Transformer substation current transformer transformation ratio on-site checking method

Abstract

The invention relates to the technical field of power system detection, and discloses a transformer substation current transformer transformation ratio field checking method and device and a distributed system. The method comprises the following steps: acquiring operation data of the current transformer through various high-precision sensors, and training and predicting by using a deep learning RNN algorithm processing unit after data storage, preprocessing and feature extraction. The data acquisition unit comprises a plurality of sensors, the data preprocessing unit processes data by using a moving average filtering algorithm and a linear interpolation method, and the feature extraction unit extracts features by using an ARMA model and a weighted summation method. The device comprises a plurality of processing units which are connected through a data bus and are arranged in a cabinet with heat dissipation and electromagnetic shielding performance. The distributed system comprises a plurality of checking devices, and is connected with the central server through a local area network, thereby achieving the comprehensive monitoring of the transformation ratio of the current transformer of the transformer substation. The method improves the verification accuracy and real-time performance, fully considers the influence of environmental factors, and is efficient, convenient and high in reliability.

Description

Technical Field

[0001] The present invention relates to the technical field of power system detection, and in particular to the technical field of on-site detection and verification of current transformer ratio of a substation. Background Art

[0002] Prior art In the field of on-site verification of the current transformer ratio in substations, traditional verification methods mainly rely on manual operation and simple electrical testing equipment. Usually, technicians use a standard current source to inject a known current into the current transformer, then measure the secondary output current, and determine the ratio by calculating the ratio of the two. In this process, multimeters, oscilloscopes and other equipment are also used to monitor parameters such as voltage and current waveforms to assist in judging the performance of the transformer. For example, in the daily maintenance of some small substations, staff will regularly carry these devices to the site to verify each transformer one by one.

[0003] Technology Development

[0004] With the continuous improvement of the automation and intelligence of power systems, the accuracy and real-time requirements for substation equipment monitoring are also increasing. On the one hand, hardware equipment is constantly upgraded, and high-precision sensors such as manganese copper shunts, platinum resistance thermometers, capacitive humidity sensors, and electromagnetic interference detection equipment are gradually used in the monitoring of current transformers, which can more accurately collect relevant data such as current, temperature, humidity, and electromagnetic interference. On the other hand, algorithms and data analysis technologies are also constantly developing, from simple numerical calculations to intelligent analysis based on machine learning and deep learning. For example, some studies have begun to try to introduce neural network algorithms into the fault diagnosis and ratio verification of transformers, and improve the accuracy of judging the operating status of transformers by learning a large amount of historical data.

[0005] Current situation, pain points and difficulties

[0006] Large manual operation errors: Traditional methods rely on manual operation, and differences in the operation level and experience of different technicians may lead to inaccurate measurement results. Moreover, manual operation is cumbersome and time-consuming, and inefficient when many transformers need to be checked in large-scale substations.

[0007] Poor real-time performance: The traditional periodic inspection method cannot monitor the transformer ratio in real time. If the transformer ratio is abnormal between two inspections, it is difficult to detect in time, which may affect the stable operation of the power system.

[0008] Insufficient assessment of environmental factors: Although existing technologies can measure environmental parameters, they often do not fully consider the comprehensive impact of environmental factors such as temperature, humidity, and electromagnetic interference on the measurement results when analyzing the transformer ratio. Environmental factors may interfere with the measurement signal and cause deviations in the measurement results, but there is currently a lack of effective methods to accurately evaluate and correct these impacts.

[0009] Single data analysis method: Previous data analysis was mostly based on simple mathematical calculations and empirical judgments, and had limited ability to handle complex nonlinear relationships and multi-factor coupling problems. For example, the transformer ratio may be affected by multiple factors, and traditional methods make it difficult to establish an accurate model to describe these complex relationships, thus affecting the accurate judgment of the transformer ratio. Summary of the invention

[0010] The invention provides a method for on-site verification of a current transformer ratio in a transformer substation, a verification device using the method, and a distributed system including the verification device.

[0011] 1. Verification Method

[0012] 1.1 Component units: including data acquisition unit, data storage unit, data preprocessing unit, feature extraction unit, and deep learning RNN algorithm processing unit. The data acquisition unit collects data and transmits it to the storage unit; the storage unit acts as a data transfer and storage hub; the preprocessing unit processes the data of the storage unit to remove abnormal values ​​and missing values; the feature extraction unit extracts features from the preprocessed data; the deep learning RNN algorithm processing unit uses the algorithm to train and learn the feature data.

[0013] 1.2 Algorithm steps

[0014] Step S01: a data acquisition module acquires data and transmits it to a data storage unit.

[0015] Step S02: the data preprocessing unit cleans the data of the data storage unit.

[0016] Step S03: The feature extraction unit extracts features from the cleaned data and transmits them to the deep learning RNN algorithm processing unit.

[0017] Step S04, the deep learning RNN algorithm processing unit performs training and prediction according to a specific formula, and transmits the prediction result to the comparison and judgment unit (in the formula is the hidden state at the current moment, and are the weight matrices from input to hidden layer and from hidden layer to hidden layer, and is the bias vector, They are the weight matrices corresponding to ambient temperature, humidity, and electromagnetic interference intensity, They are the current ambient temperature, humidity, and electromagnetic interference intensity values. is the input feature at the current moment).

[0018] 3. Other related content

[0019] 3.1. The data acquisition unit includes a manganese copper shunt, a platinum resistance thermometer, a capacitive humidity sensor, a loop antenna and a spectrum analyzer combination structure, which are used for current sampling, temperature measurement, humidity measurement and electromagnetic interference detection respectively.

[0020] 3.2 The data preprocessing unit uses a sliding average filtering algorithm to remove outliers and a linear interpolation method to fill in missing data.

[0021] 3.3 In the feature extraction unit, the time series feature extraction adopts the autoregressive moving average model (ARMA), and the environmental feature extraction is the weighted sum of the ambient temperature, humidity and electromagnetic interference intensity (the weighting coefficients are 0.3, 0.2, and 0.5 respectively).

[0022] 3.4 The method also includes pre-initialization items (such as parameter setting and equipment calibration of data acquisition unit and storage unit), intermediate inspection items (real-time monitoring of data quality and feature extraction effectiveness), and post-feedback verification items (results are fed back to the monitoring center, and abnormal situations are secondary checked and recorded).

[0023] 2. Verification device: including the above units, each unit is connected through a data bus, the data acquisition unit collects data and then processes it in turn through other units, and finally the feedback unit outputs the result. The device is installed in a cabinet with good heat dissipation and electromagnetic shielding performance, and the internal unit is modularly designed for easy installation, maintenance and upgrading.

[0024] 3. Distributed system: It includes multiple verification devices distributed in different locations of the substation, connected to the central server through a local area network. The central server summarizes, analyzes and uniformly manages the data of each device to achieve comprehensive monitoring of the current transformer ratio of the substation. Each verification device has a unique network address identifier. The central server can remotely configure parameters and query data, and has a fault warning function.

[0025] Principle Analysis

[0026] 1. Data collection: Use a variety of high-precision sensors (manganese copper shunts, platinum resistance thermometers, capacitive humidity sensors, electromagnetic interference detection equipment) to accurately collect data such as current, ambient temperature, humidity, and electromagnetic interference intensity during the operation of the current transformer, providing a basic data source for subsequent accurate analysis.

[0027] 2. Data processing flow

[0028] 1. The data preprocessing unit first removes outliers through a sliding average filtering algorithm to ensure that the data is relatively accurate, and then uses linear interpolation to fill in missing data to ensure data continuity and make the data more suitable for subsequent analysis.

[0029] 2. The feature extraction unit uses the autoregressive moving average model (ARMA) to extract time series features to reflect the time-varying law of the operating status of the current transformer. At the same time, the environmental factors are weighted and summed to obtain the comprehensive environmental characteristic value, and these feature data are input into the deep learning RNN algorithm processing unit.

[0030] 3. Deep learning RNN algorithm processing: The algorithm is trained and predicted based on a formula that includes multiple influencing factors (such as environmental factors, input features, etc.), and can learn the complex relationships in the data to accurately judge the current transformer ratio. In the formula, parameters such as different weight matrices and bias vectors are continuously adjusted during the training process to adapt to the data characteristics under different working conditions, and the final output prediction results are used for subsequent comparison and judgment.

[0031] Beneficial Effects

[0032] 1. Improve accuracy

[0033] 1.1 High-precision sensors collect data more accurately, reducing judgment errors caused by inaccurate data.

[0034] 1.2 The deep learning RNN algorithm processing unit can handle complex relationships and is more accurate in judging the change ratio than traditional simple calculation methods.

[0035] 1.3. The feature extraction unit extracts effective features, which helps to improve the accuracy of model prediction and thus improve the accuracy of ratio verification.

[0036] 2. Enhanced real-time performance: It can collect, process and analyze data in real time, and promptly detect abnormal current transformer ratios, thus avoiding the impact on the stable operation of the power system caused by the failure to detect abnormalities during regular inspection intervals.

[0037] 3. Fully consider the impact of environmental factors: By collecting ambient temperature, humidity and electromagnetic interference intensity data and setting the corresponding weight matrix in the algorithm, environmental factors are incorporated into the analysis process, which can accurately evaluate their comprehensive impact on the measurement results, thereby correcting the measurement deviation caused by environmental factors.

[0038] 4. Efficient and convenient

[0039] 4.1 Automated data collection, processing and analysis processes reduce manual operations, reduce manual operation errors, and improve verification efficiency. It is especially suitable for the verification of numerous transformers in large-scale substations.

[0040] 4.2 In the distributed system, the central server can remotely manage and configure each verification device, making it easier for operation and maintenance personnel to operate and improving management efficiency.

[0041] 5. High reliability

[0042] 5.1 Check the good heat dissipation and electromagnetic shielding performance and modular design of the device cabinet to ensure stable operation of the equipment, facilitate maintenance and upgrades, and improve the overall reliability of the equipment.

[0043] 5.2. The fault warning function of the distributed system can promptly detect verification device failures or data anomalies, facilitating timely processing by operation and maintenance personnel to ensure the continuous and effective operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 :Flowchart of on-site verification method of current transformer ratio in substation; Figure 2 : Interactive diagram of the on-site verification method of the current transformer ratio in the substation; Figure 3 : Component diagram of the method for on-site verification of current transformer ratio in substation;

[0046] Figure 4 : Schematic diagram of data acquisition unit object;

[0047] Figure 5 : Schematic diagram of data sequence processing by data preprocessing unit;

[0048] Figure 6 : Schematic diagram of feature extraction unit processing data sequence;

[0049] Figure 7 : Attached project diagram;

[0050] Figure 8 : Interactive diagram of data acquisition unit parameter settings;

[0051] Fig. 9 : Pre-initialization project interaction diagram;

[0052] Fig.10 :Intermediate inspection project flow chart;

[0053] Fig.11 : Schematic diagram of post-feedback verification project;

[0054] Fig.12 : Verification device component diagram;

[0055] Fig.13 : Distributed system deployment diagram. DETAILED DESCRIPTION

[0056] Embodiment 1, claim 1 combined Figure 1 , Figure 2 , Figure 3 , detailed explanation is as follows;

[0057] 1. Overall structure of the verification method

[0058] 1.1 System components

[0059] 1.1.1 Data acquisition unit

[0060] The data acquisition unit is the basis of the entire verification method and is responsible for obtaining various key data. It includes a variety of sensors such as manganese copper shunts, platinum resistance thermometers, capacitive humidity sensors, and a combination structure of loop antennas and spectrum analyzers. The manganese copper shunt is used to accurately collect the current data of the current transformer, providing the core basis for the subsequent ratio calculation. The platinum resistance thermometer can accurately measure the ambient temperature, and the capacitive humidity sensor can sense the ambient humidity in real time. The combination structure of the loop antenna and spectrum analyzer is used to accurately detect the frequency and intensity of electromagnetic interference in the environment. For example, in a large substation, when the current transformer is running, the manganese copper shunt continuously monitors the current size. During a peak load period, the primary side current of the current transformer is accurately collected as 500A. At the same time, the platinum resistance thermometer shows that the ambient temperature is 35°C, the capacitive humidity sensor measures the humidity as 60%RH, and the electromagnetic interference sensor detects the presence of electromagnetic interference of a certain frequency and intensity in the surrounding area. These data are collected synchronously, laying the foundation for a comprehensive analysis of the operating status of the current transformer.

[0061] 1.1.2 Data Storage Unit

[0062] The data storage unit plays a key role as a data transfer and storage hub. It receives various types of data from the data acquisition unit and temporarily stores them. In the case of the above-mentioned large substation, the collected data such as current, temperature, humidity and electromagnetic interference intensity will first be transmitted to the data storage unit. This unit is like a data cache station, ensuring the continuity of data throughout the entire verification process, providing a stable data source for subsequent data preprocessing, feature extraction and other links, and avoiding data loss or interruption affecting the verification results.

[0063] 1.1.3 Data Preprocessing Unit

[0064] The data preprocessing unit mainly processes the data in the data storage unit. First, it uses a sliding average filtering algorithm to remove abnormal values. Taking current data as an example, its filtering formula is (in is the filtered value, is the original data, Assuming the sliding window size (m = 5), the current data collected is [498, 502, 450, 505, 501] (450 may be an outlier). After the sliding average filter calculation, a smoother and more accurate current value sequence can be obtained. In addition, the unit also has a data filling function. If a small amount of missing data occurs during the data collection process, such as missing humidity data at a certain moment, it will use linear interpolation to fill it in to ensure the continuity of the data and make the data more suitable for subsequent analysis and processing.

[0065] 1.1.4 Feature Extraction Unit

[0066] The feature extraction unit extracts key features from the preprocessed data. In terms of time series feature extraction, the autoregressive moving average model (ARMA) is used. For example, by analyzing the autocorrelation and moving average of historical current data, features that can reflect the law of change of the operating status of the current transformer over time are extracted. For environmental feature extraction, the weighted sum of ambient temperature, humidity and electromagnetic interference intensity is taken as the comprehensive environmental feature value, where the weighting coefficients of ambient temperature, humidity and electromagnetic interference intensity are 0.3, 0.2 and 0.5 respectively. In the above substation case, the calculated comprehensive environmental feature value is: (0.3×35 + 0.2×60 + 0.5×electromagnetic interference intensity value). These extracted features can more effectively represent the key information of the data and provide more valuable input for the deep learning algorithm.

[0067] 1.1.5 Deep Learning RNN Algorithm Processing Unit

[0068] The deep learning RNN algorithm processing unit uses the deep learning RNN algorithm to train and learn the feature data. It is based on the formula (in is the hidden state at the current moment, and are the weight matrices from input to hidden layer and from hidden layer to hidden layer, and is the bias vector, They are the weight matrices corresponding to ambient temperature, humidity, and electromagnetic interference intensity, They are the current ambient temperature, humidity, and electromagnetic interference intensity values. is the input feature at the current moment) for training and prediction. During the training process, multiple sets of collected data are used as samples, for example, 1000 sets of current transformer data (including current, environmental data, etc.) at different moments are collected.

[0069] 1.2 RNN training process example

[0070] First, randomly initialize the weight matrix and the bias vector . Then, for each set of data:

[0071] The input features at the current moment (such as current value, environmental characteristic value, etc.), the hidden state of the previous moment (The initial moment can be set to all zero vectors) and the current moment's environmental parameters Substitute the formula to calculate the hidden state at the current moment .

[0072] According to the hidden state Calculate predicted outputs (e.g. predict the transformation ratio of a current transformer).

[0073] Calculate the error between the predicted output and the actual output (the exact ratio is known), such as the mean square error (MSE).

[0074] Using the back-propagation algorithm, the gradients of the weight matrix and bias vector are calculated based on the error.

[0075] The weight matrix and bias vector are updated according to the gradient. The update formula can use stochastic gradient descent (SGD) or its variants, such as Adagrad, Adadelta, etc. For example, the weight update formula is (in is the learning rate).

[0076] Repeat the above steps and perform multiple iterations of training on all sample data until the error converges to a range that meets the requirements. After a large number of iterations of training, the RNN algorithm can learn the complex relationships in the data and accurately determine the transformation ratio of the current transformer.

[0077] 1.3 Algorithm steps and system interaction

[0078] After the data acquisition module collects data, it is transmitted to the data storage unit according to the process (step S01), and then the data preprocessing unit cleans the data (step S02), the feature extraction unit extracts features and transmits them to the deep learning RNN algorithm processing unit (step S03), and finally the deep learning RNN algorithm processing unit performs training and prediction, and transmits the prediction results to the comparison and judgment unit (step S04). Each unit is connected through a data bus to achieve efficient data interaction and ensure the smooth progress of the entire verification process. In the actual operation of the substation, this process is continuously circulated, real-time monitoring of the operating status of the current transformer, timely detection of abnormal ratio conditions, and ensuring the stable operation of the power system.

[0079] 2. Verify device structure and function

[0080] The verification device includes the above-mentioned units, which are closely connected through the data bus. After the data acquisition unit collects data, it is processed by other units in turn, and finally the feedback unit outputs the results. The entire device is installed in a cabinet, which has good heat dissipation and electromagnetic shielding performance. For example, at the substation site, the cabinet can effectively block the influence of external electromagnetic interference on the internal device, and the heat dissipation system ensures that the device will not malfunction due to overheating during long-term operation. The internal units adopt a modular design to facilitate installation, maintenance and upgrading. When a unit needs to be maintained or upgraded, the corresponding module can be directly replaced without large-scale disassembly and reassembly of the entire device, which greatly improves work efficiency and reduces operation and maintenance costs.

[0081] 3. Distributed system architecture and advantages

[0082] The distributed system consists of multiple verification devices distributed in different locations of the substation, which are connected to the central server through a local area network. The central server summarizes and analyzes the data of each device and manages it in a unified manner to achieve comprehensive monitoring of the current transformer ratio of the entire substation. Each verification device has a unique network address identifier in the distributed system, and the central server can perform remote parameter configuration and data query operations on a single or multiple verification devices according to demand. For example, if the operation and maintenance personnel find that the current transformer data in a certain area has an abnormal trend in the monitoring center, they can remotely adjust the data acquisition parameters of the verification device in the area through the central server to further analyze the data in detail. At the same time, the distributed system has a fault warning function. When a verification device fails or the data is abnormal, the central server will send an alarm message to the operation and maintenance personnel in a timely manner. If the communication module of a verification device fails, the central server will immediately issue an alarm, and the operation and maintenance personnel can quickly go to handle it to ensure the continuous and effective operation of the system and the safe and stable power supply of the substation. This distributed system architecture improves the intelligence level and operation and maintenance efficiency of substation equipment management.

[0083] Embodiment 2, claim 2 combined Figure 4 The details are as follows;

[0084] In the on-site verification of the current transformer ratio of the substation, the data acquisition unit is the primary link of the entire process and its role is crucial. The unit consists of a manganese copper shunt, a platinum resistance thermometer, a capacitive humidity sensor, and a combination structure of a loop antenna and a spectrum analyzer. Based on Ohm's law, the manganese copper shunt measures the voltage drop across itself to accurately convert and obtain the small current signal corresponding to the large current on the primary side of the current transformer, thereby accurately sensing the current size; the platinum resistance thermometer uses the characteristics of platinum metal resistance changing with temperature. In the complex environment of the substation, whether it is hot summer or cold winter, it can accurately measure the ambient temperature around the current transformer, providing reliable data for evaluating the impact of temperature on the ratio; the capacitive humidity sensor uses the principle that the dielectric constant of the medium between the capacitor plates changes with humidity to change the capacitance value, and monitors the humidity in the humid environment of the substation in real time, assisting in a comprehensive analysis of the impact of environmental factors on the current transformer; the loop antenna of the electromagnetic interference sensor receives electromagnetic signals, and the spectrum analyzer analyzes the signal frequency and intensity. In the complex electromagnetic environment of the substation, it can timely detect signals such as transient electromagnetic interference generated by the closing or switching operation of large transformers.

[0085] The collected data such as current, temperature, humidity and electromagnetic interference intensity are transmitted to the data storage unit via the data bus, laying a comprehensive and solid information foundation for subsequent data preprocessing, feature extraction and deep learning RNN algorithm processing. If the data acquisition unit fails or the data is inaccurate, the subsequent analysis and judgment will lose reliable basis, which may lead to errors in the evaluation of the current transformer ratio. Therefore, the normal operation of the data acquisition unit and high-precision data acquisition are the key to ensuring the accuracy and effectiveness of the verification work. It works in coordination with other units to jointly ensure the stable operation and accurate measurement of current transformers in the power system.

[0086] Embodiment 3, claim 3 combined Figure 5 The details are as follows;

[0087] Data preprocessing unit for on-site verification of current transformer ratio in substations

[0088] 1. Importance of data preprocessing unit

[0089] In the on-site verification system of the current transformer ratio in the substation, the data preprocessing unit plays an indispensable role. It is a bridge connecting the data acquisition unit and the subsequent analysis and processing unit, responsible for the preliminary processing and optimization of the collected data to improve the data quality and ensure the accuracy and reliability of the entire verification work. Its processing effect directly affects the effectiveness of subsequent feature extraction and algorithm analysis, and is a key link to ensure the smooth progress of the entire verification process.

[0090] 1.1 Sliding average filter algorithm to remove outliers

[0091] 1.1.1 Algorithm Principle and Application

[0092] The data preprocessing unit uses a sliding average filtering algorithm to remove outliers. The core principle of this algorithm is to smooth the original data by calculating the average value of the data in a certain sliding window, thereby effectively eliminating interference factors in the data. Its filtering formula is: (in is the filtered value, is the original data, is the sliding window size). In the processing of substation current transformer operation data, for example, the collected current data sequence may be subject to various short-term interferences and individual abnormal values ​​may appear. Assuming that a set of current data collected is [498, 502, 450, 505,501] (where 450 may be an abnormal value), if the sliding window size is set (m = 5), according to the formula, first calculate the average value of the data in the first window ((498 + 502 + 450 + 505 + 501) / 5 = 491.2), this average value will be used as the value of the first data point after filtering, and then the window will slide backward in sequence, and the average value will be continuously updated to obtain a smoother and more accurate current value sequence. In this way, the sliding average filtering algorithm can effectively reduce the impact of abnormal values ​​on the data, make the data more stable and reliable, and provide a more accurate basis for subsequent analysis.

[0093] 1.1.2 Linear interpolation to fill missing data

[0094] 1.1.2.1 Method and necessity

[0095] During the data collection process, a small amount of missing data may occur due to various reasons. In this case, the data preprocessing unit uses linear interpolation to fill in the missing data. Linear interpolation estimates the missing values ​​based on the linear relationship between data points. For example, during the temperature data collection process, if the temperature data at a certain moment is missing, assuming that the temperature values ​​at two adjacent moments are and , the corresponding time point is and , and the time point of missing data is , the missing temperature values ​​are calculated by linear interpolation Ensuring data continuity is crucial for subsequent feature extraction and algorithm processing, because many analysis methods rely on the integrity and continuity of data, and missing data may lead to inaccurate feature extraction or algorithm failure to operate normally.

[0096] 1.1.2.2 Impact on overall data

[0097] By filling the missing data with linear interpolation, the data sequence remains continuous in time, avoiding information interruption caused by missing data. This helps subsequent operations such as time series feature extraction to accurately reflect the changing pattern of the operating status of the current transformer over time. In Figure 5, it can be seen that the data processed by the data preprocessing unit, whether it is the smoothed data after the sliding average filter or the complete sequence after filling the missing data, is more suitable for further feature extraction and deep learning RNN algorithm processing. After these operations of the data preprocessing unit, the data quality is significantly improved, providing a strong guarantee for the accuracy and reliability of the on-site verification of the current transformer ratio of the entire substation. It provides more stable, continuous and accurate data for subsequent units, allowing the entire verification system to operate more efficiently, accurately judge the ratio of the current transformer, discover potential problems in a timely manner, and ensure the safe and stable operation of the power system.

[0098] Embodiment 4, claim 4 combined Figure 6 The details are as follows;

[0099] Feature extraction unit for on-site verification of current transformer ratio in substations

[0100] 1. The core role of the feature extraction unit

[0101] In the on-site verification system of the current transformer ratio in the substation, the feature extraction unit is in a key position. It is like a data refinery, responsible for mining the most representative and valuable information from the pre-processed data, converting the original data into a feature vector that can deeply reflect the operating status of the current transformer, and providing high-quality input for the subsequent deep learning RNN algorithm processing, thereby greatly improving the accuracy and reliability of the entire verification system in judging the current transformer ratio.

[0102] 1.1 Time Series Feature Extraction (Using Autoregressive Moving Average Model - ARMA)

[0103] 1.1.1 ARMA model principle and application

[0104] The time series feature extraction in the feature extraction unit adopts the autoregressive moving average model (ARMA). The ARMA model is constructed based on the autocorrelation and moving average of historical data. For the operation data of the current transformer, such as the sequence of changes in the current value over time, it has certain inherent laws. The ARMA model extracts features that can characterize the law of changes in the operation status of the current transformer over time by analyzing the impact of the data at the past moment on the data at the current moment (autoregressive part) and the impact of the random error at the past moment on the data at the current moment (moving average part). In practical applications, suppose we have a series of data points of the current value of the current transformer changing over time. The ARMA model calculates the corresponding coefficients according to the set autoregressive order (p) and moving average order (q). For example, when p = 2 and q = 1, the model predicts the current value at the current moment based on the current values ​​at the previous two moments and the random error at the previous moment, and minimizes the error between the predicted value and the actual value by continuously adjusting the coefficients. In this way, the ARMA model can extract dynamic features hidden in the time series data, which are essential for accurately grasping the operation trend and state changes of the current transformer.

[0105] 1.1.2 Significance of current transformer condition assessment

[0106] The time series features extracted by the ARMA model can help us detect abnormal changes in the operation of the current transformer in a timely manner. For example, if the core of the current transformer is magnetically saturated, the change pattern of its current value may deviate from the normal mode. The ARMA feature can keenly capture this change and provide a strong basis for timely warning and fault diagnosis. In Figure 6, it can be seen that the time series data processed by the ARMA model is converted into a set of representative eigenvalues. These eigenvalues ​​can more intuitively reflect the change trend of the operating status of the current transformer and provide key information for subsequent analysis and processing.

[0107] 1.1.3 Environmental feature extraction (weighted summation)

[0108] 1.1.3.1 Weighted summation method

[0109] In terms of environmental feature extraction, the weighted sum of ambient temperature, humidity and electromagnetic interference intensity is used as the comprehensive environmental feature value. The weighted coefficients of ambient temperature, humidity and electromagnetic interference intensity are 0.3, 0.2 and 0.5 respectively. This weighted summation method is based on the in-depth study and analysis of the influence of different environmental factors on the current transformer. For example, the electromagnetic interference intensity may have a greater impact on the measurement accuracy of the current transformer, so a higher weight of 0.5 is given; the temperature and humidity have a relatively smaller impact, but it cannot be ignored, and weights of 0.3 and 0.2 are given respectively. For the environmental data at a certain moment, assuming that the temperature is 30°C, the humidity is 50%RH, and the electromagnetic interference intensity is a certain value, the comprehensive environmental feature value is calculated by weighted summation = 0.3×30 + 0.2×50 + 0.5×electromagnetic interference intensity value.

[0110] 1.1.3.2 Necessity of considering environmental factors

[0111] Considering environmental factors is crucial for accurately evaluating the current transformer ratio. Changes in ambient temperature may cause changes in the resistance of the current transformer winding, which in turn affects the ratio; changes in humidity may affect the insulation performance and indirectly affect the measurement accuracy; and electromagnetic interference directly interferes with the measurement signal. The comprehensive environmental characteristic value obtained by weighted summation can quantify the combined impact of these environmental factors, so that subsequent algorithm processing can more accurately correct the measurement deviation caused by environmental factors, thereby improving the accuracy of current transformer ratio verification. The feature extraction unit effectively extracts time series features and comprehensive environmental features by reasonably using the ARMA model and weighted summation method, providing rich and accurate feature information for the entire verification system, further enhancing the system's ability to judge the current transformer ratio and ensuring the reliable operation of current transformers in power systems.

[0112] Embodiment 5, claim 5 combines Figure 7 The details are as follows;

[0113] Full process guarantee for on-site verification of current transformer ratio in substations

[0114] 1. Preliminary preparation lays the foundation

[0115] Before the on-site verification of the current transformer ratio of the substation is started, the pre-initialization project is the key to ensure accurate measurement. The parameter settings of the data acquisition unit and the data storage unit must be carefully planned, the sampling frequency must be flexibly set according to the operating frequency of the current transformer, the range can be switched manually or automatically, and the industrial Ethernet protocol is used to ensure high-speed and stable data transmission. At the same time, the equipment calibration link is indispensable. The standard current source calibrates the smart meter to an accuracy of 0.1%, and the standard temperature and humidity chamber and the electromagnetic shielding room calibrate the environmental sensors respectively to provide an accurate benchmark for subsequent measurements. This series of pre-operations is like the cornerstone of a building, laying a solid foundation for the entire verification work.

[0116] 2. Intermediate monitoring to ensure quality

[0117] During data preprocessing and feature extraction, the intermediate inspection items monitor data quality and feature validity in real time. The degree of data dispersion is monitored by calculating the variance and standard deviation. Once the preset threshold is exceeded, the parameters of the data collection and preprocessing links are immediately checked. At the same time, the model prediction accuracy is compared before and after feature extraction. If the improvement is less than 5%, the feature extraction unit parameters or algorithms are optimized. This process is like a quality control, ensuring that the data is reliable and the features are valid, which will escort the subsequent accurate judgment of the current transformer ratio.

[0118] 3. Post-verification and strengthening of results

[0119] After determining the transformation ratio result, the post-feedback verification project is quickly launched. When the result is fed back to the monitoring center, the wireless communication module automatically detects the switching signal to ensure stable transmission. For abnormal situations, the high-precision handheld calibrator performs a secondary verification with higher measurement accuracy, records abnormal information in detail and stores backup. This not only provides accurate data, but also provides a basis for operation and maintenance personnel to trace back and analyze, maintain and improve equipment. It is an important closed loop of the entire verification process, which effectively guarantees the stable and reliable operation of the substation current transformer.

[0120] Embodiment 6, claim 6 combined Figure 8 The details are as follows;

[0121] Key pre-processing steps for on-site verification of current transformer ratio in substations

[0122] 1. Accurate parameter settings ensure data collection

[0123] In the pre-initialization stage of on-site verification of the current transformer ratio of the substation, the accuracy of parameter settings is of great significance to data collection. The sampling frequency needs to be flexibly adjusted according to the operating frequency of the current transformer. For example, when the operating frequency is 50Hz, it is often set to 200Hz or higher to obtain enough data points to depict the current waveform to prevent data distortion caused by low frequency from affecting subsequent analysis. The range can be switched manually or automatically. The appropriate fixed value can be manually set under stable working conditions. The range is automatically switched under complex and variable loads to avoid data overflow and ensure that the data is complete and accurate. The industrial Ethernet protocol is used to ensure high-speed and stable data transmission, and effectively resist interference in complex electromagnetic environments. Its high transmission rate and strong real-time performance ensure timely and accurate data transmission, laying the foundation for the smooth operation of the verification system, so that the collected data can truly reflect the operating status of the current transformer.

[0124] 2. Strict equipment calibration ensures measurement accuracy

[0125] Equipment calibration is also indispensable. Use a standard current source to calibrate the smart meter to an accuracy of 0.1%. By outputting a precisely known current and comparing the measured value with the standard value, the internal parameters are adjusted to provide a reliable data basis for the ratio calculation. For environmental sensors, a standard temperature and humidity chamber (temperature and humidity control accuracy of ±0.1°C and ±1%RH respectively) is used to calibrate the temperature and humidity sensors. The output value is compared with the standard value at different set temperatures and humidities to adjust the parameters. An electromagnetic shielding room (shielding effectiveness of more than 80dB) is used to calibrate the electromagnetic interference sensor. A known interference signal is applied to compare the measurement with the standard value to adjust the parameters. This ensures that the environmental sensor can measure accurately and provides accurate data support for analyzing the impact of environmental factors on the ratio, thereby improving the overall accuracy and reliability of the verification system.

[0126] Embodiment 7, claim 7 combined Fig. 9 The details are as follows;

[0127] Key points for controlling the current transformer ratio on-site inspection in substations

[0128] 1. Data quality monitoring ensures data reliability

[0129] Data quality monitoring plays a vital role in the on-site verification of the current transformer ratio in substations. The degree of data dispersion is measured by calculating the variance and standard deviation, which is based on the difference between the data point and the average value. If the variance is greater than the preset threshold or the standard deviation exceeds the set range, it indicates that the data quality is worrying, which may be caused by sensor failure or electromagnetic interference. This requires rechecking the data acquisition unit parameters, such as sampling frequency, sensor connection, and data preprocessing unit filtering algorithm parameters, to ensure that the data is stable and reliable, providing a solid foundation for subsequent analysis.

[0130] 2. Feature extraction and evaluation ensure accurate judgment

[0131] The effectiveness evaluation of feature extraction is also critical. Based on the comparison of model prediction accuracy before and after feature extraction, if the accuracy is significantly improved, it proves that feature extraction is effective; conversely, if the improvement is less than 5%, the feature extraction unit parameters or algorithms need to be adjusted. For the ARMA model of time series features, adjust the autoregressive and moving average orders to optimize feature extraction, and reasonably adjust the environmental feature weighting coefficients to make the features more realistic, thereby improving the accuracy of the verification system's judgment on the transformation ratio and ensuring the stable operation of the current transformer.

[0132] Embodiment 8, claim 8 combined Fig.10 The details are as follows;

[0133] Post-security measures for on-site verification of current transformer ratio in substations

[0134] 1. Result feedback communication ensures data transmission

[0135] In the post-feedback verification process of the current transformer ratio on-site verification of the substation, the communication guarantee of the result feedback is extremely critical. The wireless communication module relies on the automatic detection and switching function of signal strength to ensure stable data transmission in a complex electromagnetic environment. It monitors the signal in real time, such as once a second. When the signal strength of the main communication network (such as Wi-Fi) is lower than the set threshold (such as -70dBm), it automatically switches to the backup communication network (such as 4G). Its intelligent switching path can be clearly seen from Figure 10, ensuring that the verification results are delivered to the monitoring center in time, so that the operation and maintenance personnel can respond quickly.

[0136] 2. Secondary verification record storage helps operation and maintenance

[0137] In order to determine the abnormal situation, the secondary verification uses a high-precision handheld current transformer calibrator (with an accuracy of 0.05% higher than the main measurement equipment) to measure again to obtain more accurate data to confirm the abnormality. At the same time, the abnormal time, data and secondary verification results are recorded in detail, first stored in the local database, and then backed up to the monitoring center server in real time or regularly to prevent data loss. This provides a strong basis for operation and maintenance personnel to trace and analyze abnormalities, maintain and improve equipment, and effectively ensure the long-term stable operation of current transformers.

[0138] Embodiment 9, claim 9 combined Fig.11 The details are as follows;

[0139] In-depth analysis of the on-site verification device for current transformer ratio in substations

[0140] 1. Overall architecture and hardware design

[0141] The on-site verification device for the current transformer ratio of the substation is composed of multiple units connected by a data bus. The data acquisition unit collects current, environment and other data and transmits them to the storage unit. The preprocessing unit cleans the data. The feature extraction unit obtains key features. After the deep learning RNN algorithm processing unit calculates, it is judged by the comparison and judgment unit, and finally the feedback unit outputs the result. The entire device is placed in a cabinet with good heat dissipation and electromagnetic shielding performance, and each unit is modularly designed. Heat dissipation prevents overheating failures, and shielding against electromagnetic interference ensures data accuracy. Modularity facilitates maintenance and upgrading. If the sensor is damaged, the module can be replaced separately. When optimizing the algorithm or expanding the function, only the corresponding module needs to be upgraded. According to Figure 11, it can be seen that the units are closely connected and cooperated, laying a solid foundation for verification.

[0142] 2. Implementation of unit coordination and verification functions

[0143] Each unit plays a key role in the verification and works closely together. The acquisition and storage unit provides and transfers data, which is the cornerstone of subsequent analysis. Inaccurate data or storage problems will lead to evaluation errors. The preprocessing unit improves data quality, the feature extraction unit provides effective input, the deep learning algorithm processing unit accurately analyzes the relationship, the comparison and judgment unit determines the result, and the feedback unit presents the status in a timely manner. Each unit performs its duties according to the process, and jointly achieves accurate and efficient verification of the transformation ratio to ensure the safe and stable operation of the substation. Figure 11 shows its collaborative workflow.

[0144] Embodiment 10, claim 10 combined Fig.12 , Fig.13 The details are as follows;

[0145] Distributed System Analysis of On-site Verification of Current Transformer Ratio in Substations

[0146] 1. System architecture and connections

[0147] The distributed system for on-site verification of the current transformer ratio of the substation consists of verification devices distributed in different locations and a central server. The verification devices are connected to the central server through a local area network. Each verification device contains multiple functional units working together, and has a unique network address identifier, which is convenient for remote operation of the central server. The operation and maintenance personnel can adjust parameters and query data as needed, reduce the number of on-site operation and maintenance, and improve the level of intelligent and refined management. The connection architecture can be clearly seen from Figure 12 to ensure efficient operation and accurate monitoring of the system.

[0148] 2. Functional advantages and guarantees

[0149] The central server aggregates and analyzes the data of the verification device, comprehensively monitors the operation of the current transformer, combines multi-source data to evaluate performance, and integrates the data view as shown in Figure 13 to help operation and maintenance personnel discover patterns and warn of faults. The system fault warning function is sensitive and can detect abnormalities in the verification device in time and notify the operation and maintenance personnel, reducing the risk of monitoring errors, enhancing the stability and reliability of the system, and effectively ensuring the safe operation of substation equipment and continuous and stable power supply.

Claims

1. A method for on-site verification of the current transformer ratio of a substation, comprising a data acquisition unit, a data storage unit, a data preprocessing unit, a feature extraction unit, and a deep learning RNN algorithm processing unit in: The data acquisition unit collects data and is connected to the data storage unit to provide data; The data storage unit stores the data collected by the data collection unit, is connected with the data collection unit and the data preprocessing unit, and serves as a data transfer and storage hub; The data preprocessing unit processes the data of the data storage unit to remove abnormal values ​​and missing values, which is the data transition processing link; The feature extraction unit is used to extract features from the preprocessed data, and connects the data preprocessing unit and the deep learning RNN algorithm processing unit; The function of the deep learning RNN algorithm processing unit is to use the deep learning RNN algorithm to train and learn the feature data; Also includes: Algorithm steps: Step S01, the data acquisition module collects data and transmits it to the data storage unit; Step S02, the data preprocessing unit cleans the data of the data storage unit; Step S03, the feature extraction unit extracts features from the cleaned data and transmits them to the deep learning RNN algorithm processing unit; Step S04, the deep learning RNN algorithm processing unit is based on The formula is used for training and prediction, and the prediction results are transmitted to the comparison and judgment unit. in is the hidden state at the current moment, and are the weight matrices from input to hidden layer and from hidden layer to hidden layer, and is the bias vector, , , They are the weight matrices corresponding to ambient temperature, humidity, and electromagnetic interference intensity, , , They are the current ambient temperature, humidity, and electromagnetic interference intensity values. is the input feature at the current moment.

2. The data acquisition unit according to claim 1, characterized in that: It also includes a manganese copper shunt for current sampling, a platinum resistance thermometer as the temperature sensor, a capacitive humidity sensor as the humidity sensor, and an electromagnetic interference sensor using a combination of a loop antenna and a spectrum analyzer to sense the frequency and intensity of electromagnetic interference in the environment.

3. The data preprocessing unit according to claim 1, characterized in that: The data preprocessing unit uses a sliding average filtering algorithm to remove outliers, and its filtering formula is: , in is the filtered value, is the original data, The unit also has a data filling function, which uses linear interpolation to fill in a small amount of missing data to ensure data continuity.

4. The feature extraction unit according to claim 1, characterized in that: It also includes that the time series feature extraction in the feature extraction unit adopts an autoregressive moving average model (ARMA), and the time series features are extracted by analyzing the autocorrelation and moving average of historical data to reflect the change pattern of the operating status of the current transformer over time. When extracting the environmental features, the ambient temperature, humidity and electromagnetic interference intensity are weighted and summed as the comprehensive environmental feature value, wherein the weighting coefficients of the ambient temperature, humidity and electromagnetic interference intensity are (0.3), (0.2) and (0.5) respectively.

5. The on-site verification method for the current transformer ratio of a substation according to claims 1 to 4, characterized in that: The method further comprises, Pre-initialization project: Before data collection, the data acquisition unit and data storage unit are parameterized and calibrated. The parameter setting includes setting the sampling frequency, range, and data transmission protocol. The equipment is calibrated using a standard current source to calibrate the smart meter with a calibration accuracy of (0.1%). A standard temperature and humidity chamber and an electromagnetic shielding room are used to calibrate the environmental sensor. Intermediate inspection items: During data preprocessing and feature extraction, real-time monitoring of data quality and effectiveness of feature extraction is performed. Data quality monitoring determines the degree of data dispersion by calculating the variance and standard deviation of the data. Feature extraction effectiveness inspection compares the prediction accuracy of the model after feature extraction with the accuracy before feature extraction. If the accuracy is improved, it is considered effective. Post-feedback verification items: After determining the transformation ratio result, the result will be fed back to the monitoring center, and the abnormal situation will be secondary checked and recorded. The secondary check uses spare measuring equipment to measure the abnormal current transformer again, and the records include abnormal time, abnormal data and secondary check results.

6. The pre-initialization project according to claim 5, characterized in that: In the parameter setting of the data acquisition unit, the sampling frequency can be adjusted according to the working frequency of the current transformer, the range can be switched manually or automatically, and the data transmission protocol adopts the industrial Ethernet protocol to ensure high-speed and stable data transmission. During the equipment calibration process, the output current accuracy of the standard current source can reach The temperature and humidity control accuracy of the standard temperature and humidity chamber are and , the shielding effectiveness of the electromagnetic shielding room is above.

7. The intermediate inspection item according to claim 5, characterized in that: During data quality monitoring, if the variance of the data is greater than the preset variance threshold or the standard deviation is greater than the preset standard deviation threshold, the data quality is judged to be poor and the parameter settings of the data acquisition unit or the data preprocessing unit need to be rechecked. During the feature extraction validity check, if the prediction accuracy of the model after feature extraction is not improved enough compared with that before feature extraction, , then it is necessary to readjust the parameters or algorithm of the feature extraction unit.

8. The post-feedback verification item according to claim 5, characterized in that: The wireless communication module used for feedback to the monitoring center has the function of automatic signal strength detection and switching. When the main communication network signal is weak, it automatically switches to the backup communication network. The backup measurement equipment used for secondary verification is a high-precision handheld current transformer calibrator, which has higher measurement accuracy than the main measurement equipment. , the recorded information is stored in the local database and backed up to the monitoring center server to prevent data loss.

9. A verification device using the verification method according to claims 1 to 8, characterized in that: It includes a data acquisition unit, a data storage unit, a data preprocessing unit, a feature extraction unit, a deep learning RNN algorithm processing unit, a comparison and judgment unit, and a feedback unit. The units are connected through a data bus. After the data acquisition unit collects data, it is processed by other units in turn and finally the feedback unit outputs the result. The entire device is installed in a cabinet. The cabinet has good heat dissipation and electromagnetic shielding performance. The internal units adopt a modular design to facilitate installation, maintenance and upgrading.

10. A distributed system comprising the verification device according to claim 9, characterized in that: It includes multiple verification devices distributed in different locations of the substation. These devices are connected to the central server through a local area network. The central server summarizes and analyzes the data of each device and manages it in a unified manner to achieve comprehensive monitoring of the current transformer ratio of the entire substation. Each verification device has a unique network address identifier in the distributed system. The central server can perform remote parameter configuration and data query operations on a single or multiple verification devices as needed, and has a fault warning function. When a verification device fails or the data is abnormal, an alarm message is sent to the operation and maintenance personnel in a timely manner.

Images