060] For the identified time section in which high-frequency harmonic interference exists, in order to realize the quantification and positioning of the deviation of electric energy metering data under the influence of interference, an error fitting method based on the standard active power curve template is constructed, which accurately reflects the disturbance degree of high-frequency interference on the metering accuracy by extracting the active power deviation vector, including the following steps:

[0061] In the time section that has been identified as existing high-frequency harmonic interference, the active power measurement values recorded continuously by the corresponding electric energy metering node in the time section are extracted to construct the current active power curve. The curve takes time as the horizontal axis and the active power output by the electric energy metering chip every period as the vertical axis, forming a power change curve with time sequence characteristics. In order to improve the fitting accuracy, the extracted data should be subjected to linear interpolation processing to ensure the consistency of the curve under different sampling frequencies. In addition, in order to facilitate alignment with the template data, the current curve is normalized to make its dimension uniform and have stronger fitting comparability.

[0062] From the running database of the platform to which the electric energy metering node belongs, the historical active power curve template of the node under normal operating conditions is called. The selection of the template curve needs to consider the terminal device type corresponding to the electric energy metering node, the running time section (such as weekdays and non-weekdays), the load level, and the historical load characteristics, etc. dimensions, from which the most suitable standard active power change mode for the current running scene is matched. The template curve can be generated through historical big data clustering analysis and typical curve extraction mechanism, and has the characteristics of strong representativeness, stable time sequence, and clear shape, etc. It is a key reference for evaluating the deviation degree of the current curve.

[0063] The generation of the template curve can be realized by using clustering analysis and typical curve extraction mechanism based on a large amount of historical active power data accumulated in the electric energy metering platform. The specific steps are as follows:

[0064] The preliminarily screened and classified active power curves collected historically are classified according to the terminal device type to which the electric energy metering node belongs, the running time section, the load level, etc. to ensure that the input data have similarity.

[0065] By using time series clustering algorithm (such as K-means dynamic time warping variants, DBSCAN, etc.), the curves with similar power change trend are aggregated into several sub-class clusters, and in each sub-class cluster, one or more most representative typical curves are extracted as the standard template under this running state through center curve extraction, mean curve extraction or morphological matching optimization.

[0066] The template curve has obvious time sequence stability, clear load change rule and less error interference, so it can be used as a reliable benchmark for fitting comparison in practical application. Its role is to provide a reference form for error identification of the current running power curve, so as to effectively determine whether there is an abnormal deviation and assist in judging whether high-frequency interference has a substantial impact on the measurement accuracy. It is the key basis in the entire error traceability analysis chain.

[0067] By using the time series fitting error based method, the current active power curve and the standard active power curve template are analyzed for error fitting. In this scheme, the dynamic time warping algorithm (Dynamic Time Warping) or window alignment fitting strategy is preferably used to align the shapes of the two power curves on the time axis, and the deviations between the power values at each time point are calculated point by point to construct a power deviation sequence. To enhance the ability to describe the impact of disturbance, further extract multiple statistical features such as peak deviation, average deviation and deviation volatility rate from the deviation sequence to comprehensively constitute the active power deviation vector. This vector can be regarded as the deviation signature of the power curve of the measurement node under the influence of high-frequency harmonic interference, which quantitatively reflects the degree and form of interference.

[0068] The implementation process of fitting analysis includes the following key steps:

[0069] The current active power curve and the selected standard active power curve template are normalized respectively to ensure the consistency of the two curves in numerical dimension and time scale, so as to eliminate the deviation influence brought by different running scenarios or measurement devices.

[0070] The dynamic time warping algorithm (Dynamic Time Warping, DTW) is used to align the two curves on the time axis, that is, by constructing a distance matrix, comparing the distances between the sampling points of the curves, and finding the alignment path with the smallest cumulative distance among all possible paths, the morphological matching of unequal length or non-synchronous fluctuation data is realized. If the curve length is consistent or the synchronization requirement is higher, the window alignment fitting strategy can be used to calculate the amplitude difference between the two curves at the same time position point by point through setting a fixed length sliding window. After alignment, the power value difference between the current curve and the template curve at each time point is extracted to construct a power deviation sequence that changes with time.

[0071] The sequence truly reflects the deviation of the power measurement curve from the normal reference curve in the high-frequency interference section, and is the basis for subsequent construction of power deviation vector, judgment of abnormal power error causes and generation of correction parameters. Through this refined fitting method, the dynamic deviation of the power measurement data under the influence of high-frequency disturbance can be described in detail, and the accuracy of error positioning and the intelligence of disturbance response are enhanced.

[0072] The extracted active power deviation vector is in one-to-one correspondence with the interference time section, and is stored in a structured way in the deviation traceability data set for subsequent use by the abnormal cause discrimination module. The deviation vector not only contains the overall deviation characteristics of the disturbance section, but also retains the dynamic response mode of the power curve under disturbance, which can provide key quantitative basis for judging whether the disturbance is caused by the abnormal response behavior of the power measurement chip. In addition, this step bypasses the drawbacks of traditional absolute threshold-based judgment by template matching and fitting, avoids misjudgment caused by the complexity of power grid load changes, has good adaptability and non-obvious judgment strategy, and significantly enhances the traceability analysis capability of the power measurement system in complex interference environment.

[0073] The core role of this step is to quantitatively analyze and extract the active power deviation of the power measurement node under the influence of high-frequency harmonic interference, providing accurate and traceable data basis for subsequent abnormal error identification and dynamic correction. In the actual operation of the smart grid, affected by the complex power environment, the power measurement data may deviate for a short time during the interference period, but whether these deviations are caused by high-frequency harmonic interference and whether they have a systematic impact are often difficult to identify by simply determining the absolute value. Through this step, the active power change curve of the current power measurement node is first extracted in the identified interference time section, and compared with the standard active power curve template formed by the node under historical operating conditions, which can effectively build a reference baseline that conforms to the characteristics of the node and represents normal operation behavior. On this basis, the deviation between the current power curve and the standard template on the time axis and amplitude axis is identified through fitting analysis, and multiple statistical features of the power deviation vector are extracted, such as peak error, mean error, and fluctuation frequency, thereby realizing the structured description of abnormal behavior of power measurement under interference. The deviation vector not only quantifies the degree of abnormality of the current data, but also serves as an input feature for the subsequent abnormal classification model to assist in determining whether the error is caused by the distortion of the measurement chip. In summary, this step plays a key role in converting disturbance perception into quantitative expression in the overall method system, and is a bridge link from "finding disturbance" to "identifying error nature", significantly improving the accuracy, interpretability and practicality of power measurement error analysis.

[0074] S3, inputting the active power deviation vector, the terminal device type corresponding to the current electric energy metering node, the instantaneous load change characteristic, and the high-frequency harmonic spectrum characteristic into an electric energy anomaly cause classification model, and outputting a classification result of whether the power deviation is caused by the high-frequency harmonic interference on the electric energy metering chip, and marking as an abnormal electric energy error segment;

[0075] For the extracted active power deviation vector, in order to further determine whether the deviation is caused by the abnormal response behavior of the electric energy metering chip under the high-frequency harmonic interference environment, an electric energy anomaly cause classification model based on multi-feature fusion is proposed. By constructing a discrimination mechanism with device type, load dynamic characteristic and spectrum characteristic as input, intelligent identification of power deviation characteristics and accurate marking of abnormal error segments are realized. The specific embodiment includes the following steps:

[0076] The terminal device type information corresponding to the current node in the electric energy metering system is extracted as a static prior attribute affecting the electric energy fluctuation characteristic and input into the classification model. Different types of terminal devices (such as electric welding devices, industrial air conditioners, electric vehicle charging piles, etc.) have obviously different load characteristics and anti-interference ability when running, which may cause different forms of spectrum disturbance and power curve fluctuation. Therefore, the device type is set as an independent input feature in the model, which helps to improve the adaptability and recognition accuracy of the model in multiple scenarios.

[0077] The instantaneous load change characteristic in the current interference time segment is extracted, mainly including load current change rate, voltage fluctuation amplitude, active power rise / fall gradient, etc., as time-sensitive features reflecting the dynamic response behavior of the load. These features can be obtained by calculating the first or second derivative of the electric energy parameters in the continuous time segment, combined with the burstiness of the disturbance and the nonlinear load characteristics, which helps to determine whether the power deviation is caused by the abnormal operation of the load itself or whether it has obvious external interference characteristics.

[0078] Based on the voltage spectrum and current spectrum obtained by the foregoing frequency domain transformation, the high-frequency harmonic spectrum characteristic is extracted, including the energy proportion of high-order harmonic components, the peak frequency position of harmonics, the spectrum change rate, etc. These features can be used to describe the spectrum form and intensity change of the interference signal, as an important basis for distinguishing the source of interference. By jointly inputting the spectrum characteristic and the power deviation vector, the recognition ability of the model for the specific causal path of "the metering chip misjudgment caused by harmonic interference" can be enhanced.

[0079] The active power deviation vector, terminal device type, instantaneous load change feature, and high-frequency harmonic spectrum feature are input into the trained power anomaly cause classification model. The model can be constructed based on a supervised learning algorithm (such as gradient boosting tree, random forest, or lightweight neural network), and the cause classification result of the power deviation is output. If the model judges that the power deviation is highly related to high-frequency harmonic interference and does not conform to the logic of the load itself, the corresponding time segment is marked as "abnormal power error segment under harmonic interference influence". This segment marking will serve as the basis for subsequent error correction parameter generation, and at the same time, a complete closed-loop path from interference identification, error quantification to cause attribution is established at the data level.

[0080] To build a power anomaly cause classification model based on a supervised learning algorithm, a high-quality training sample set is first needed, which is composed of labeled active power deviation events in historical power measurement data. Each sample contains four types of input features: active power deviation vector, terminal device type, instantaneous load change feature, and high-frequency harmonic spectrum feature. The label is the cause type of the corresponding deviation event (such as "normal load fluctuation", "high-frequency harmonic interference", "measurement chip failure", etc.). Then, the structured sample data is normalized, encoded (such as one-hot encoding of device type), and cleaned to ensure the uniformity and effectiveness of the input features. Subsequently, a suitable supervised learning algorithm is selected, such as gradient boosting tree, which can handle nonlinear and heterogeneous feature data and has strong interpretability. Random forest has strong robustness and anti-overfitting ability, while lightweight neural network is suitable for large-scale and diverse data scenarios. During model training, cross-validation and hyperparameter tuning are used to improve model generalization ability. After training, the model performance is evaluated by precision, recall, F1 score, and other indicators, and the cause classification results output by the model are logically verified and reviewed by experts to ensure their stability and credibility in practical applications. Finally, the model can be integrated into the power data processing platform to realize automatic cause identification, dynamic labeling, and response driving of real-time power deviation data.

[0081] The core role of this step is to classify the active power deviation phenomenon occurring in the electric energy metering process through an intelligent discrimination mechanism, to determine whether the deviation is caused by the abnormal response behavior of the electric energy metering chip under high-frequency harmonic interference environment, so as to realize accurate labeling of abnormal electric energy error segments. In the actual power grid operation environment, active power deviation may be caused by many factors, such as normal load fluctuation, device start-stop, power grid voltage disturbance, and external high-frequency harmonic interference. Only the numerical deviation of power data cannot accurately determine the cause, which is easy to cause false alarm, false correction, and even misidentify electricity stealing behavior. Therefore, by building an electric energy abnormal cause classification model integrating multi-dimensional features, different sources of abnormal behavior can be effectively distinguished at the data level. Specifically, the model takes the active power deviation vector as the core behavior feature, the terminal device type as the static attribute of the operating environment, the instantaneous load change feature as the dynamic response index, and the high-frequency harmonic spectrum feature as the external interference label, forming a complete feature input vector. With the help of a supervised learning model (such as gradient boosting tree, random forest or lightweight neural network), the deviation cause is classified and identified. When the model output indicates that the deviation is caused by the metering error caused by high-frequency interference, the current time period is marked as an “abnormal electric energy error segment”, providing accurate target range for subsequent correction mechanism. Through this step, an intelligent decision-making path from power deviation perception to cause analysis is established, breaking the logical chain of interference identification, error diagnosis and data correction, significantly improving the understanding ability and response accuracy of the electric energy metering system to complex interference environment. It is the key link connecting perception and processing in the entire line loss intelligent detection system.

[0082] S4, for the abnormal electric energy error segment, according to its time duration interval, numerical deviation direction and numerical deviation amplitude characteristics, generate corresponding high-frequency harmonic interference correction factors, and build a short-period data correction weight sequence of the electric energy metering node level;

[0083] To realize the quantitative correction of the abnormal electric energy error segment of the electric energy metering node under the action of high-frequency harmonic interference, a correction factor generation method based on abnormal segment characteristics is proposed, and a short-period data correction weight sequence of the electric energy metering node level is built, which is used for accurate repair of electric energy data. This implementation includes the following steps:

[0084] For the identified abnormal power error segments, the key feature parameters are extracted, including time duration interval, numerical offset direction and numerical offset amplitude. The time duration interval is used to measure the action time of the interference, which is an important basis for controlling the distribution of correction strength on the time axis; the numerical offset direction indicates whether the error is high or low, which is the key to determining the correction increase and decrease direction; the numerical offset amplitude reflects the severity of the error, which is the core parameter to determine the correction amplitude. These feature parameters are obtained by statistical calculation on the power deviation data, ensuring that the correction processing is targeted and accurate.

[0085] According to the above characteristics, a correction factor corresponding to each abnormal segment is constructed. Specifically, if the offset direction is positive, the correction factor is negative, and vice versa, so that the correction effect presents a symmetric compensation characteristic. The absolute value of the correction factor is set according to the size of the offset amplitude. The greater the offset amplitude, the stronger the correction factor. Considering the influence of the duration of the interference, the effect of the correction factor can be distributed according to the time gradient within the segment, so that the correction weight of the beginning and the end of the segment gradually changes, thereby avoiding the distortion of the power curve structure caused by abrupt correction.

[0086] Each correction factor is finely mapped to the corresponding abnormal segment interval in time sequence, and is divided into several short period time intervals, such as one interval per second or per half second. In each interval, different weights are assigned according to its relative position in the whole time, forming a set of continuous and time-sensitive correction weights. This process ensures that each correction weight corresponds to a specific power measurement point, achieving accurate adjustment at the sampling point level. In this way, a correction weight sequence with both segment continuity and time flexibility is constructed, ensuring the smoothness and structural stability of data correction.

[0087] The role of this step is to quantitatively correct the identified abnormal power error segments, and to construct a correction mechanism with time sequence perception ability and interference sensitivity, so as to accurately restore the true state of the electric energy measurement data under the influence of high-frequency harmonic interference. In the actual operation of the smart grid environment, high-frequency harmonic interference has the characteristics of suddenness, intermittence and locality, which may cause the electric energy measurement chip to produce systematic deviation in a short time. Only by marking abnormal segments and shielding or excluding related data, not only the key energy consumption information cannot be preserved, but also the continuity and analysis integrity of the overall power curve may be damaged.

[0088] Therefore, the step extracts the core features of each abnormal segment, such as the time duration interval, the offset direction, and the offset amplitude, constructs a targeted correction factor, and further decomposes it into a continuous correction weight sequence at the sampling granularity, to realize flexible adjustment of the original data. This correction method not only considers the numerical size of the error, but also integrates the interference persistence and trend information, which can dynamically compensate for the sampling points affected by interference without destroying the original data structure, significantly improving the smoothness and reliability of data repair. The final correction result will be used in subsequent key links such as power grid line loss calculation and early warning analysis, effectively avoiding global evaluation deviation caused by local abnormal error diffusion, and ensuring the continuity and accuracy of intelligent power grid data processing flow. Therefore, this step plays a key role in the transition from "abnormal identification" to "error correction" in the entire method system, and is an important technical link to support the dynamic error compensation mechanism.

[0089] S5, based on the short-period data correction weight sequence, performing segment-level weighted correction on the original electric energy measurement data, calculating the electric quantity consistency verification matrix between multiple electric energy measurement nodes, and correcting the systematic error caused by high-frequency harmonic interference in the electric energy measurement data of each node;

[0090] To solve the problem of systematic deviation of local electric energy measurement error caused by high-frequency harmonic interference in multi-node power grid data, a segment-level weighted correction method based on short-period data correction weight sequence is proposed, and further through the electric quantity consistency verification mechanism, the overall correction and balance of the measurement data between multiple electric energy measurement nodes are realized. Specifically, the following steps are included:

[0091] According to the short-period data correction weight sequence constructed above, it is applied to the original electric energy data of the corresponding electric energy measurement node, and the measurement values of the abnormal electric energy error segments are adjusted at the segment level. The specific method is as follows: for each sampling point affected by interference, the corrected power value is calculated according to the correction weight in the time sequence, and a new power time sequence is obtained. This correction operation is carried out in a point-to-point manner to ensure the time continuity and physical rationality of the data. In the external segment, i.e. the part of the data not affected by interference, the original measurement value is kept unchanged, thereby constructing a complete power curve integrating the original data and the corrected segment, and realizing the full curve reservation based on local interference compensation.

[0092] The corrected power data of each electric energy metering node is uniformly converted into equal-period electric quantity data (such as per-minute electric quantity, per-5-minute electric quantity, etc.), and the electric quantity data between different nodes is aligned on the time axis to facilitate cross-node comparison. In an actual power grid structure, multiple electric energy metering nodes are often located in the same power supply branch or power transformation unit, and the total incoming line and branch outgoing line should follow the electric quantity conservation relationship. Based on this relationship, the electric quantity logical association structure is extracted, and the electric quantity consistency relationship between nodes is constructed, that is, a plurality of pairs of "parent-child node" or "adjacent node" combinations are formed, which are used for subsequent consistency verification.

[0093] According to the above node combination relationship, an electric quantity consistency verification matrix is constructed, and each element of the matrix is used to describe the electric quantity difference between two related electric energy metering nodes in a specific time period. For example, in a typical bus-branch architecture, the electric quantity of the bus node is taken as a reference, and compared with the sum of the electric quantities of all the branch nodes hanging below, to calculate the electric quantity difference value per unit time. Under normal operating conditions, the difference value should remain within an acceptable error tolerance interval; if the difference value deviates for a long time, it indicates that there is a residual error or the metering data of one or more nodes has not been completely corrected. Through the verification matrix, the global metering consistency level can be quantitatively evaluated, and the possible error concentration nodes can be identified.

[0094] The construction process of the electric quantity consistency verification matrix is based on the physical connection relationship between the electric energy metering nodes in the power grid topology structure, and the electric quantity data between the nodes is systematically compared according to the principle of electric quantity conservation. First, according to the power grid operation model, it is determined which upper node each node logically belongs to or whether there is a parallel relationship, forming a plurality of groups of "electric quantity associated node pairs", such as incoming line nodes and their subordinate branch nodes. Then, in a uniform time period (such as every 5 minutes or every 15 minutes), the electric quantity data corresponding to each node pair is respectively summarized. For each associated node group, the difference between the reference node electric quantity (such as the bus or the upper node) and the sum of the target node electric quantities is calculated, and the difference value is filled into the corresponding position of the verification matrix. Each row of the matrix represents a time period, and each column represents a group of electric quantity comparison relationship, and the values in the matrix reflect the size and direction of the electric quantity difference. Through analysis of the matrix, it can be quickly identified whether there is unreasonable energy distribution between the electric energy metering nodes, and whether the correction value of a node exists deviation that has not been eliminated, which is an important basis for realizing global data consistency evaluation and correction adjustment.

[0095] According to the deviation degree in the consistency check matrix of electric quantity, the short period correction weight of the node with prominent deviation is further adjusted. The adjustment is based on the principle of minimum deviation, and the correction amplitude is fine-tuned without destroying the shape of the single node data curve, so as to minimize the overall error. After the adjustment is completed, the new corrected electric quantity sequence is regenerated, and the check matrix operation is performed again until the overall consistency error of the system converges to the tolerance threshold range. Through the above process, the transition from single node segment correction to global consistency correction of multiple nodes is realized, which effectively improves the physical accuracy and structural rationality of the data, and suppresses the secondary measurement error diffusion caused by unbalanced interference correction.

[0096] The core role of this step is to further improve the overall data consistency and logical closed loop of the entire power grid measurement system on the basis of completing the local error correction of a single electric energy measurement node, so as to realize the overall elimination of systematic errors caused by high-frequency harmonic interference. In a complex distribution network, there is a clear physical power supply relationship between multiple electric energy measurement nodes, such as main lines and branches, transformer bus and outgoing line, distribution transformer and end user. The electric quantity of these nodes should meet the principle of energy conservation in statistics. However, due to the regional and asymmetric nature of high-frequency harmonic interference, its influence on different nodes is not consistent. Even if the local segment correction at the node level is completed, if the relationship between nodes is not coordinated, energy distribution imbalance at the system level may still occur, which will affect the accuracy of line loss analysis and load assessment. Therefore, by combining the original electric energy measurement data of each node with its corresponding short period correction weight sequence, the segment level weighted adjustment can be realized, and fine-grained error compensation can be achieved within the interference segment. Further, by constructing an electric quantity consistency check matrix, the electric quantities of multiple nodes are compared in time sequence, and the electric quantity difference between them is quantified, so as to find the distribution characteristics of residual error between nodes. This process not only helps to identify nodes with insufficient correction or over-correction, but also realizes the secondary balance repair of errors in the whole network. Finally, this step realizes the transition from "local correction" to "system consistency optimization", which not only preserves the repair accuracy of each measurement point, but also ensures the consistency and physical rationality of the whole network data. It is a key step to ensure the accuracy of smart grid line loss detection, power tracing and energy efficiency evaluation.

[0097] S6, input the corrected electric energy measurement data into the power grid line loss analysis module, update the warning judgment threshold and error tolerance boundary in the line loss trend model, realize the closed-loop correction of electric energy measurement interference data, and realize the real-time stable identification and response control of smart grid line loss anomaly;

[0098] To realize the closed-loop correction of power measurement data under the influence of high-frequency harmonic interference and ensure the accuracy and response ability of line loss analysis in the interference environment, a processing flow based on dynamic updating of line loss trend model after correction is proposed. The flow re-constructs the line loss judgment basis by taking the corrected power measurement data as input, updates the early warning threshold and tolerance boundary, realizes the stable identification and real-time regulation of line loss anomaly in smart grid, and includes the following steps:

[0099] The power measurement data after short-period weight correction and node consistency leveling processing is re-collected according to the power grid topology, and the corresponding line loss value of each power supply unit or distribution branch is recalculated, including the difference between the incoming power and each outgoing power in a unit time period, and then the corrected basic line loss data sequence is obtained. To ensure data continuity and model input consistency, the sequence needs to be time-aligned and dimension-unified with historical line loss data to ensure that there is no abnormal mutation or scale mismatch in the model input layer.

[0100] Based on the corrected line loss data, the existing line loss trend model is updated. The line loss trend model is a mathematical or statistical model used to describe the line loss change law of a power grid region at different time periods, which often includes time series model, regression prediction model or multi-factor regression model, etc. In the updating process, the time window corresponding to the correction data segment is preferentially introduced, and the real data in the window is used to replace the original noisy data, and the model parameters such as sliding mean, volatility, change gradient and other key indicators are recalculated to ensure that the model reflects the actual operation state of the current network more truly, stably and accurately.

[0101] Based on the updated line loss trend model, the reference threshold and error tolerance boundary of line loss anomaly early warning judgment are dynamically adjusted. Traditional early warning mechanism often uses static threshold or empirical setting, which is prone to false alarm or missed alarm in complex electromagnetic interference environment. By introducing the latest correction data and combining the dynamic output results of the trend model, the adaptive adjustment of the early warning threshold can be realized. For example, according to the confidence interval, residual distribution and other indicators output by the model, a multi-level threshold response system under different operating conditions is set; the tolerance boundary is relaxed when the load fluctuation is large, and the error allowed interval is compressed under stable conditions, so as to improve the identification ability and interference tolerance ability of the system to real anomalies.

[0102] The dynamically updated judgment system is integrated with the original line loss analysis process to form a closed-loop operation mechanism with self-perception, self-correction and self-adjustment. When new data flows in, the system first completes error identification and correction operations, then embeds the correction results into the line loss analysis process, triggers the retraining of the trend model and refreshes the threshold value, and finally uses the real-time judgment results to drive the alarm mechanism, abnormal marking or automatic operation and maintenance decision. Through this mechanism, even if the electric energy metering data is affected by periodic or intermittent high-frequency interference, the continuity and reliability of the line loss evaluation results can be ensured, and the operation stability and abnormal response capability of the smart grid in complex environment can be significantly enhanced.

[0103] The role of this step is to truly integrate the electric energy metering error identification and correction results completed in the early stage for high-frequency harmonic interference into the power grid line loss analysis system, and to build a closed-loop response mechanism based on data quality improvement, so as to realize the overall improvement of stability, accuracy and self-adaptive capability in the line loss anomaly identification of the smart grid. The line loss analysis of the smart grid usually relies on the power and energy data of each electric energy metering node for calculation and trend modeling, but in the operation environment with frequent high-frequency electromagnetic interference, the original metering data may have periodic deviation, local mutation or systematic distortion, and if used directly for line loss modeling and early warning threshold setting, it will lead to false alarm, missed alarm, and even affect energy scheduling and operation and maintenance strategy judgment. Through this step, the system first inputs the high-quality electric energy metering data after short-period weight adjustment and consistency check into the line loss analysis process, eliminating the error fragments caused by interference from the data source layer and enhancing the stability of trend analysis. Then, by updating the historical data window and prediction structure of the line loss trend model, the model is closer to the current operation condition, avoiding the accumulation of deviation caused by relying on distorted historical data. On this basis, the early warning judgment threshold and error tolerance boundary are dynamically adjusted to have a sensitive but not excessive response capability under different load states, equipment operation periods or power grid topology changes, realizing the transformation of the early warning mechanism from static judgment to dynamic perception. Finally, this step forms a whole-process closed-loop control link of "interference identification-error correction-model updating-threshold adjustment-stable identification", so that the power grid still has stable, accurate and real-time line loss monitoring capability in complex electromagnetic environment, significantly improving the perception ability and emergency control level of the smart grid to operation anomalies, which is a key link to support the high reliability and high intelligence of the smart grid operation.

[0104] Through the above intelligent power grid line loss intelligent detection method based on Internet of Things big data, a closed-loop processing mechanism from period identification of high-frequency harmonic interference, error tracing, cause discrimination to fine correction and consistency verification of whole network data can be realized, effectively solving the problems of power misjudgment, line loss misidentification and false alarm caused by interference in the traditional electric energy metering system under complex electromagnetic environment. The method not only has multi-dimensional feature fusion interference perception ability, but also introduces dynamic correction factor and weight distribution strategy to realize error compensation at the segment level and node level, and through the electric quantity consistency matrix and trend model updating mechanism, the accuracy and stability of the corrected data in the whole network analysis are comprehensively guaranteed. Finally, the intelligent power grid can still maintain high sensitivity, high accuracy and sustainable response ability to line loss abnormalities under complex operation scenes such as high-frequency interference and nonlinear load concentration, and the reliability, precision and intelligent level of the line loss analysis system are improved.

[0105] The above formulas are dimensionless values calculated, and the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0106] The above only describes certain exemplary embodiments of the present application by way of illustration, and it is self-evident that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present application.

[0107] It should be noted that in this paper, if there are relationship terms such as first and second, they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0108] It should be understood that in various embodiments of the present application, the size of the sequence number of each process described above does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0109] Those skilled in the art can clearly understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0110] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0111] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment according to actual needs.

[0112] In addition, the functional units in each embodiment of the present application can be integrated into one processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit.

[0113] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0114] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without deviating from the spirit and scope of the present application. Therefore, the above figures and description are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

Claims

1. The intelligent power grid line loss intelligent detection method based on Internet of Things big data, characterized in that, Comprising the following steps: S1, extracting the voltage signal and current signal of each electric energy metering node within the set time window, performing frequency domain transformation to obtain the frequency spectrum, calculating the energy aggregation index of the high frequency band, and identifying the high frequency harmonic interference time section according to the energy aggregation degree change; S2, based on the identified high frequency harmonic interference time section, calling the historical standard active power curve template of the corresponding electric energy metering node, fitting the current active power curve with the template, and extracting the active power deviation vector; S3, inputting the active power deviation vector, terminal device type, instantaneous load change characteristic and high frequency harmonic spectrum characteristic into the abnormal cause classification model, outputting whether the power deviation is caused by high frequency harmonic interference on the electric energy metering chip, and marking the abnormal electric energy error segment; S4, for the abnormal electric energy error segment, generating a high frequency harmonic interference correction factor according to the time duration interval, numerical offset direction and amplitude, and constructing an electric energy metering node level short period data correction weight sequence; Step S4 includes: extracting the time duration interval, numerical offset direction and numerical offset amplitude of the abnormal electric energy error segment; generate the corresponding correction factor based on the offset direction and offset amplitude, and set the correction gradient according to the time distribution within the segment; map the correction factor to the abnormal segment and divide it into short period time sections; assign correction weights in each time section to construct a short period data correction weight sequence at the electric energy metering node level; S5, based on the correction weight sequence, the original electric energy metering data is corrected by segment weighting, and an electric quantity consistency check matrix among multiple electric energy metering nodes is calculated to correct the systematic error as a whole; Step S5 includes: based on the short period data correction weight sequence, the abnormal electric energy error segment in the original electric energy metering data is corrected by segment level weighting to generate a corrected power sequence; convert the corrected power data of each electric energy metering node into equal period electric quantity data, and construct an electric quantity consistency check matrix among multiple metering nodes; based on the deviation degree of the electric quantity difference between each node in the electric quantity consistency check matrix, adjust the correction weight of the corresponding node to generate new corrected electric quantity data; repeat the checking and correcting process until the overall consistency error of the system converges to within the tolerance threshold range; S6, input the corrected data into the line loss analysis module, update the early warning judgment threshold and error tolerance boundary in the line loss trend model, realize the closed loop correction of electric energy metering interference data and real-time stable identification and response of line loss anomaly. 2.The IoT big data-based smart grid line loss intelligent detection method according to claim 1, characterized in that, Step S1 includes: within the set time window, continuously collect the voltage signal and current signal of the electric energy metering node, and perform window function weighting processing to eliminate the influence of frequency spectrum leakage; perform fast Fourier transform on the weighted processed signal to obtain voltage frequency spectrum and current frequency spectrum, and calculate the frequency energy accumulation value of the high frequency band; calculate the frequency energy aggregation index to form time series data; based on dynamic threshold and disturbance duration judgment, identify the time section of high frequency harmonic interference, and generate interference event log for subsequent error compensation. 3.The IoT big data-based smart grid line loss intelligent detection method of claim 1, wherein, Step S2 includes: Based on the identified high-frequency harmonic interference time section, the active power measurement values recorded continuously by the corresponding electric energy metering node between the start time and the end time of the high-frequency harmonic interference are extracted, a current active power curve is constructed, and the active power curve is normalized; The standard active power curve template under the historical operating conditions of the corresponding node is called; The current active power curve and the standard template are time-series fitted, shape alignment is performed using a dynamic time warping algorithm or a window alignment strategy, and point-by-point power difference values are calculated; The statistical features of the power difference value sequence are extracted, an active power deviation vector is constructed, and used to quantify the influence of high-frequency harmonic interference on the metering node. 4.The IoT big data-based smart grid line loss intelligent detection method of claim 1, wherein, Step S3 includes: Extracting the active power deviation vector, the terminal device type, the instantaneous load change feature, and the high-frequency harmonic spectrum feature; The extracted features are uniformly input into the trained electric energy anomaly cause classification model, which is constructed based on a supervised learning algorithm; Outputting the classification result of whether the power deviation is caused by high-frequency harmonic interference on the electric energy metering chip or not; Marking the time section corresponding to the classification result as an abnormal electric energy error segment, and using it as the basis for generating subsequent correction parameters. 5.The IoT big data-based smart grid line loss intelligent detection method of claim 1, wherein, Step S6 includes: The corrected electric energy metering data is collected according to the power grid topology, and the basic line loss data sequence of each power supply unit is recalculated; The corrected basic line loss data sequence is input into the line loss trend model, the original data in the corresponding time period is replaced, and the model parameters are updated; Based on the updated line loss trend model, the line loss anomaly early warning judgment threshold and the error tolerance boundary are adjusted; The updated judgment result is integrated with the line loss analysis process to form a closed-loop correction and real-time response regulation mechanism for electric energy metering data.

Images