A power grid measurement data quality intelligent monitoring method

Abstract

The present application relates to the technical field of data monitoring, and particularly relates to a power grid measurement data quality intelligent monitoring method.The method comprises the following steps: filling and cleaning and normalizing various types of measurement data in a power grid by real-time acquisition to obtain power grid measurement multi-source normalized data; performing multi-dimensional quality characteristic analysis and measurement quality evaluation on the power grid measurement multi-source normalized data to obtain power grid measurement data quality grade scores; obtaining power grid operation corresponding seasonal, periodic and real-time working condition data to perform power grid quality dynamic threshold statistics to obtain power grid operation quality dynamic thresholds; performing measurement anomaly detection based on the power grid measurement data quality grade scores and combining the power grid operation quality dynamic thresholds, while performing anomaly classification and intelligent early warning monitoring to generate corresponding power grid measurement data anomaly early warning information and perform corresponding power grid safe operation decision support work.The present application can comprehensively, accurately and intelligently monitor power grid measurement data quality.

Description

Technical Field

[0001] This invention relates to the field of data monitoring technology, and in particular to an intelligent monitoring method for the quality of power grid measurement data. Background Technology

[0002] With the development of intelligent and automated power systems, power grid operation monitoring and data acquisition have become crucial means to ensure the safe, stable, and efficient operation of the power grid. Power grid measurement data includes important parameters such as voltage, current, and power, which provide the foundation for real-time power system dispatching, load forecasting, fault detection, and safety assessment. Currently, methods for monitoring and correcting power grid data quality mainly rely on traditional data verification methods, such as data consistency checks, range checks, and outlier detection. However, these methods are usually based on simple rules and manually set thresholds, which cannot adapt to the complex nonlinear relationships and variable operating environments in power grid data. Therefore, existing monitoring methods struggle to effectively identify and handle various types of data quality problems when faced with large-scale, high-frequency power grid measurement data, and often suffer from high false alarm and false negative rates. Summary of the Invention

[0003] Therefore, the present invention needs to provide an intelligent monitoring method for power grid measurement data quality to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for intelligent monitoring of power grid measurement data quality includes the following steps:

[0005] Step S1: By collecting various measurement data in the power grid in real time, and filling, cleaning and normalizing the various measurement data in the power grid, multi-source normalized data of power grid measurement is obtained.

[0006] Step S2: Perform multi-dimensional quality feature analysis on the multi-source normalized data of power grid measurements to obtain a multi-dimensional quality feature set of power grid measurements; evaluate the measurement quality of various types of measurement data in the power grid based on the multi-dimensional quality feature set of power grid measurements to obtain a power grid measurement data quality level score;

[0007] Step S3: Obtain seasonal, periodic and real-time operating condition data corresponding to power grid operation, and perform dynamic threshold statistics on power grid quality based on the seasonal, periodic and real-time operating condition data corresponding to power grid operation to obtain dynamic thresholds for power grid operation quality.

[0008] Step S4: Based on the power grid measurement data quality level score and combined with the power grid operation quality dynamic threshold, perform measurement anomaly detection on various types of measurement data in the power grid to obtain power grid measurement quality anomaly data results; perform anomaly classification and intelligent early warning monitoring on the power grid measurement quality anomaly data results to generate corresponding power grid measurement data anomaly early warning information, and execute corresponding power grid safety operation decision support work.

[0009] Furthermore, step S1 includes the following steps:

[0010] Step S11: Collect various measurement data in the power grid in real time, including real-time data on voltage, current, active power, reactive power, and frequency of substations, transmission lines, and distribution substations, as well as auxiliary data corresponding to the status of data acquisition equipment and communication links.

[0011] Step S12: Perform deduplication processing on various types of measurement data in the power grid to obtain multi-source deduplicated power grid measurement data;

[0012] Step S13: Use regression to fit and fill in the missing data corresponding to various measurements in the multi-source deduplication data of power grid measurements, so as to obtain the missing data of multi-source power grid measurements.

[0013] Step S14: Perform format normalization processing on various types of measurement data in the multi-source missing data of power grid measurement to convert measurement data with different formats and units into standard formats and units to obtain normalized multi-source data of power grid measurement.

[0014] Furthermore, step S2 includes the following steps:

[0015] Step S21: Perform time-series synchronization processing on the normalized data of multiple power grid measurement sources to obtain the corresponding time-series data of multiple power grid measurement sources within the same time series range;

[0016] Step S22: By setting a time-series sliding window, and performing sliding smoothing filtering on the corresponding multi-source time-series data of power grid measurements within the same time-series range based on the time-series sliding window, smoothed multi-source time-series data of power grid measurements are obtained.

[0017] Step S23: Perform time-domain feature statistics on the multi-source time-series smoothed data of power grid measurements to statistically calculate the mean, variance, maximum value, minimum value, skewness, kurtosis and rate of change of various types of measurement data, and obtain the time-domain statistical characteristics of the multi-source power grid measurements.

[0018] Step S24: Perform Fourier transform on various types of measurement data within the multi-source normalized data of power grid measurements to generate frequency bands for various types of measurement data; perform frequency domain feature statistics on various types of measurement data frequency bands to obtain the frequency domain statistical characteristics of multi-source power grid measurements;

[0019] Step S25: Perform correlation feature analysis on various types of measurement data within the multi-source normalized data of power grid measurements to analyze the degree of correlation between different measurement data and obtain the multi-source correlation features of power grid measurements.

[0020] Step S26: Analyze the equipment operation status data of the corresponding data acquisition equipment in the multi-source normalized data of power grid measurement to obtain the operation status characteristics of the power grid acquisition equipment; evaluate the link stability of the corresponding communication link status data in the multi-source normalized data of power grid measurement to obtain the stability characteristics of the power grid communication link.

[0021] Step S27: Combine the time-domain statistical characteristics of multiple power grid measurement sources, the frequency-domain statistical characteristics of multiple power grid measurement sources, the correlation characteristics of multiple power grid measurement sources, the operating status characteristics of power grid acquisition equipment, and the stability characteristics of power grid communication links into the same set to obtain a multi-dimensional quality feature set of power grid measurement.

[0022] Step S28: Based on the multi-dimensional quality feature set of power grid measurements, evaluate the measurement quality of various types of measurement data in the power grid to obtain the power grid measurement data quality level score.

[0023] Furthermore, the link stability assessment of the corresponding communication link status data within the multi-source normalized data of power grid measurements described in step S26 includes the following steps:

[0024] The bandwidth timing data, delay timing data, and packet loss timing data of each link in the power grid are obtained by using communication link status data.

[0025] Basic stability index analysis is performed on the bandwidth time series data, delay time series data and packet loss time series data corresponding to each link in the power grid, so as to statistically analyze the instantaneous bandwidth, delay and packet loss volatility corresponding to each link and obtain the basic stability index of the power grid corresponding to each link.

[0026] Periodic fluctuation detection is performed on communication link status data. The long-term abnormal fluctuation range of each link is evaluated by modeling with an autoregressive model or a moving average model. The periodic fluctuation data segments of each link are analyzed, and the fluctuation amplitude, frequency and trend of each link are quantified to obtain the power grid volatility index of each link.

[0027] Correlation analysis is performed on the bandwidth time-series data, delay time-series data and packet loss time-series data corresponding to each link in the power grid to analyze the correlation between the bandwidth, delay and packet loss indicators corresponding to each link and obtain the power grid status correlation indicators corresponding to each link.

[0028] By constructing the topology of the power grid communication network and evaluating the stability contribution of each link to the entire power grid communication network based on the topology, the stability contribution weight of each link to the entire power grid communication network can be obtained.

[0029] Based on the stability contribution weight of each link to the entire power grid communication network, and using the analytic hierarchy process (AHP) to comprehensively evaluate the stability of each link, the corresponding basic stability index, power grid volatility index, and power grid state correlation index are divided into three levels: basic stability, volatility, and correlation. The stability level of each link in the power grid is calculated by combining the stability contribution weight of each link, so as to obtain the stability characteristics of the power grid communication links.

[0030] Furthermore, the analysis process for the instantaneous bandwidth fluctuation rate corresponding to each link is as follows:

[0031] By setting the corresponding sliding window size to 3-5 time points, and dividing the bandwidth time series data corresponding to each link in the power grid into bandwidth time series segments based on the sliding window size, the bandwidth time series window data segments corresponding to each link are obtained.

[0032] By selecting any two adjacent bandwidth time-series window data segments for each link, the corresponding bandwidth fluctuation amplitude is calculated to obtain the bandwidth fluctuation amplitude under any two adjacent time-series windows for each link.

[0033] Instantaneous volatility is calculated by statistically analyzing the bandwidth fluctuation amplitude between any two adjacent time windows of each link, so as to obtain the instantaneous volatility at the time point between any two adjacent time windows of each link, and thus obtain the instantaneous bandwidth volatility of each link.

[0034] Furthermore, step S28 includes the following steps:

[0035] Step S281: By treating each quality feature in the multi-dimensional quality feature set of power grid measurement as a node in the graph, and using graph convolutional network to analyze the correlation between each feature node, and at the same time performing spatial feature correlation graph analysis on each quality feature in the multi-dimensional quality feature set of power grid measurement based on the correlation between each feature node, a spatial correlation graph of power grid measurement quality features is generated.

[0036] Step S282: Use a weighted fusion neural network to perform in-depth quality assessment modeling of spatial correlation feature nodes in the spatial correlation map of power grid measurement quality features, so as to generate a multi-dimensional quality assessment model for power grid measurement.

[0037] Step S283: Input various types of measurement data in the power grid into the power grid measurement multidimensional quality assessment model for measurement quality assessment. Calculate the correlation coefficient between the corresponding measurement data and each spatially associated feature node based on the weighted fusion neural network, and calculate the corresponding quality level score to obtain the power grid measurement data quality level score.

[0038] Furthermore, step S3 includes the following steps:

[0039] Step S31: Obtain seasonal, periodic, and real-time operating data corresponding to the power grid operation;

[0040] Step S32: Analyze the power grid load level based on the seasonal and periodic operating conditions data corresponding to the power grid operation to obtain the power grid operating condition load level;

[0041] Step S33: Calculate the power generation ratio of the real-time operating data corresponding to the power grid operation to obtain the power generation ratio under the power grid operation conditions;

[0042] Step S34: Perform dynamic threshold statistics on grid quality based on the load level of grid operation and the power generation ratio of grid operation to obtain the dynamic threshold of grid operation quality.

[0043] Furthermore, step S32 includes the following steps:

[0044] The power load spectrum is converted from the seasonal and periodic operating conditions data corresponding to the power grid operation to generate the power grid load sine wave signal spectrum corresponding to the power grid operation load under seasonal and periodic conditions;

[0045] By performing statistical analysis on the load variation spectrum of the power grid load sinusoidal signal under seasonal and periodic conditions, the seasonal and periodic load variation trends of the power grid operating conditions can be obtained.

[0046] Load level prediction analysis is conducted based on the seasonal and periodic load change trends of the power grid operating conditions to obtain the load level of the power grid operating conditions.

[0047] Furthermore, step S33 includes the following steps:

[0048] The power generation load distribution of different power generation units in the real-time operating data corresponding to the power grid operation is analyzed to obtain the power generation load distribution of different power generation units under the power grid operation conditions.

[0049] Power transmission mining and analysis are performed on the real-time operating data of the power grid to obtain the power transmission relationship between different power generation units under the power grid operating conditions.

[0050] Based on the power transmission relationship between different power generation units under different grid operating conditions, the power dispatch attenuation assessment is carried out between different power generation units in the real-time operating condition data corresponding to the grid operation, so as to obtain the power dispatch attenuation coefficient between different power generation units under different grid operating conditions.

[0051] Based on the power generation dispatch attenuation coefficient between different power generation units under different power grid operating conditions, the power generation ratio of the power generation load distribution corresponding to different power generation units under different power grid operating conditions is calculated to obtain the power generation ratio under the power grid operating conditions.

[0052] Furthermore, step S4 includes the following steps:

[0053] Step S41: Based on the power grid measurement data quality level score and combined with the power grid operation quality dynamic threshold, perform measurement anomaly detection on various types of measurement data in the power grid. When the power grid measurement data quality level score is between 0 and 60 points or the corresponding measurement data in the power grid exceeds the power grid operation quality dynamic threshold, the corresponding measurement data is judged as data anomaly until all types of measurement data are detected to obtain the power grid measurement quality anomaly data result.

[0054] Step S42: Perform anomaly classification processing on each abnormal measurement data in the power grid measurement quality anomaly data results to obtain various types of power grid abnormal measurement data;

[0055] Step S43: Perform power grid anomaly source analysis on various types of power grid anomaly measurement data to obtain the causes of anomalies corresponding to various types of anomalies, including anomalies at individual data points as well as large-scale data anomalies caused by equipment failures, communication interruptions and power grid failures.

[0056] Step S44: Intelligent early warning monitoring is performed based on the causes of anomalies in various types of abnormal measurement data. If an anomaly is found in an individual data point, a yellow warning is issued; if a large-scale data anomaly is caused by equipment failure, communication interruption, or power grid failure, a red warning is issued. Corresponding power grid measurement data anomaly warning information is generated and pushed to the relevant power grid operation and maintenance management cloud platform via SMS, email, and in-station messages to execute corresponding power grid safety operation decision support work, including power grid operation mode adjustment plans, equipment maintenance plans, and communication link optimization plans.

[0057] The beneficial effects of this invention are:

[0058] The intelligent monitoring method for power grid measurement data quality proposed in this invention, compared with existing technologies, offers the following advantages: by collecting various measurement data from the power grid in real time, such as voltage, current, frequency, and power, it can comprehensively and in real-time monitor the operating status of the power grid, ensuring that all important data can be obtained in a timely manner. However, power grid data may be missing or abnormal due to factors such as noise, equipment failure, or environmental interference. Therefore, data filling and cleaning is crucial. This process can effectively eliminate erroneous information in the data, making it more reliable and accurate. Data normalization ensures that various measurement data can be compared under the same standard, eliminating dimensional differences between different measurement data. This allows power grid data from different sources to be processed uniformly on the same platform, improving the efficiency and accuracy of data processing, thereby enhancing the reliability and stability of the power grid. Secondly, by conducting multi-dimensional quality characteristic analysis on multi-source normalized data of power grid measurements, the quality characteristics of various types of measurement data can be comprehensively evaluated, including indicators such as the basic stability, volatility, and correlation of the measurement data. These quality characteristics will help identify potential quality problems and provide a basis for data quality assessment. Through measurement quality assessment, each measurement data of the power grid can be assigned a quality level score, thereby clearly understanding the quality status of each type of data. For example, some data may have large errors due to equipment failure or transmission delays, while other data may have high accuracy and timeliness. Through measurement quality level scoring, decision-makers can quickly identify high-quality and low-quality data and then take targeted measures, such as monitoring low-quality data more frequently or correcting it in more detail. This can effectively identify and handle various types of data quality problems, thereby improving the overall operating efficiency and reliability of the power grid. Then, by considering the seasonality, periodicity, and real-time operating data of the power grid, we can gain a deeper understanding of the power grid's performance in different time periods and under different operating conditions. Based on these characteristic data, we can statistically determine the dynamic quality thresholds, which can help to more accurately set the standards and safety boundaries for power grid operation. These dynamic thresholds can not only reflect the standard range of the power grid during stable operation, but also flexibly adapt to changes in the power grid under different operating conditions. Through this process, we can monitor the power grid's operating status more precisely, thereby improving the power grid's ability to respond to various emergencies.Finally, by utilizing the power grid measurement data quality rating and the power grid operation quality dynamic threshold, anomaly detection of measurement data can be performed, effectively identifying data that does not meet expected standards. This process not only promptly detects data anomalies but also categorizes them according to different types of anomalies. For example, some anomalies are caused by sensor failure or data loss, while others are caused by problems within the power grid itself (such as equipment failure or load anomalies). Based on the intelligent early warning monitoring process, these anomalies will be automatically identified and converted into anomaly warning information, including yellow and red warnings, for relevant personnel to take measures. This can significantly improve the efficiency of power grid monitoring, avoid the lag and oversight of manual operation, and ensure that the power grid can respond immediately when problems occur. The combination of anomaly detection and early warning can also provide important support for the safe operation decision-making of the power grid, such as timely scheduling of power equipment for maintenance or adjustment of power grid load, thereby reducing the false alarm rate and missed alarm rate of power grid measurement data quality anomalies. Attached Figure Description

[0059] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0060] Figure 1 This is a flowchart illustrating the steps of the intelligent monitoring method for power grid measurement data quality of the present invention.

[0061] Figure 2 for Figure 1 A detailed flowchart of step S1. Detailed Implementation

[0062] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0063] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides an intelligent monitoring method for power grid measurement data quality, the method comprising the following steps:

[0064] Step S1: By collecting various measurement data in the power grid in real time, and filling, cleaning and normalizing the various measurement data in the power grid, multi-source normalized data of power grid measurement is obtained.

[0065] Step S2: Perform multi-dimensional quality feature analysis on the multi-source normalized data of power grid measurements to obtain a multi-dimensional quality feature set of power grid measurements; evaluate the measurement quality of various types of measurement data in the power grid based on the multi-dimensional quality feature set of power grid measurements to obtain a power grid measurement data quality level score;

[0066] Step S3: Obtain seasonal, periodic and real-time operating condition data corresponding to power grid operation, and perform dynamic threshold statistics on power grid quality based on the seasonal, periodic and real-time operating condition data corresponding to power grid operation to obtain dynamic thresholds for power grid operation quality.

[0067] Step S4: Based on the power grid measurement data quality level score and combined with the power grid operation quality dynamic threshold, perform measurement anomaly detection on various types of measurement data in the power grid to obtain power grid measurement quality anomaly data results; perform anomaly classification and intelligent early warning monitoring on the power grid measurement quality anomaly data results to generate corresponding power grid measurement data anomaly early warning information, and execute corresponding power grid safety operation decision support work.

[0068] In the embodiments of this invention, please refer to Figure 1 The diagram shown is a flowchart illustrating the steps of the intelligent monitoring method for power grid measurement data quality according to the present invention. In this example, the intelligent monitoring method for power grid measurement data quality includes the following steps:

[0069] Step S1: By collecting various measurement data in the power grid in real time, and filling, cleaning and normalizing the various measurement data in the power grid, multi-source normalized data of power grid measurement is obtained.

[0070] In this embodiment of the invention, smart meters, sensors, and other data acquisition devices deployed at various nodes such as substations, transmission lines, and distribution substations collect various measurement data of the power grid in real time at a frequency of once per second. This data covers electrical quantities such as voltage, current, active power, reactive power, and frequency, as well as auxiliary data such as the status of the data acquisition devices and the status of communication links. The raw data is transmitted to the data center in the form of digital signals and stored in the raw data table of the database. For missing values ​​in the raw data, linear interpolation is used to fill them in. For example, the voltage data of a certain distribution substation from 10:00 to 10:00... There are missing values ​​between 1 and 2. Based on the voltage values ​​at 9:59 and 10:01, the voltage filler value at 10:00 is obtained through linear calculation. For duplicate data, the timestamp and unique identifier of the data are used for deduplication to ensure the uniqueness of the data. In the normalization process, for voltage data, voltage values ​​of different ranges are uniformly mapped to the 0-1 interval. For example, the 220V voltage value is converted into the corresponding normalized value according to a specific formula. After filling, cleaning and normalization, multi-source normalized data of power grid measurement is formed and stored in a special data table to provide an accurate and standardized data foundation for subsequent analysis.

[0071] Step S2: Perform multi-dimensional quality feature analysis on the multi-source normalized data of power grid measurements to obtain a multi-dimensional quality feature set of power grid measurements; evaluate the measurement quality of various types of measurement data in the power grid based on the multi-dimensional quality feature set of power grid measurements to obtain a power grid measurement data quality level score;

[0072] In this embodiment of the invention, multi-dimensional quality characteristic analysis is performed on the multi-source normalized data of the power grid. In the time domain analysis, taking the current data of a certain transmission line as an example, its mean, variance, maximum value, minimum value, and other statistical quantities are calculated. The calculation shows that the mean current of the line is 500A and the variance is 25, reflecting that the current fluctuates within a certain range. In the frequency domain analysis, Fourier transform is used to convert the voltage data from the time domain to the frequency domain to analyze the energy distribution of different frequency components. If the energy proportion in the 50Hz frequency band reaches 80%, it indicates that the main frequency component of the power grid is stable at 50Hz, but there is harmonic interference at other frequencies. In the correlation analysis, the Pearson correlation coefficient between voltage and current is calculated. If the correlation coefficient is 0.9, it indicates that the two have a strong positive correlation. Simultaneously, the status data of the data acquisition equipment is analyzed, such as the equipment's operating time and temperature, to determine whether the equipment is in an abnormal operating state. The status data of the communication link is evaluated, and indicators such as bandwidth, latency, and packet loss rate are calculated to determine the stability of the link. These time-domain, frequency-domain, correlation, and equipment and link status features are integrated to form a multi-dimensional quality feature set of power grid measurements. Based on this feature set, a preset evaluation model and weights are used to evaluate the measurement quality of various types of measurement data in the power grid, and finally obtain the power grid measurement data quality level score. The score range is 0-100 points, such as 0-60 points as poor, 61-75 points as medium, 76-90 points as good, and 91-100 points as excellent. The higher the score, the better the data quality.

[0073] Step S3: Obtain seasonal, periodic and real-time operating condition data corresponding to power grid operation, and perform dynamic threshold statistics on power grid quality based on the seasonal, periodic and real-time operating condition data corresponding to power grid operation to obtain dynamic thresholds for power grid operation quality.

[0074] In an embodiment of the present invention, seasonal, periodic, and real-time operating condition data corresponding to the operation of the power grid are obtained from a historical database and a real-time monitoring module. The seasonal data covers power consumption loads, power generation powers, etc. in each season in the past 5 years. The periodic data includes the power grid operation data for each week in the past 104 weeks. The real-time operating condition data updates the operation parameters of each node of the power grid in real time. Taking the seasonal data as an example, the load change trend during the peak summer power consumption period is analyzed. By calculating indicators such as the average load and load growth rate at the same time period in summer of different years, the load level in the future summer is predicted. For the periodic data, the load change law for each day in a week is studied to determine the load difference between weekdays and weekends. Combining the real-time operating condition data, such as the power generation power of each power generation unit and the transmission power of the transmission line at a certain moment, time series analysis methods and mathematical models are used to perform dynamic threshold statistics on the power grid quality. For example, for the voltage threshold, according to the voltage fluctuation range under different operating conditions in the historical data and combining the current operating state of the power grid, the calculated dynamic threshold of the voltage is ±8% of the rated voltage; for the frequency threshold, considering the impact of the increasing proportion of renewable energy power generation on the frequency stability, the dynamic threshold of the frequency is determined to be 50 ± 0.2 Hz, so as to obtain the dynamic threshold of the power grid operation quality, which is used to measure whether the power grid operation parameters are normal.

[0075] Step S4: Perform measurement anomaly detection on various types of measurement data in the power grid based on the power grid measurement data quality grade score and in combination with the dynamic threshold of the power grid operation quality to obtain the power grid measurement quality anomaly data result; perform anomaly classification and intelligent early warning monitoring on the power grid measurement quality anomaly data result to generate the corresponding power grid measurement data anomaly early warning information, and execute the corresponding power grid safe operation decision support work.

[0076] In the embodiments of the present invention, measurement anomaly detection is performed on various types of measurement data in the power grid by scoring the quality level of power grid measurement data and combining it with the dynamic threshold of power grid operation quality. For example, the quality level score of the voltage measurement data of a certain substation is 55 points, which is lower than 60 points. At the same time, the actual voltage value is 200V, exceeding the range of the dynamic threshold of 220V ± 8%. This voltage measurement data is determined as abnormal data. According to the data acquisition sequence, all measurement data is detected one by one, and the abnormal data is recorded in the abnormal data table to form the power grid measurement quality abnormal data result. The abnormal data result is subjected to abnormal classification processing. If the current data of a certain transmission line has an isolated abnormal value at a certain moment while other related data is normal, it is classified as an abnormal individual data point; if the voltage, current, power, etc. data of multiple distribution transformers in a certain area are abnormal at the same time and show a similar change trend, it is determined as a large-area data anomaly caused by equipment failure, communication interruption or power grid failure. For different types of anomalies, intelligent early warning monitoring is carried out. For abnormal individual data points, a yellow early warning message is generated, detailing the location, parameters and preliminary reasons of the abnormal data, such as "The current of a certain line is abnormal at 14:23, and the reason is short-term load fluctuation". For large-area data anomalies, a red early warning message is generated, such as "Equipment failure has occurred in the power grid of Area B, and multiple measurement data are abnormal". The early warning information is pushed to the power grid operation and maintenance management cloud platform through methods such as text messages, emails and in-station messages. According to the early warning information, operation and maintenance personnel perform corresponding power grid safe operation decision support work, such as arranging maintenance personnel to repair for equipment failure; organizing technical personnel to check and repair the communication link for communication interruption; adjusting the power grid operation mode for power grid failure to ensure the stable operation of the power grid.

[0077] Further, as an embodiment of the present invention, referring to Figure 2 shown, it is Figure 1 the detailed step flow schematic diagram of step S1 in

[0078] Step S11: Real-time collect various types of measurement data in the power grid, including the voltage, current, active power, reactive power, frequency real-time data corresponding to substations, transmission lines and distribution transformers, as well as the auxiliary data corresponding to the data acquisition device status and communication link status;

[0079] In the embodiments of the present invention, various types of measurement data are collected in real time through data collection devices such as smart meters and sensors deployed in substations, transmission lines, and distribution areas. For example, in a certain substation, the smart meter collects voltage, current, active power, reactive power, and frequency data every 15 minutes. These data are transmitted to the data collection terminal in the form of digital signals. At the same time, the status information of the data collection devices themselves (such as device operating status, battery power, etc.) and the communication link status information (such as signal strength, transmission delay, etc.) are also collected. Taking a certain transmission line as an example, the sensor monitors the current and voltage of the line in real time, and the data collection frequency is once per second. These data are transmitted to the data center through the communication network. During the transmission process, the communication link status data is also recorded. For example, during a certain data transmission process, the signal strength is 80% and the transmission delay is 50 milliseconds. All the collected data is aggregated and stored in the database of the data center, forming an original data set containing various types of measurement data and auxiliary data.

[0080] Step S12: Perform deduplication processing on various types of measurement data in the power grid to obtain multi-source deduplicated measurement data of the power grid;

[0081] In the embodiments of the present invention, by reading the original data set from the database of the data center and using the Python programming language in combination with the pandas library for deduplication processing. First, load the original data into a pandas DataFrame. Assume that the data set contains 10,000 records, including duplicate data rows. Use the duplicated() function to find duplicate rows in the DataFrame. For example, duplicate_rows = df.duplicated(), which returns a boolean Series where True indicates that the row is a duplicate row. Then, use the drop_duplicates() function to delete the duplicate rows, such as df = df.drop_duplicates(). In this way, the deduplicated data set is obtained. During the deduplication process, some special situations may be encountered. For example, some data rows are the same in some columns but different in other columns. For this situation, which row of data to retain can be selected according to specific requirements. For example, for voltage data, if there are two records with the same voltage value but different collection times, the latest record can be selected according to the order of collection time. After the deduplication processing, finally, the multi-source deduplicated measurement data of the power grid is obtained and stored in a new database table.

[0082] Step S13: Use the regression method to perform fitting and filling processing on the missing data corresponding to various types of measurements in the multi-source deduplicated measurement data of the power grid to obtain multi-source missing filled measurement data of the power grid;

[0083] In an embodiment of the present invention, by performing fitting and filling processing on the missing data in the deduplicated dataset, a linear regression model in the scikit-learn library of Python is used. First, the dataset is divided into a feature matrix X and a target vector y, where the feature matrix contains all columns except the column of missing data, and the target vector is the column where the missing data is located. For example, for voltage data, assuming there are 100 data points, and 10 of them are missing, the features of the remaining 90 data points (such as time, active power, reactive power, etc.) are used as the feature matrix X, and the voltage values are used as the target vector y. Then, the LinearRegression() function is used to create a linear regression model, and the fit() function is used to train the model, such as model = LinearRegression().fit(X, y). After training, the predict() function is used to predict the missing data. For example, y_pred = model.predict(X_missing), where X_missing is the feature matrix of the missing data. Finally, the predicted values are filled into the positions of the missing data to obtain the filled dataset. In practical applications, situations with a large amount of missing data may be encountered. At this time, more complex regression models, such as polynomial regression models or decision tree regression models, can be used. In addition, other methods, such as interpolation methods or machine learning-based methods, can also be combined to improve the accuracy of filling missing data.

[0084] Step S14: Perform format normalization processing on various measurement data in the multi-source missing filling data of power grid measurements, so as to uniformly convert the measurement data corresponding to different formats and different units into a standard format and unit, in order to obtain multi-source normalized data of power grid measurements.

[0085] In an embodiment of the present invention, by performing format normalization on the filled dataset and using the pandas library and numpy library in Python, first, the measurement data in different formats is uniformly converted. For example, for time data, there are different formats such as "YYYY-MM-DD HH:MM:SS" and "MM / DD / YYYY HH:MM:SS". The to_datetime() function in pandas is used to convert it to a unified format, such as df['time'] = pd.to_datetime(df['time']). Then, the measurement data in different units is unit-converted. For example, for voltage data, such as volts (V) and kilovolts (kV), the unit conversion can be performed by using the where() function in numpy, such as df['voltage']=np.where(df['voltage_unit']=='kV', df['voltage']*1000, df['voltage']). Finally, the formats and units of all measurement data are unified into standard formats and units, obtaining the multi-source normalized power grid measurement data, which is stored in the final database table, providing an accurate data basis for subsequent data quality analysis and monitoring.

[0086] Further, step S2 includes the following steps:

[0087] Step S21: Perform time series synchronization processing on the multi-source normalized power grid measurement data to obtain the corresponding multi-source time series data of the power grid measurement within the same time series range;

[0088] In an embodiment of the present invention, in this step, the multi-source normalized power grid measurement data includes measurement data such as voltage, current, and power from different locations such as substations, transmission lines, and distribution areas, as well as auxiliary data such as the status of data acquisition devices and the status of communication links. Since the acquisition times of these data are different, time series synchronization processing is required. Taking a regional power grid as an example, the voltage data acquisition time intervals of 3 substations are 5 minutes, 10 minutes, and 15 minutes respectively. Based on the time stamp, all data is resampled according to the minimum time interval (5 minutes). For the missing time point data, linear interpolation is used for filling. For example, there is no voltage data record for a certain substation between 10:00 and 10:05. The system calculates the voltage value at the intermediate time point through linear calculation based on the voltage values at 10:00 and 10:05. After resampling and interpolation processing, all measurement data is within the same time series range, forming the multi-source time series data of the power grid measurement, ensuring the consistency and accuracy of subsequent analysis.

[0089] Step S22: Set a time-series sliding window and perform sliding smoothing filtering on the multi-source time-series data of power grid measurements corresponding to the same time range to obtain smoothed multi-source time-series data of power grid measurements;

[0090] In an embodiment of the present invention, for the multi-source time-series data of power grid measurements in the same time range, set the size of the time-series sliding window to 10 time points. Taking the current time-series data of a certain substation as an example, this data contains 100 time-point records. Starting from the first time point, every time 10 consecutive time-point data are selected as a window. Within each window, a moving average filtering algorithm is used for smoothing. The specific operation is to add the 10 current data in the window and then divide by 10 to obtain the average value of this window. This average value is used as the smoothed data at the middle time point of the window. When the window slides backward by one time point, repeat the above calculation process. For example, the first window contains the data of the 1st - 10th time points. After calculating the average value, it is used as the smoothed data at the 5th time point. When the window slides to the 2nd - 11th time points, calculate the new average value as the smoothed data at the 6th time point. In this way, sliding smoothing filtering is performed on all measurement data to eliminate noise and abnormal fluctuations in the data, and finally obtain the smoothed multi-source time-series data of power grid measurements, making the data curve smoother and more stable, which is convenient for subsequent feature extraction.

[0091] Step S23: Perform time-domain feature statistics on the smoothed multi-source time-series data of power grid measurements to statistically calculate the mean, variance, maximum value, minimum value, skewness, kurtosis, and change rate corresponding to various measurement data, and obtain the multi-source time-domain statistical features of power grid measurements;

[0092] In an embodiment of the present invention, through performing time-domain feature statistics on the smoothed multi-source time-series data of power grid measurements, taking the active power data of a certain transmission line as an example, this data contains 200 time-point records after smoothing. Calculate the mean, variance, maximum value, minimum value, skewness, kurtosis, and change rate of this data in sequence. The mean is obtained by adding the active power values of 200 time points and then dividing by 200. The variance is used to measure the degree of dispersion of the data and is obtained by calculating the sum of the squares of the differences between each data point and the mean and then dividing by the number of data points. The maximum value and minimum value are directly found from the data. The skewness reflects the degree of asymmetry of the data distribution, that is , where is the number of data points, represents the i-th data point, is the mean of the data, is the standard deviation, and the kurtosis measures the degree of peakedness of the data distribution, that is ; The rate of change is calculated as the ratio of the difference between data at adjacent time points to the data at the previous time point. The same statistical calculations are performed on all measured data, and finally, the multi-source time-domain statistical characteristics of power grid measurements are obtained. These characteristics can reflect the basic characteristics and change laws of the data in the time dimension.

[0093] Step S24: Perform Fourier transform on various types of measured data in the multi-source normalized data of power grid measurements to generate frequency segments of various types of measured data; perform frequency-domain feature statistics on the frequency segments of various types of measured data to obtain the multi-source frequency-domain statistical characteristics of power grid measurements;

[0094] In the embodiment of the present invention, for various types of measured data in the multi-source normalized data of power grid measurements, such as the voltage data of a certain distribution transformer area, Fourier transform is used to convert it from the time domain to the frequency domain. Fourier transform decomposes the signal of voltage changing with time into the superposition of sine and cosine waves of different frequencies, thereby generating the frequency segment of voltage data. In the frequency domain, the system performs feature statistics on each frequency segment. For example, calculate the energy proportion of each frequency segment, that is, the ratio of the signal energy in the frequency segment to the total energy; count the frequency values and amplitudes of the main frequency components. Suppose that after Fourier transform of the voltage data, the energy proportion in the 50Hz frequency segment is 70% and the amplitude is 220V, and the energy proportion in the 100Hz frequency segment is 20% and the amplitude is 50V. The same Fourier transform and frequency-domain feature statistics are performed on all measured data, and finally, the multi-source frequency-domain statistical characteristics of power grid measurements are obtained. These characteristics help to analyze the frequency characteristics of the data and detect abnormal conditions such as the presence of harmonics.

[0095] Step S25: Perform correlation feature analysis on various types of measured data in the multi-source normalized data of power grid measurements to analyze the degree of association between different measured data and obtain the multi-source correlation characteristics of power grid measurements;

[0096] In the embodiment of the present invention, by performing correlation feature analysis on various types of measured data in the multi-source normalized data of power grid measurements, taking the voltage and current data of a certain substation as an example, the system calculates the Pearson correlation coefficient between the two. By substituting the corresponding time point values of the voltage data and the current data into the Pearson correlation coefficient formula, the calculated correlation coefficient between the two is 0.9, indicating a strong positive correlation between the voltage and the current. The degree of association between different types of measured data will also be analyzed, such as analyzing the correlation between active power and reactive power, frequency and voltage, etc. In addition to the Pearson correlation coefficient, methods such as the Spearman rank correlation coefficient may also be used for verification. By comprehensively analyzing the degree of association between different measured data, the multi-source correlation characteristics of power grid measurements are finally obtained. These characteristics can reveal the internal connections between different physical quantities in the power grid and provide a basis for power grid operation analysis and fault diagnosis.

[0097] Step S26: Analyze the operating status of the data acquisition devices corresponding to the power grid measurement multi-source normalized data to obtain the operating status characteristics of the power grid acquisition devices; evaluate the link stability of the corresponding communication link status data in the power grid measurement multi-source normalized data to obtain the power grid communication link stability characteristics.

[0098] In the embodiment of the present invention, by analyzing the operating status of the data acquisition devices corresponding to the power grid measurement multi-source normalized data, taking a certain data acquisition device as an example, its status data includes information such as device operating duration, battery power, temperature, etc. These data are judged according to preset thresholds. For example, if the device operating duration exceeds 1000 hours, maintenance is required; when the battery power is lower than 20%, it is regarded as insufficient power; when the temperature exceeds 50°C, it means the device is overheated. If a device has an operating duration of 1200 hours, a battery power of 15%, and a temperature of 55°C, it is determined that the device has problems such as too long operating time, insufficient power, and overheating, and the abnormal status of the device is recorded. By analyzing the status data of all data acquisition devices, the operating status characteristics of the power grid acquisition devices are obtained. For the communication link status data, the link stability is evaluated according to the corresponding method. For example, by calculating the instantaneous volatility, long-term fluctuation index, correlation index, etc. of the bandwidth, delay, and packet loss rate of each link, combined with the stable contribution weight of the link in the power grid communication network topology, the stability level of each link is comprehensively evaluated by using the analytic hierarchy process, and finally the power grid communication link stability characteristics are obtained, providing a reference for ensuring the reliability of power grid data transmission.

[0099] Step S27: Merge the power grid measurement multi-source time-domain statistical characteristics, power grid measurement multi-source frequency-domain statistical characteristics, power grid measurement multi-source correlation characteristics, power grid acquisition device operating status characteristics, and power grid communication link stability characteristics into the same set to obtain the power grid measurement multi-dimensional quality characteristic set.

[0100] In the embodiments of the present invention, by combining the multi-source time-domain statistical features of power grid measurement, the multi-source frequency-domain statistical features of power grid measurement, the multi-source correlation features of power grid measurement, the operation state features of power grid acquisition equipment, and the stability features of power grid communication links, a set is established in tabular form. The first column records the feature types, such as "time-domain statistical feature - mean value", "frequency-domain statistical feature - proportion of 50Hz energy", "correlation feature - correlation coefficient between voltage and current", "equipment operation state feature - operation duration of equipment 1", "communication link stability feature - stability level of link 1"; the second column records the corresponding values. For example, the value corresponding to "time-domain statistical feature - voltage mean value of a certain substation" is 225V, the value corresponding to "frequency-domain statistical feature - proportion of 50Hz energy of current in a certain transmission line" is 80%, the value corresponding to "correlation feature - correlation coefficient between active power and reactive power in a certain distribution area" is 0.7, the value corresponding to "equipment operation state feature - battery power of equipment 2" is 30%, and the value corresponding to "communication link stability feature - stability level of link 2" is good. In this way, all features are integrated into the same set, and finally the multi-dimensional quality feature set of power grid measurement is obtained, providing a comprehensive data basis for subsequent measurement quality assessment.

[0101] Step S28: Based on the multi-dimensional quality feature set of power grid measurement, perform measurement quality assessment on various types of measurement data in the power grid to obtain the measurement data quality level score of the power grid.

[0102] In an embodiment of the present invention, the measurement quality of various measurement data in the power grid is evaluated by using the multi-dimensional quality feature set of power grid measurement obtained based on step S27. First, according to historical data and expert experience, weights are set for each quality feature. For example, the weight of the mean value in the time-domain statistical features is set to 0.15, the weight of the energy proportion of the main frequency band in the frequency-domain statistical features is set to 0.1, the weight of the voltage-current correlation coefficient in the correlation features is set to 0.1, the weight of the device operation duration in the device operation state features is set to 0.2, and the weight of the link stability level in the communication link stability features is set to 0.25. Then, the value of each quality feature is multiplied by its corresponding weight, and all the products are added together to obtain a comprehensive quality score. Suppose the values of each quality feature of a certain set of measurement data are as follows: the time-domain mean value is 220V, the energy proportion of 50Hz in the frequency domain is 75%, the voltage-current correlation coefficient is 0.8, the device operation duration is 800 hours (no abnormality), and the link stability level is excellent (corresponding score is 4 points). Then the quality level score of this set of data is: (220×0.15)+(0.75×0.1)+(0.8×0.1)+(1×0.2)+(4×0.25)=33 + 0.075 + 0.08 + 0.2 + 1 = 34.355 points. Finally, the quality level score is divided into different levels according to a predetermined standard, such as 0 - 60 points for poor, 61 - 75 points for medium, 76 - 90 points for good, and 91 - 100 points for excellent, so as to obtain the quality level score of the power grid measurement data. This score can comprehensively reflect the quality status of the power grid measurement data, help operation and maintenance personnel discover data quality problems in a timely manner, and ensure the accuracy of power grid operation monitoring and data analysis.

[0103] Further, the link stability evaluation of the corresponding communication link state data in the power grid multi-source normalized data in step S26 includes the following steps:

[0104] Obtain the bandwidth time-series data, delay time-series data, and packet loss time-series data corresponding to each link in the power grid through the communication link state data;

[0105] In an embodiment of the present invention, bandwidth time series data, delay time series data, and packet loss time series data corresponding to each link are obtained by parsing communication link status data. Assume that the communication link status data is stored in a database in JSON format, and each data record contains information such as link ID, timestamp, bandwidth, delay, and packet loss rate. The data in the database is read using the pandas library in Python. Then, the data is grouped according to the link ID, and the changing data of the bandwidth, delay, and packet loss rate of each link over time are extracted respectively to form corresponding time series data. For example, for the data with link ID "Link-01", the bandwidth time series data is bandwidth_time_series = data[data['link_id'] == 'Link-01']['bandwidth'], the delay time series data is delay_time_series = data[data['link_id'] == 'Link-01']['delay'], and the packet loss time series data is packet_loss_time_series = data[data['link_id'] == 'Link-01']['packet_loss_rate']. These time series data will be used as the basis for subsequent analysis.

[0106] Preferably, basic stability index analysis is performed on the bandwidth time series data, delay time series data, and packet loss time series data corresponding to each link in the power grid to statistically analyze the instantaneous bandwidth, delay, and packet loss volatility corresponding to each link, and obtain the basic power grid stability index corresponding to each link;

[0107] In an embodiment of the present invention, by performing basic stability index analysis on the obtained bandwidth time series data, delay time series data, and packet loss time series data of each link, statistical analysis is carried out using the numpy library and pandas library in Python. Taking the bandwidth time series data as an example, the volatility of the instantaneous bandwidth is calculated. First, the first-order difference of the bandwidth is calculated, that is, the difference in bandwidth between adjacent time points. Then, the standard deviation of the bandwidth is calculated to measure the degree of bandwidth fluctuation. The standard deviation of the bandwidth is divided by the average value of the bandwidth to obtain the instantaneous volatility of the bandwidth. The same method is used to calculate the instantaneous volatility of delay and packet loss, and the basic power grid stability index corresponding to each link is obtained. These indexes reflect the stability of the link in the short term. The smaller the volatility, the more stable the link.

[0108] Preferably, periodic fluctuation detection is performed on the communication link status data to model through an autoregressive model or a moving average model to evaluate the long-term abnormal fluctuation interval corresponding to each link, analyze the corresponding periodic fluctuation data segments of each link, and quantify the fluctuation amplitude, frequency, and change trend corresponding to each link to obtain the power grid volatility index corresponding to each link;

[0109] In an embodiment of the present invention, by performing periodic fluctuation detection on communication link state data, an autoregressive model (AR) or a moving average model (MA) in the statsmodels library of Python is used for modeling. Taking the delayed time series data as an example, first, the stationarity of the data is tested, and the unit root test is performed using the adfuller function. For example:

[0110] python

[0111] from statsmodels.tsa.stattools import adfuller

[0112] def adf_test(series):

[0113] result = adfuller(series)

[0114] print('ADF Statistic: {}'.format(result[0]))

[0115] print('p - value: {}'.format(result[1]))

[0116] print('Critical Values:')

[0117] for key, value in result[4].items():

[0118] print('\t{}: {}'.format(key, value))

[0119] if result[1] <= 0.05:

[0120] print("The series is stationary.")

[0121] else:

[0122] print("The series is non - stationary.")

[0123] adf_test(delay_time_series)

[0124] If the data is stationary, the AR or MA model can be directly used for modeling. If the data is non - stationary, differencing processing needs to be performed to make it stationary. Assume that the delayed time - series data becomes stationary after first - order differencing, and the AR model is used for modeling. For example:

[0125] python

[0126] from statsmodels.tsa.ar_model import AutoReg

[0127] model = AutoReg(delay_time_series_diff, lags = 10)

[0128] model_fit = model.fit()

[0129] The long - term trend of the delay is obtained through model prediction, and the error between the predicted value and the actual value is calculated. According to the error, the long - term abnormal fluctuation intervals corresponding to each link are analyzed. At the same time, by analyzing the parameters of the model, the fluctuation amplitude, frequency, and change trend corresponding to each link are quantified. Finally, the power - grid volatility indicators corresponding to each link are obtained. These indicators reflect the fluctuation conditions of the link in the long term and help to discover potential periodic problems.

[0130] Preferably, correlation analysis is performed among the bandwidth time - series data, delay time - series data, and packet - loss time - series data corresponding to each link in the power grid to analyze the correlation among the bandwidth, delay, and packet - loss indicators corresponding to each link, and the power - grid state correlation indicators corresponding to each link are obtained;

[0131] In the embodiment of the present invention, correlation analysis is performed among the bandwidth time - series data, delay time - series data, and packet - loss time - series data corresponding to each link in the power grid to calculate the correlation coefficient between them by using the corr function in the pandas library of Python. The same method is used to calculate the correlation coefficients between bandwidth and packet - loss, and between delay and packet - loss, and the correlation among the bandwidth, delay, and packet - loss indicators corresponding to each link, that is, the power - grid state correlation indicators. The value range of the correlation coefficient is between - 1 and 1. The closer the absolute value is to 1, the stronger the correlation between the two indicators. If the correlation coefficient between bandwidth and delay is 0.8, it means that there is a strong positive correlation between bandwidth and delay, that is, when the bandwidth increases, the delay also increases.

[0132] Preferably, by constructing the topological structure corresponding to the power - grid communication network and performing link - network stability contribution evaluation according to the topological structure corresponding to the power - grid communication network, the stability contribution weights of each link to the entire power - grid communication network are obtained;

[0133] In an embodiment of the present invention, by constructing a topological structure corresponding to a power grid communication network and evaluating the contribution of the link network to stability based on the topological structure, a topological structure is constructed by using the networkx library of Python. Assuming that the power grid communication network consists of nodes and edges, the nodes represent devices such as substations and distribution areas, and the edges represent communication links. First, an empty graph object is created. Then, according to the actual network connection situation, nodes and edges are added, and weights are assigned to the edges. The weights can represent the bandwidth, delay or other related attributes of the links. Next, the betweenness_centrality function is used to calculate the betweenness centrality of each link. The betweenness centrality reflects the importance of the link in the network. The higher the betweenness centrality, the greater the contribution of the link to the stability of the network. Finally, according to the values of the betweenness centrality, the contribution weights of each link to the stability of the entire power grid communication network are obtained. These weights will be used for the subsequent comprehensive evaluation of link stability.

[0134] Preferably, based on the contribution weights of each link to the stability of the entire power grid communication network and using the analytic hierarchy process, a comprehensive evaluation of link stability is carried out on the power grid basic stability index, power grid volatility index and power grid state correlation index corresponding to each link, so as to divide the stability index corresponding to the link into three levels, including basic stability, volatility and correlation, and calculate the stability level corresponding to each link in the power grid in combination with the contribution weights of each link to obtain the stability characteristics of the power grid communication link.

[0135] In the embodiments of the present invention, based on the contribution weights of each link to the stability of the entire power grid communication network, the analytic hierarchy process is used to comprehensively evaluate the link stability of the power grid basic stability index, the power grid volatility index, and the power grid state correlation index corresponding to each link. First, a hierarchical structure model is constructed, and the link stability evaluation is divided into three levels: the target level (link stability), the criterion level (basic stability, volatility, correlation), and the scheme level (each link). Then, a judgment matrix is constructed, and according to expert experience or actual data, the relative importance between the criterion level and the scheme level is judged. Next, the eigenvector and the maximum eigenvalue of the judgment matrix are calculated, and the eigenvector is normalized to obtain the weight vector of each criterion. Then, according to the basic stability index, volatility index, and correlation index of each link, combined with the weight vector, the stability score of each link is calculated. Suppose the basic stability index of link 1 is 0.8, the volatility index is 0.6, and the correlation index is 0.7, that is, link1_score = 0.8×weights[0]+0.6×weights[1]+0.7×weights[2]. Finally, according to the stability score, the stability index corresponding to the link is divided into different levels. For example, a score above 0.8 is stable, a score between 0.6 and 0.8 is relatively stable, and a score below 0.6 is unstable. In this way, the stability characteristics of the power grid communication link are finally obtained, providing a reference basis for the operation and maintenance of the power grid.

[0136] Further, the analysis process of the instantaneous bandwidth volatility corresponding to each link is specifically as follows:

[0137] By setting the corresponding sliding window size to 3-5 time points, and based on this sliding window size, the bandwidth time series data corresponding to each link in the power grid is divided into bandwidth time series segments to obtain each bandwidth time series window data segment corresponding to each link;

[0138] In an embodiment of the present invention, for the bandwidth time-series data corresponding to each link in the power grid, the sliding window size is set to 3 - 5 time points. For example, it is set to 4 time points. Taking a link numbered "Link-001" as an example, the bandwidth time-series data of this link contains the bandwidth values of 100 consecutive time points recorded in minutes. The data is stored in the "Link-001_bandwidth" table in the system database. Each row corresponds to the bandwidth data of one time point, and the format is "timestamp, bandwidth value (Mbps)". Starting from the first time point, it is divided according to the rule of taking 4 time points as a group. That is, the first group contains the bandwidth data of the 1st - 4th time points, the second group contains the bandwidth data of the 2nd - 5th time points, and so on, until the last group contains the bandwidth data of the 97th - 100th time points. Each group forms a bandwidth time-series window data segment. In this way, the 100-time-point bandwidth time-series data of the "Link-001" link is divided into 97 bandwidth time-series window data segments. Other links are also divided into bandwidth time-series segments in the same way and with the same window size. Finally, each bandwidth time-series window data segment corresponding to each link in the power grid is obtained. These data segments provide basic data units for subsequent analysis of bandwidth fluctuations.

[0139] Preferably, for any two adjacent bandwidth time-series window data segments corresponding to each link, the corresponding bandwidth fluctuation amplitudes are calculated respectively to obtain the bandwidth fluctuation amplitudes corresponding to any two adjacent time series windows of each link;

[0140] In an embodiment of the present invention, based on the previously obtained link bandwidth time series window data segments, taking the "Link-001" link as an example, from its 97 bandwidth time series window data segments, any two adjacent data segments are selected, such as the first group (including the bandwidth data of the first to fourth time points) and the second group (including the bandwidth data of the second to fifth time points). The maximum and minimum values of the bandwidth within these two adjacent bandwidth time series window data segments are calculated respectively. Suppose the maximum value of the bandwidth within the first group of data segments is 50 Mbps and the minimum value is 45 Mbps; the maximum value of the bandwidth within the second group of data segments is 52 Mbps and the minimum value is 48 Mbps. The calculation is performed using the formula "fluctuation amplitude = (maximum value of the latter group - minimum value of the latter group) - (maximum value of the former group - minimum value of the former group)", that is, the bandwidth fluctuation amplitude corresponding to these two adjacent bandwidth time series window data segments is (52 - 48) - (50 - 45) = -1 Mbps. In the same way, all adjacent two bandwidth time series window data segments of the "Link-001" link are calculated to obtain the bandwidth fluctuation amplitude corresponding to any two adjacent time series windows of this link. For other links in the power grid, the above operations are also performed in sequence, so as to obtain the bandwidth fluctuation amplitude corresponding to any two adjacent time series windows of each link. These fluctuation amplitudes reflect the degree of change of the bandwidth within adjacent time windows.

[0141] Preferably, based on the bandwidth fluctuation amplitudes corresponding to any two adjacent time series windows of each link, instantaneous volatility statistical calculation is performed to calculate the instantaneous volatility at the time points between any two adjacent time series windows of each link, and the corresponding instantaneous bandwidth volatility of each link is obtained.

[0142] In an embodiment of the present invention, based on the bandwidth fluctuation amplitudes corresponding to any two adjacent timing windows of each link obtained previously, taking the "Link-001" link as an example, the instantaneous volatility at the time point between any two adjacent timing windows of this link is calculated. Assuming that the data segments of two adjacent bandwidth timing windows are the nth group and the (n + 1)th group respectively, and the time point between them is the (n + 4)th time point (taking the window size as 4 time points as an example), the calculation formula for the instantaneous volatility is "instantaneous volatility = (bandwidth fluctuation amplitude between two adjacent windows / average bandwidth of the previous window) × 100%". Assuming that the average bandwidth of the nth group of windows is 47.5 Mbps and the bandwidth fluctuation amplitude between two adjacent windows is -1 Mbps, then the instantaneous volatility at the (n + 4)th time point is (-1 / 47.5) × 100% ≈ -2.11%. According to this calculation method, the time points between all adjacent two timing windows of the "Link-001" link are calculated to obtain the instantaneous volatility of this link at each time point. For other links in the power grid, the instantaneous volatility at the time point between any two adjacent timing windows is also calculated based on their respective bandwidth fluctuation amplitudes and window average bandwidths, and finally the instantaneous bandwidth volatility corresponding to each link is obtained. These volatility data can intuitively reflect the change stability of the bandwidth of each link in a short time, providing key indicators for evaluating the basic stability of the power grid communication link.

[0143] Further, step S28 includes the following steps:

[0144] Step S281: Regarding each quality feature in the multi-dimensional quality feature set of power grid measurement as a node in the graph, and using a graph convolutional network to analyze the correlation degree between each feature node, and at the same time, performing a spatial feature correlation graph analysis on each quality feature in the multi-dimensional quality feature set of power grid measurement according to the correlation degree between each feature node, so as to generate a spatial correlation graph of power grid measurement quality features;

[0145] In an embodiment of the present invention, by obtaining a multi-dimensional quality feature set of power grid measurements, which includes various quality features such as the instantaneous volatility of voltage and current, the delay and packet loss rate of communication links, etc., these quality features are regarded as nodes in a graph one by one. For example, "voltage instantaneous volatility" is regarded as one node, and "link delay" is regarded as another node. The graph convolutional network (GCN) is used to analyze these nodes. Taking the measurement data of a certain power grid area as an example, this area contains the measurement information of 10 substations, corresponding to a set of 20 quality feature nodes. The GCN analyzes the degree of association between nodes by constructing the adjacency matrix and feature matrix of the nodes. The adjacency matrix describes the connection relationship between nodes. If there is a potential connection between two quality features, the corresponding matrix position is set to 1, otherwise it is 0; the feature matrix stores the quality feature values corresponding to each node. Through the calculation of the multi-layer GCN network, the degree of association between each feature node is analyzed. For example, it is found that the degree of association between the "voltage instantaneous volatility" node and the "line current fluctuation" node is 0.7, indicating a strong association between the two. Based on these calculation results of the degree of association, a spatial feature association graph analysis is performed on all quality features to determine the connection weights and directions between nodes, and finally a spatial association graph of power grid measurement quality features is generated. This graph is presented in a visual graph, where nodes represent quality features, edges represent the associations between features, and the thickness of the edges represents the strength of the degree of association, providing a structured data basis for subsequent quality assessment modeling.

[0146] Step S282: Use a weighted fusion neural network to perform deep quality assessment modeling on the spatial association feature nodes in the spatial association graph of power grid measurement quality features to generate a multi-dimensional quality assessment model for power grid measurement;

[0147] In an embodiment of the present invention, based on the generated spatial correlation map of power grid measurement quality characteristics, a weighted fusion neural network is used to perform deep modeling on the quality assessment of spatial correlation feature nodes in the map. The weighted fusion neural network consists of multiple convolutional layers, pooling layers, and fully connected layers. The information of the spatial correlation feature nodes is used as the input of the neural network. The feature values of each node are input into the network after being normalized. Taking two correlation nodes, "voltage instantaneous volatility" and "link delay", as examples, their feature values are -2.11% and 50 milliseconds respectively, and after normalization, they are input into the network. The convolutional layer extracts the local correlation information of the node features through convolutional kernels of different sizes, and the pooling layer performs dimensionality reduction on the data to retain key features. During the network training process, historical power grid measurement data and their corresponding actual quality levels are used as training samples. For example, 1000 sets of measurement data and their manually labeled quality levels (divided into four levels: excellent, good, medium, and poor, corresponding to 4, 3, 2, and 1 points respectively) within the past month are selected. The network weights are adjusted through the backpropagation algorithm to minimize the error between the quality assessment result output by the network and the actual quality level. After multiple rounds of training, the network learns the complex relationships between various spatial correlation feature nodes, and finally generates a multi-dimensional quality assessment model for power grid measurement that can accurately evaluate the quality of power grid measurement.

[0148] Step S283: Input various types of measurement data in the power grid into the multi-dimensional quality assessment model for power grid measurement to perform measurement quality assessment, calculate the correlation coefficients between the corresponding measurement data and each spatial correlation feature node according to the weighted fusion neural network, and calculate the corresponding quality level scores by weighting, so as to obtain the quality level score of the power grid measurement data.

[0149] In an embodiment of the present invention, various measurement data collected in real time from the power grid, such as voltage, current, active power, reactive power, and frequency data of a certain substation at a certain moment, as well as data such as the bandwidth, delay, and packet loss rate of the communication link, are input into the previously generated multi-dimensional quality assessment model of the power grid measurement. The input data is processed through a weighted fusion neural network to calculate the correlation coefficients between the input data and each spatial correlation feature node in the quality feature space association map. For example, the correlation coefficient between the input voltage data and the "voltage instantaneous volatility" node is calculated as 0.8, and the correlation coefficient with the "link delay" node is 0.3. According to the weights of each feature node in the map and the calculated correlation coefficients, weighted calculation is performed. Assuming the weight of the "voltage instantaneous volatility" node is 0.6 and the weight of the "link delay" node is 0.4, the partial quality score corresponding to this measurement data is (0.8×0.6 + 0.3×0.4)×100 = 60 points. Considering the weighted calculation results of all relevant feature nodes comprehensively, the final quality grade score is obtained. The quality grade score is divided into different grades according to a predetermined standard, such as 0 - 60 points for poor, 61 - 75 points for medium, 76 - 90 points for good, and 91 - 100 points for excellent, so as to obtain the quality grade score of the power grid measurement data. This score can intuitively reflect the quality status of the power grid measurement data and provide an important reference basis for power grid operation monitoring and maintenance.

[0150] Further, step S3 includes the following steps:

[0151] Step S31: Obtain seasonal, periodic, and real-time operating condition data corresponding to the power grid operation;

[0152] In an embodiment of the present invention, through sensors, smart meters, and data acquisition devices deployed at various nodes of the power grid, seasonal, periodic, and real-time operating condition data corresponding to the operation of the power grid are acquired. The seasonal data is collected over a span of the past 5 years with an annual cycle, covering the four seasons of spring, summer, autumn, and winter, and includes data such as daily power consumption load, power generation power, voltage, and current within each season, which is stored in the "Power Grid Seasonal Data" table of the system database. The periodic data has a weekly cycle and records information such as the change in power consumption load every 24 hours within a week and the power distribution at different time periods, which is stored in the "Power Grid Periodic Data" table. The real-time operating condition data is collected once per second through a real-time communication network for data such as the power generation power of each power generation unit in the power grid, the transmission power of the transmission line, and the power consumption load of the distribution area, and is updated in real time to the "Power Grid Real-Time Data" table. For example, at the moment of 10:00:00 on a certain day, the power generation power of thermal power plant A is collected in real time as 500 MW, the current of a certain transmission line is 1000 A, and the power consumption load of a certain distribution area is 200 kW. These data are immediately stored in the "Power Grid Real-Time Data" table; at the same time, the power consumption load data at the same moment every day in the past week will be regularly summarized and supplemented to the "Power Grid Periodic Data" table; and each month, the data of the current month will be classified by season and sorted into the "Power Grid Seasonal Data" table to ensure the integrity and timeliness of the data.

[0153] Step S32: Analyze the seasonal and periodic operating condition data corresponding to the operation of the power grid to obtain the load level of the power grid operation condition;

[0154] In an embodiment of the present invention, through analyzing the data in the "Power Grid Seasonal Data" table and the "Power Grid Periodic Data" table for the load level of the power grid, for the seasonal data, taking summer as an example, the power consumption load data of each month in the past 5 years in summer is extracted, and the daily power consumption load is classified and summarized according to time periods (such as 0-6 hours, 6-12 hours, 12-18 hours, 18-24 hours), and the average power consumption load, maximum power consumption load, and minimum power consumption load of each time period are calculated. Suppose it is calculated that the average power consumption load from 6-12 hours in summer in the past 5 years is 800 MW, the maximum power consumption load reaches 1200 MW, and the minimum power consumption load is 500 MW. For the periodic data, analyze the change rule of the power consumption load every day within a week, calculate the average power consumption load on weekdays (Monday to Friday) and weekends (Saturday and Sunday), and compare to find that the average power consumption load on weekdays is 600 MW and the average power consumption load on weekends is 400 MW. Through comprehensive analysis of the seasonal and periodic data, a curve of the power consumption load changing with time is drawn, and combined with a trend prediction model, the predicted values of the load levels of the power grid operation conditions at different time periods within the next quarter are obtained. For example, it is predicted that the load level from 6-12 hours on weekdays next month will reach 650 MW, providing a basis for the resource allocation and operation planning of the power grid.

[0155] Step S33: Calculate the power generation ratio of the power grid for the real-time operating conditions corresponding to the power grid operation to obtain the power generation ratio of the power grid operation conditions.

[0156] In the embodiment of the present invention, by calculating the power generation ratio of the power grid based on the real-time operating conditions data in the "power grid real-time data" table obtained previously, all power generation units in the power grid are first classified, including thermal power plants, hydropower plants, wind power plants, photovoltaic power stations, etc. Taking a certain moment as an example, the real-time data shows that the power generation power of thermal power plant B is 300 MW, the power generation power of hydropower plant C is 150 MW, the power generation power of wind power plant D is 80 MW, and the power generation power of photovoltaic power station E is 50 MW. Calculate the total power generation power of the power grid as 300 + 150 + 80 + 50 = 580 MW. Then calculate the power generation ratio of each power generation unit respectively. The power generation ratio of thermal power plant B is 300÷580×100%≈51.72%, the power generation ratio of hydropower plant C is 150÷580×100%≈25.86%, the power generation ratio of wind power plant D is 80÷580×100%≈13.79%, and the power generation ratio of photovoltaic power station E is 50÷580×100%≈8.62%. These data are summarized and calculated once per hour to form the power generation ratio data of each power generation unit at different time periods, and are recorded in the "power grid power generation ratio data" table, clearly presenting the contribution ratio of different power generation units in the total power generation under the power grid operation conditions, which is convenient for monitoring the energy structure and power generation stability of the power grid.

[0157] Step S34: Conduct dynamic threshold statistics on the power grid quality according to the load level of the power grid operation conditions and the power generation ratio of the power grid operation conditions to obtain the dynamic threshold of the power grid operation quality.

[0158] In an embodiment of the present invention, by statistically analyzing the dynamic thresholds of grid quality based on the previously obtained grid operation condition load level and the power generation ratio in the grid operation condition, a dynamic threshold calculation model is established. This model comprehensively considers the load level, power generation ratio, and the fluctuation range of historical data. Taking the calculation of the voltage quality threshold as an example, when the grid operation condition load level is high, such as reaching 120% of the predicted value (assuming the predicted value is 650 MW, and the load reaches 780 MW at this time), and the thermal power plant power generation ratio exceeds 60%, combined with the voltage fluctuation range in the historical data under the conditions of high load and high thermal power generation ratio (assuming the historical data shows that the voltage fluctuation range is ±10% of the rated voltage at this time), through the method of weighted average, the calculated dynamic threshold of voltage quality is ±8% of the rated voltage. For the frequency quality threshold, when the load level is lower than 80% of the predicted value (i.e., below 520 MW), and the power generation ratio of renewable energy such as wind power and photovoltaic exceeds 40%, the working condition for calculating the dynamic threshold of frequency quality is met. This is because under this combination of load and energy structure, the grid frequency is easily affected by the intermittency and volatility of renewable energy generation, and it is necessary to re-determine a suitable frequency threshold range. To extract the frequency change situation in the same or similar working conditions (i.e., the load level is lower than 80% of the predicted value and the renewable energy generation ratio exceeds 40%) from historical data, assume that the grid system has collected 100 sets of data that meet this working condition in the past 3 years, and recorded the statistical quantities such as the fluctuation range, average value, and standard deviation of the frequency in these data. After analysis, it is found that the frequency in these data mainly distributes between 49.8 Hz and 50.2 Hz, the average value is 50 Hz, and the standard deviation is 0.1 Hz. Based on the statistical analysis results of historical data, considering providing a certain safety margin for grid operation and ensuring the stability requirements of frequency, taking the average value of historical data, 50 Hz, as the center, and selecting twice the standard deviation (i.e., 0.1 Hz × 2 = 0.2 Hz) as the fluctuation range, thus determining the dynamic threshold of frequency quality as 50 ± 0.2 Hz. Such a set threshold can not only reflect the actual fluctuation characteristics of frequency under this working condition, but also avoid frequent triggering of warnings due to small fluctuations in frequency to a certain extent, ensuring the accuracy and effectiveness of grid operation monitoring. The calculated dynamic thresholds of various grid quality indicators, such as voltage, frequency, power factor, etc., are stored in the "Grid Quality Dynamic Threshold" table, updated in real time and used to monitor the grid operation quality. When the actual operation data exceeds the threshold range, the system immediately issues a warning so that the staff can take measures in time to ensure the stable operation of the grid.

[0159] Further, step S32 includes the following steps:

[0160] Perform power load spectrum conversion on the seasonal and periodic working condition data corresponding to grid operation to generate the grid load sine wave signal spectrum corresponding to the grid operation power load under seasonal and periodic conditions;

[0161] In an embodiment of the present invention, by collecting the electricity load data of a provincial power grid over the past three years, the data covering different seasons, months, dates, and various time periods of a day, and storing them in the "Grid Load Historical Data" table of the system database, each record contains information such as time stamps and load power (kW). The Fourier transform is used to perform electricity load spectrum conversion on the seasonal and periodic operating condition data corresponding to the grid operation. Taking the seasonal data with a one-year cycle as an example, the load power data for each hour within a year is used as a set of input data. The Fourier transform decomposes these load signals that change over time into the superposition of sine and cosine waves with different frequencies, thereby generating the grid load sine wave signal spectrum corresponding to the electricity load under seasonality in the grid operation. For periodic operating condition data, such as the load changes on each day within a week, the Fourier transform is also performed with the load data for each hour within a week as the input to obtain the grid load sine wave signal spectrum corresponding to the period. For example, after the Fourier transform, in the seasonal spectrum, it is found that the low-frequency components related to the one-year cycle account for a relatively large proportion, indicating that seasonality has a significant impact on the load; in the periodic spectrum, the frequency components related to the one-week cycle are prominent, reflecting the periodic pattern of the electricity load within a week. These spectra are presented in the form of visual graphs and data tables for subsequent analysis.

[0162] Preferably, statistical analysis of the load changes is performed on the grid load sine wave signal spectra corresponding to the electricity load under seasonality and periodicity in the grid operation to obtain the seasonal and periodic load change trends of the grid operation conditions;

[0163] In an embodiment of the present invention, by performing statistical analysis of the load changes on the previously generated grid load sine wave signal spectra. Taking the seasonal spectrum as an example, the different frequency components in the spectrum are sorted according to the energy magnitude, and key attention is paid to the load change characteristics corresponding to the frequency components with a relatively large energy proportion. It is found that in the seasonal spectrum, the frequency components corresponding to the winter heating period and the peak period of summer air conditioner use have relatively high energy. By calculating the energy changes of these key frequency components in the same season of different years, a curve of energy change over time is plotted. For example, it is found that the energy of the frequency component corresponding to the winter heating period has increased year by year in the past three years, indicating an increasing trend in the winter electricity load. For the periodic spectrum, the load changes on different dates within a week and at different time periods within a day are statistically analyzed, and statistical quantities such as the average load and load standard deviation at the same time period every day are calculated to analyze the periodic fluctuation pattern of the load within a week. For example, it is found that the electricity consumption peaks on weekdays are concentrated at 9-11 am and 2-5 pm, while the electricity consumption peaks on weekends are relatively scattered and the overall load level is lower. Through these analyses, the seasonal and periodic load change trends of the grid operation conditions are obtained, providing a basis for load forecasting.

[0164] Preferably, load level prediction and analysis are performed according to the seasonal and periodic load change trends of the grid operation conditions to obtain the load level of the grid operation conditions.

[0165] In the embodiment of the present invention, according to the previously obtained seasonal and periodic load change trends of the grid operation conditions, a time series prediction method is used for load level prediction and analysis. Taking the seasonal load change trend as an example, the seasonal load data of the past three years are selected as training samples, and the moving average method, exponential smoothing method or more complex ARIMA model is used for modeling. Assuming the ARIMA model is used, first, the stationarity test of the seasonal load data is carried out. If the data is not stationary, differencing is performed to make it stationary. Then, the parameters of the model are determined through the autocorrelation function (ACF) and partial autocorrelation function (PACF), such as the p, d, q values in ARIMA (p, d, q). After model training and parameter optimization, a model suitable for this seasonal load prediction is obtained. The trained model is used to predict the future seasonal load level. For example, the load level during the winter heating period of the next year is predicted. According to the model output results, appropriate adjustments are made in combination with the actual situation, such as considering factors such as the access of new large industrial users and changes in residents' electricity consumption habits. For periodic load prediction, a suitable time series model is also used, and the load change rules at different dates within a week and different time periods within a day are combined for prediction. Finally, the load level of the grid operation conditions is obtained, and the prediction results are presented in the form of tables and charts, including information such as the predicted load power and prediction error range for different time periods, providing decision support for grid dispatching, equipment planning, etc.

[0166] Further, step S33 includes the following steps:

[0167] Perform an analysis of the power generation load distribution of different power generation units in the real-time condition data corresponding to the grid operation to obtain the power generation load distribution corresponding to different power generation units in the grid operation conditions;

[0168] In an embodiment of the present invention, by retrieving the real-time operating condition data corresponding to the power grid operation, these data are from the monitoring devices of each power generation unit in the power grid and stored in the "Real-time Power Generation Data" table of the system database. Each record contains information such as the power generation unit number, timestamp, and power generation load power (MW). Taking a regional power grid as an example, this power grid includes 10 power generation units in total, namely thermal power plants, hydropower plants, wind farms, and photovoltaic power stations. The power generation load data of different power generation units are sorted by timestamp, with an hourly statistical period, and the average power generation load of each power generation unit within each hour is calculated. For example, at a certain moment, the power generation load power of thermal power plant A is 300 MW, the power generation load power of hydropower plant B is 150 MW, and the power generation load power of wind farm C is only 50 MW due to low wind speed. Through the statistics of data for multiple hours, the power generation load distribution of each power generation unit at different time periods is obtained, and these data are plotted into a bar chart, with the abscissa being the power generation unit number and the ordinate being the power generation load power, clearly showing the power generation load distribution corresponding to different power generation units under the power grid operation conditions and intuitively presenting the differences in the power generation contributions of each power generation unit in the power grid.

[0169] Preferably, power transmission mining analysis is carried out between different power generation units in the real-time operating condition data corresponding to the power grid operation to obtain the power generation power transmission relationship between different power generation units under the power grid operation conditions;

[0170] In an embodiment of the present invention, based on the previously sorted real-time operating condition data of the power grid operation, power transmission mining analysis is carried out between different power generation units by using the power flow analysis method. In the power grid, each power generation unit is interconnected through transmission lines, and power is transmitted on these lines. According to the topological structure of the power grid, a network model including power generation unit nodes and transmission lines is constructed. Taking two adjacent power generation units as an example, thermal power plant D and hydropower plant E are connected by a transmission line. According to the real-time monitored voltage and current data at both ends of the line, combined with the line parameters (such as resistance and reactance), Kirchhoff's laws and Ohm's law are used to calculate the power transmission direction and magnitude on this line. For example, in the power grid, thermal power plant D and hydropower plant E are connected by a transmission line. It is known that the resistance of this line R = 5Ω and the reactance X = 10Ω. The real-time monitored voltage data at both ends of the line are: the voltage on the side of thermal power plant D , the voltage on the side of hydropower plant E , calculate the line current. According to Kirchhoff's voltage law (KVL), in a closed loop, the sum of the voltage drops across all components is equal to zero. For this transmission line, the voltage equation can be listed: , where is the line current, is the imaginary unit, substitute the known data into the equation to solve for the current , first perform the operation on the voltage phasor: , and the line impedance =5 + 10 , then the current , through complex number operations, multiply both the numerator and denominator by the conjugate complex number of the denominator 5 - 10 , and we get , then the current in phasor form is . The direction of the current is reflected by the phasor angle. A negative angle indicates that the current flows from the high - voltage side (side D of the thermal power plant) to the low - voltage side (side E of the hydropower plant). Calculate the magnitude of power transmission according to the power calculation formula (where is the complex power, is the voltage phasor, is the conjugate complex number of the current phasor). Calculate the transmission power using the voltage on side D of the thermal power plant and the calculated current. The conjugate complex number of the current is = , then the complex power . Therefore, the active power P = Re(S)=209 MW, and the reactive power Q = Im(S)=418 Mvar. It can be concluded that the active power transmitted from the thermal power plant D to the hydropower plant E on this line is 209 MW, and the reactive power is 418 Mvar. This indicates that part of the power of the thermal power plant D is transmitted to the hydropower plant E through this line. Perform the same calculation and analysis on all the connection lines between the power generation units in the power grid to determine the power generation power transmission relationship between different power generation units and record it in matrix form. The elements in the matrix represent the power transmission amount between two power generation units, so as to comprehensively understand the flow direction and distribution of power generation power in the power grid.

[0171] Preferably, based on the power generation power transmission relationship between different power generation units under different grid operating conditions, perform a power generation dispatch attenuation assessment on different power generation units in the real - time operating condition data corresponding to the grid operation to obtain the power generation dispatch attenuation coefficient between different power generation units under different grid operating conditions;

[0172] In the embodiment of the present invention, through the power generation dispatch attenuation assessment based on the previously obtained power generation power transmission relationship between different power generation units under different grid operating conditions, the power generation dispatch attenuation mainly considers the influence of factors such as transmission line losses and equipment conversion efficiency on power generation power, and establishes a power generation dispatch attenuation assessment model. This model includes a transmission line loss calculation module and an equipment efficiency correction module. Taking a transmission line with a length of 100 kilometers as an example, it is known that the line resistance is 0.1 Ω / km. When 100 MW of power is transmitted through this line, according to Joule's law P = I 2 R (where I is the current and R is the resistance), combined with the relationship between power, voltage, and current P = UI, calculate the power loss of the line. Assuming that the calculated line loss is 5 MW, then the power transmission efficiency of this line is , that is, the power generation dispatching attenuation coefficient is 1 - 95% = 0.05. For the internal equipment of the power generation unit, such as generators and transformers in the power plant, the output power is corrected according to the rated parameters and actual operating efficiency of the equipment. By analyzing and calculating each line and equipment on the power transmission path between all power generation units in the power grid one by one, the power generation dispatching attenuation coefficient between different power generation units under different grid operating conditions is finally obtained and recorded in tabular form. The table contains information such as power generation unit pairs (such as "thermal power plant F - wind farm G"), transmission paths, attenuation coefficients, etc., providing data support for accurately evaluating the power generation dispatching effect.

[0173] Preferably, based on the power generation dispatching attenuation coefficient between different power generation units under different grid operating conditions, the power generation proportion in the grid is calculated between the power generation load distributions corresponding to different power generation units under different grid operating conditions, and the power generation proportion in the grid operating conditions is obtained.

[0174] In the embodiment of the present invention, by calculating the power generation proportion in the grid for the power generation load distributions corresponding to different power generation units based on the previously obtained power generation dispatching attenuation coefficient between different power generation units under different grid operating conditions. First, the actual power generation contribution of each power generation unit is corrected according to the power generation dispatching attenuation coefficient. For example, if the original power generation load of power generation unit H is 200 MW and the power generation dispatching attenuation coefficient during its power transmission to other units is 0.1, then the corrected actual power generation contribution is , and at the same time, the total power generation load of the power grid is calculated, and the corrected actual power generation contributions of all power generation units are added together. Assuming that the total sum of the corrected actual power generation contributions of 10 power generation units in the power grid is 1500 MW and the corrected actual power generation contribution of power generation unit I is 150 MW, then the power generation proportion of power generation unit I in the power grid is 150÷1500×100% = 10%. The same calculation is performed for each power generation unit, and finally the power generation proportion of each power generation unit under the grid operating conditions is obtained, forming a power generation proportion report. The report not only contains the power generation proportion values of each power generation unit, but also compares and analyzes the change trends of the power generation proportion at different times and under different operating conditions, providing an important basis for the power generation dispatching decision-making and energy structure optimization of the power grid.

[0175] Further, step S4 includes the following steps:

[0176] Step S41: Based on the power grid measurement data quality grade score and combined with the power grid operation quality dynamic threshold, measurement anomaly detection is performed on various types of measurement data in the power grid. When the power grid measurement data quality grade score is between 0 - 60 points or the corresponding measurement data in the power grid exceeds the power grid operation quality dynamic threshold, the corresponding measurement data is determined as data anomaly until all types of measurement data are detected to obtain the power grid measurement quality anomaly data result;

[0177] In an embodiment of the present invention, by retrieving the quality grade scores of each measurement data from the "Grid Measurement Data Quality Grade Score" table, and at the same time obtaining the dynamic thresholds corresponding to various measurement data such as voltage, frequency, power factor, etc. from the "Grid Quality Dynamic Threshold" table. Taking the grid measurement data at a certain moment as an example, the quality grade score of the voltage measurement data of the distribution transformer area numbered M-001 is 55 points, which is between 0 and 60 points. According to the determination rule, this voltage measurement data is preliminarily determined to be abnormal. For the frequency measurement data of the transmission line numbered M-002, its quality grade score is 75 points, but the actual measured value is 50.5 Hz, which exceeds the range of the dynamic threshold of 50±0.2 Hz, and is also determined to be abnormal data. According to the time sequence of data collection and the device number, the above detection operations are sequentially performed on all the measurement data in the grid, and the data that meets the abnormal determination conditions is marked as abnormal data and recorded in the "Grid Measurement Quality Abnormal Data Temporary Table" until the detection of various measurement data is completed, and finally a complete grid measurement quality abnormal data result is formed, providing an accurate data basis for subsequent processing.

[0178] Step S42: Perform abnormal classification processing on each abnormal measurement data in the grid measurement quality abnormal data result to obtain various types of grid abnormal measurement data;

[0179] In an embodiment of the present invention, by classifying the data in the "Grid Measurement Quality Abnormal Data Result" according to the preset abnormal classification rules. Taking voltage abnormal data as an example, if the voltage value of a certain data point fluctuates greatly instantaneously, and the difference from the data at the previous and subsequent time points is obvious, and after checking, no abnormalities are found in other relevant measurement data, then this data is classified as an individual data point abnormality; if the voltage measurement data of multiple distribution transformer areas in a certain area are continuously low or high at the same time, and the current, power and other data in this area also show corresponding abnormal change trends, it is determined to be a large-area data abnormality caused by equipment failure, communication interruption or grid failure. For frequency abnormal data, if the frequency measurement data of a single power generation unit is abnormal while the overall grid frequency is not significantly affected, it is classified as an individual data point abnormality; if the frequency measurement data of multiple power generation units deviate from the normal range at the same time and the grid frequency regulation system shows an abnormal response, it is judged as a large-area data abnormality. By detailed analysis and comparison of each abnormal measurement data, it is accurately classified into the corresponding abnormal type to clearly present the distribution of various abnormal data, facilitating subsequent targeted analysis and processing.

[0180] Step S43: Perform grid abnormal traceability analysis on various types of grid abnormal measurement data to obtain the causes of abnormalities corresponding to various abnormal measurement data, including individual data point abnormalities and large-area data abnormalities caused by equipment failure, communication interruption and grid failure;

[0181] In the embodiments of the present invention, through the power grid anomaly traceability analysis of various previously classified abnormal measurement data, for individual data point anomalies, the measurement data, equipment operation status data, and communication link status data within a certain period before and after the data point are retrieved. For example, if there is an abnormal spike in the current data of a certain distribution transformer area at a certain moment, by checking the equipment operation logs before and after that moment, it is found that due to the startup of a large device near that moment, the instantaneous current fluctuates, and it is determined that this is the reason for the individual data point anomaly. For large-area data anomalies, through comprehensive analysis by combining the power grid topology structure, equipment operation parameters, and historical fault records. If the voltage measurement data of multiple substations in a certain area is simultaneously abnormally low, and the power transmission in this area is abnormal, through inspection, it is found that a short-circuit fault has occurred in a certain key transmission line in this area, resulting in a voltage drop, which in turn affects the measurement data of surrounding equipment, and it is determined that this is the reason for the large-area data anomaly caused by equipment failure. If the measurement data in a certain area has not been updated for a long time, and the communication link status shows a signal interruption, it is judged as a large-area data anomaly caused by communication interruption. The cause of the anomaly corresponding to each abnormal measurement data is detailedly recorded in the "Power Grid Abnormal Measurement Data Traceability Result Table" to provide a clear basis for subsequent early warnings and decision-making.

[0182] Step S44: Perform intelligent early warning monitoring according to the causes of anomalies corresponding to various abnormal measurement data. If it is an individual data point anomaly, a yellow early warning is issued; if it is a large-area data anomaly caused by equipment failure, communication interruption, and power grid failure, a red early warning is issued to generate corresponding power grid measurement data anomaly early warning information, and the power grid measurement data anomaly early warning information is pushed to the relevant power grid operation and maintenance management cloud platform through text messages, emails, and in-site messages to execute corresponding power grid safe operation decision support work, including power grid operation mode adjustment plans, equipment maintenance plans, and communication link optimization plans.

[0183] In the embodiments of the present invention, intelligent early warning monitoring is carried out according to the cause of the anomaly recorded in the "Grid Abnormal Measurement Data Traceability Result Table". For the situation where individual data points are determined to be abnormal, such as the power factor data of a certain transmission line occasionally deviating slightly from the normal range, a yellow early warning message is immediately generated, including the equipment number, measurement parameter, abnormal time, and preliminary cause of the abnormal data, such as "Equipment number L-003, the power factor data was abnormal at 14:23 on October 15, 2024, due to transient load fluctuations". If it is detected that a large-area data anomaly is caused by equipment failure, communication interruption, or grid failure, such as multiple measurement data anomalies in a certain regional grid due to substation equipment failure, a red early warning message is quickly generated to elaborate on the abnormal situation, such as "Equipment failure occurred in Grid A of Region A at 9:00 on October 16, 2024, involving abnormal data such as voltage, current, and power of multiple substations". And the early warning message is pushed to the relevant grid operation and maintenance management cloud platform through three methods: text message, email, and in-station message. After receiving the early warning message, the operation and maintenance personnel execute the corresponding grid safe operation decision-making work according to the preset decision support plan. For equipment failure, start the equipment maintenance plan and arrange professional personnel to go to the faulty equipment for maintenance; for communication interruption, formulate a communication link optimization plan, check the communication fault points and repair them; for grid failure, adjust the grid operation mode to ensure the stable operation of the grid, and ensure that timely and effective countermeasures can be taken when abnormal situations occur to reduce the impact on grid operation.

Claims

1. A method for intelligent monitoring of power grid measurement data quality, characterized in that, Includes the following steps: Step S1: By collecting various measurement data in the power grid in real time, and filling, cleaning and normalizing the various measurement data in the power grid, multi-source normalized data of power grid measurement is obtained. Step S2: Perform multi-dimensional quality feature analysis on the multi-source normalized data of power grid measurements to obtain a multi-dimensional quality feature set of power grid measurements; The measurement quality is assessed based on a multi-dimensional quality feature set of power grid measurements to obtain a power grid measurement data quality level score; wherein the measurement quality assessment includes the following steps: By treating each quality feature in the multidimensional quality feature set of power grid measurement as a node in a graph, and using graph convolutional networks to analyze the correlation between each feature node, spatial feature correlation graph analysis is performed on each quality feature in the multidimensional quality feature set of power grid measurement based on the correlation between each feature node, so as to generate a spatial correlation graph of power grid measurement quality features. A weighted fusion neural network is used to perform in-depth quality assessment modeling of spatial correlation feature nodes within the spatial correlation map of power grid measurement quality features, so as to generate a multi-dimensional quality assessment model for power grid measurements. Various measurement data in the power grid are input into the power grid measurement multidimensional quality assessment model for measurement quality assessment. The correlation coefficient between the corresponding measurement data and each spatially associated feature node is calculated based on the weighted fusion neural network, and the corresponding quality level score is calculated by weighting. The power grid measurement data quality level score is obtained. Step S3: Obtain seasonal, periodic and real-time operating condition data corresponding to power grid operation, and perform dynamic threshold statistics on power grid quality based on the seasonal, periodic and real-time operating condition data corresponding to power grid operation to obtain dynamic thresholds for power grid operation quality. Step S4: Based on the power grid measurement data quality level score and combined with the power grid operation quality dynamic threshold, perform measurement anomaly detection on various types of measurement data in the power grid to obtain power grid measurement quality anomaly data results; perform anomaly classification and intelligent early warning monitoring on the power grid measurement quality anomaly data results to generate corresponding power grid measurement data anomaly early warning information, and execute corresponding power grid safety operation decision support work.

2. The intelligent monitoring method for power grid measurement data quality according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Collect various measurement data in the power grid in real time, including real-time data on voltage, current, active power, reactive power, and frequency of substations, transmission lines, and distribution substations, as well as auxiliary data corresponding to the status of data acquisition equipment and communication links. Step S12: Perform deduplication processing on various types of measurement data in the power grid to obtain multi-source deduplicated power grid measurement data; Step S13: Use regression to fit and fill in the missing data corresponding to various measurements in the multi-source deduplication data of power grid measurements, so as to obtain the missing data of multi-source power grid measurements. Step S14: Perform format normalization processing on various types of measurement data in the multi-source missing data of power grid measurement to convert measurement data with different formats and units into standard formats and units to obtain normalized multi-source data of power grid measurement.

3. The intelligent monitoring method for power grid measurement data quality according to claim 1, characterized in that, The process of obtaining the multidimensional quality feature set of power grid measurements in step S2 includes the following steps: Step S21: Perform time-series synchronization processing on the normalized data of multiple power grid measurement sources to obtain the corresponding time-series data of multiple power grid measurement sources within the same time series range; Step S22: By setting a time-series sliding window, and performing sliding smoothing filtering on the corresponding multi-source time-series data of power grid measurements within the same time-series range based on the time-series sliding window, smoothed multi-source time-series data of power grid measurements are obtained. Step S23: Perform time-domain feature statistics on the multi-source time-series smoothed data of power grid measurements to statistically calculate the mean, variance, maximum value, minimum value, skewness, kurtosis and rate of change of various types of measurement data, and obtain the time-domain statistical characteristics of the multi-source power grid measurements. Step S24: Perform Fourier transform on various types of measurement data within the multi-source normalized data of power grid measurements to generate frequency bands for various types of measurement data; perform frequency domain feature statistics on various types of measurement data frequency bands to obtain the frequency domain statistical characteristics of multi-source power grid measurements; Step S25: Perform correlation feature analysis on various types of measurement data within the multi-source normalized data of power grid measurements to analyze the degree of correlation between different measurement data and obtain the multi-source correlation features of power grid measurements. Step S26: Analyze the equipment operation status data of the corresponding data acquisition equipment in the multi-source normalized data of power grid measurement to obtain the operation status characteristics of the power grid acquisition equipment; evaluate the link stability of the corresponding communication link status data in the multi-source normalized data of power grid measurement to obtain the stability characteristics of the power grid communication link. Step S27: Combine the time-domain statistical characteristics of multiple power grid measurement sources, the frequency-domain statistical characteristics of multiple power grid measurement sources, the correlation characteristics of multiple power grid measurement sources, the operating status characteristics of power grid acquisition equipment, and the stability characteristics of power grid communication links into the same set to obtain a multi-dimensional quality feature set of power grid measurement.

4. The intelligent monitoring method for power grid measurement data quality according to claim 3, characterized in that, Step S26, which involves evaluating the link stability of the corresponding communication link status data within the multi-source normalized data of power grid measurements, includes the following steps: The bandwidth timing data, delay timing data, and packet loss timing data of each link in the power grid are obtained by using communication link status data. Basic stability index analysis is performed on the bandwidth time series data, delay time series data and packet loss time series data corresponding to each link in the power grid, so as to statistically analyze the instantaneous bandwidth, delay and packet loss volatility corresponding to each link and obtain the basic stability index of the power grid corresponding to each link. Periodic fluctuation detection is performed on communication link status data. The long-term abnormal fluctuation range of each link is evaluated by modeling with an autoregressive model or a moving average model. The periodic fluctuation data segments of each link are analyzed, and the fluctuation amplitude, frequency and trend of each link are quantified to obtain the power grid volatility index of each link. Correlation analysis is performed on the bandwidth time-series data, delay time-series data and packet loss time-series data corresponding to each link in the power grid to analyze the correlation between the bandwidth, delay and packet loss indicators corresponding to each link and obtain the power grid status correlation indicators corresponding to each link. By constructing the topology of the power grid communication network and evaluating the stability contribution of each link to the entire power grid communication network based on the topology, the stability contribution weight of each link to the entire power grid communication network can be obtained. Based on the stability contribution weight of each link to the entire power grid communication network, and using the analytic hierarchy process (AHP) to comprehensively evaluate the stability of each link, the corresponding basic stability index, power grid volatility index, and power grid state correlation index are divided into three levels: basic stability, volatility, and correlation. The stability level of each link in the power grid is calculated by combining the stability contribution weight of each link, so as to obtain the stability characteristics of the power grid communication links.

5. The intelligent monitoring method for power grid measurement data quality according to claim 4, characterized in that, The analysis process for the instantaneous bandwidth fluctuation rate corresponding to each link is as follows: By setting the corresponding sliding window size to 3-5 time points, and dividing the bandwidth time series data corresponding to each link in the power grid into bandwidth time series segments based on the sliding window size, the bandwidth time series window data segments corresponding to each link are obtained. By selecting any two adjacent bandwidth time-series window data segments for each link, the corresponding bandwidth fluctuation amplitude is calculated to obtain the bandwidth fluctuation amplitude under any two adjacent time-series windows for each link. Instantaneous volatility is calculated by statistically analyzing the bandwidth fluctuation amplitude between any two adjacent time windows of each link, so as to obtain the instantaneous volatility at the time point between any two adjacent time windows of each link, and thus obtain the instantaneous bandwidth volatility of each link.

6. The intelligent monitoring method for power grid measurement data quality according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Obtain seasonal, periodic, and real-time operating data corresponding to the power grid operation; Step S32: Analyze the power grid load level based on the seasonal and periodic operating conditions data corresponding to the power grid operation to obtain the power grid operating condition load level; Step S33: Calculate the power generation ratio of the real-time operating data corresponding to the power grid operation to obtain the power generation ratio under the power grid operation conditions; Step S34: Perform dynamic threshold statistics on grid quality based on the load level of grid operation and the power generation ratio of grid operation to obtain the dynamic threshold of grid operation quality.

7. The intelligent monitoring method for power grid measurement data quality according to claim 6, characterized in that, Step S32 includes the following steps: The power load spectrum is converted from the seasonal and periodic operating conditions data corresponding to the power grid operation to generate the power grid load sine wave signal spectrum corresponding to the power grid operation load under seasonal and periodic conditions; By performing statistical analysis on the load variation spectrum of the power grid load sinusoidal signal under seasonal and periodic conditions, the seasonal and periodic load variation trends of the power grid operating conditions can be obtained. Load level prediction analysis is conducted based on the seasonal and periodic load change trends of the power grid operating conditions to obtain the load level of the power grid operating conditions.

8. The intelligent monitoring method for power grid measurement data quality according to claim 6, characterized in that, Step S33 includes the following steps: The power generation load distribution of different power generation units in the real-time operating data corresponding to the power grid operation is analyzed to obtain the power generation load distribution of different power generation units under the power grid operation conditions. Power transmission mining and analysis are performed on the real-time operating data of the power grid to obtain the power transmission relationship between different power generation units under the power grid operating conditions. Based on the power transmission relationship between different power generation units under different grid operating conditions, the power dispatch attenuation assessment is carried out between different power generation units in the real-time operating condition data corresponding to the grid operation, so as to obtain the power dispatch attenuation coefficient between different power generation units under different grid operating conditions. Based on the power generation dispatch attenuation coefficient between different power generation units under different power grid operating conditions, the power generation ratio of the power generation load distribution corresponding to different power generation units under different power grid operating conditions is calculated to obtain the power generation ratio under the power grid operating conditions.

9. The intelligent monitoring method for power grid measurement data quality according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Based on the power grid measurement data quality level score and combined with the power grid operation quality dynamic threshold, perform measurement anomaly detection on various types of measurement data in the power grid. When the power grid measurement data quality level score is between 0 and 60 points or the corresponding measurement data in the power grid exceeds the power grid operation quality dynamic threshold, the corresponding measurement data is judged as data anomaly until all types of measurement data are detected to obtain the power grid measurement quality anomaly data result. Step S42: Perform anomaly classification processing on each abnormal measurement data in the power grid measurement quality anomaly data results to obtain various types of power grid abnormal measurement data; Step S43: Perform power grid anomaly source analysis on various types of power grid anomaly measurement data to obtain the causes of anomalies corresponding to various types of anomalies, including anomalies at individual data points as well as large-scale data anomalies caused by equipment failures, communication interruptions and power grid failures. Step S44: Intelligent early warning monitoring is performed based on the causes of anomalies in various types of abnormal measurement data. If an anomaly is found in an individual data point, a yellow warning is issued; if a large-scale data anomaly is caused by equipment failure, communication interruption, or power grid failure, a red warning is issued. Corresponding power grid measurement data anomaly warning information is generated and pushed to the relevant power grid operation and maintenance management cloud platform via SMS, email, and in-station messages to execute corresponding power grid safety operation decision support work, including power grid operation mode adjustment plans, equipment maintenance plans, and communication link optimization plans.

Images