A method and system for detecting the position of the tap changer of an on-load tap changer

Abstract

The application provides a method and system for detecting the position of a tap changer of an on-load tap changer, the method comprising: collecting vibration signals, current signals and angular displacement signals during the switching process of the on-load tap changer; conditioning the vibration signals, current signals and angular displacement signals, and preprocessing the conditioned vibration signals, conditioned current signals and conditioned angular displacement signals; extracting key features from the processed vibration signals, processed current signals and processed angular displacement signals; identifying the tap position and switching process of the on-load tap changer according to the key features; and performing abnormal alarm according to the tap position and switching process of the on-load tap changer. The method extracts key features from the vibration signals, current signals and angular displacement signals during the switching process of the on-load tap changer, identifies the tap position and switching process of the on-load tap changer according to the key features, and realizes real-time and accurate detection of the tap position and switching process of the on-load tap changer.

Description

Technical Field

[0001] This application relates to the field of transformer testing technology, and in particular to a method and system for detecting the switching position of an on-load tap changer. Background Technology

[0002] On-load tap changers are key devices for voltage regulation of transformers under load. They achieve stable output voltage regulation by changing the turns ratio of the transformer windings. The normal operation of on-load tap changers is crucial for ensuring power grid voltage quality, and the reliability of their mechanical condition directly affects the operational safety of the transformer.

[0003] Currently, traditional tap changer position detection methods for on-load tap changer position detection and status monitoring mostly employ simple mechanical indicators or position sensors, providing only limited position information and failing to accurately identify the real-time position and switching process of the switch. When detecting single-to-double or double-to-single tap changers, traditional methods cannot provide sufficient time accuracy, and detection systems based on single sensors are easily affected by interference in complex electromagnetic environments, leading to signal distortion. Especially when extracting vibration signal features, traditional linear analysis methods struggle to effectively capture the nonlinear and complex dynamic characteristics of the signal. Existing detection methods typically require operation under power outage conditions or separation of the tap changer from the transformer windings, a complex and time-consuming process. Furthermore, some online monitoring methods suffer from computational complexity and poor real-time performance, making efficient operation on embedded devices difficult.

[0004] With the development of smart grid technology, higher requirements are placed on the fast and accurate voltage regulation characteristics of on-load tap changers, and there is an urgent need for a method and system that can detect tap position switching in real time and with precision. Summary of the Invention

[0005] To address one of the aforementioned technical deficiencies, this application provides a method and system for detecting the tap changer position.

[0006] The first aspect of this application provides a method for detecting the tap changer position of an on-load tap changer, the method comprising: Vibration signals, current signals, and angular displacement signals were collected during the on-load tap changer switching process; The vibration signal, current signal, and angular displacement signal are conditioned to obtain conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal; The conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal are preprocessed to obtain the processed vibration signal, processed current signal, and processed angular displacement signal. Key features are extracted based on the processed vibration signal, processed current signal, and processed angular displacement signal; Identify the tap position and switching process of the on-load tap changer based on key features; An alarm will be triggered based on the tap position and switching process of the on-load tap changer.

[0007] Optionally, the conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal are preprocessed to obtain the processed vibration signal, processed current signal, and processed angular displacement signal, including: The integrity of the conditioned vibration signal, conditioned current signal and conditioned angular displacement signal is checked, and the first processing is performed based on the check results to obtain the first processed signal. The sliding window method is used to evaluate the signal quality of the signal after the first processing at each time period. Based on the signal quality, a second processing is performed to obtain the signal after the second processing. Anomaly detection is performed on the signal after the second processing to remove abnormal data, resulting in the signal after the third processing. The signal after the third processing is decomposed into multiple layers based on the wavelet basis function, and the signal and noise are separated at multiple scales. Noise is then denoised using threshold rules to obtain the processed vibration signal, processed current signal, and processed angular displacement signal.

[0008] Optionally, key features are extracted based on the processed vibration signal, processed current signal, and processed angular displacement signal, including: Spatial features of the processed vibration signal are extracted based on a deep separable convolutional neural network. The deep separable convolutional neural network has a 1×3 convolutional kernel, a stride of 1, zero padding, and a batch normalization layer and a ReLU activation function are added after each convolutional layer. The time-series features of the processed vibration signal, processed current signal, and processed angular displacement signal are extracted based on the Long Short-Term Memory (LSTM) network; the LSM network contains 128 hidden units.

[0009] Optionally, the tap position and switching process of the on-load tap changer can be identified based on key features, including: Identify the peak location of the vibration signal based on key features; Determine the time interval between consecutive peaks; Extract the statistical characteristics of each time interval; the statistical characteristics include mean, variance, skewness and kurtosis; Identify the tap position and switching process of on-load tap changers based on statistical characteristics.

[0010] In a second aspect, this application provides an on-load tap changer position detection system, which includes: a multi-sensor acquisition network, a signal conditioning unit, a data processing system, and an edge computing platform; A multi-sensor acquisition network is used to acquire vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process; The signal conditioning unit is used to condition the vibration signal, current signal and angular displacement signal to obtain the conditioned vibration signal, conditioned current signal and conditioned angular displacement signal; The data processing system is used to preprocess the conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal to obtain the processed vibration signal, processed current signal, and processed angular displacement signal. The edge computing platform is used to extract key features based on the processed vibration signal, processed current signal, and processed angular displacement signal; identify the tap position and switching process of the on-load tap changer based on the key features; and generate abnormal alarms based on the tap position and switching process of the on-load tap changer.

[0011] Optionally, the multi-sensor acquisition network includes: a vibration sensing module, a current sensing module, and a position sensing module; The vibration sensing module is used to collect vibration signals during the on-load tap changer switching process; The current sensing module is used to collect the current signal of the circuit during the on-load tap changer switching process; The position sensing module is used to collect angular displacement signals during the on-load tap changer switching process.

[0012] Optionally, the vibration sensing module includes: three vibration acceleration sensors, respectively arranged on the drive mechanism base of the on-load tap changer, the selector chamber connection flange and the top cover of the transition resistance chamber; the response range of any vibration acceleration sensor is 0.5Hz-10kHz and the sensitivity is 100mV / g. The current sensing module is a Hall current sensor, which is installed on the main circuit of the on-load tap changer, with an accuracy of ±0.5% and a bandwidth of 200kHz. The position sensing module is an angular displacement sensor, installed on the transmission assembly of the on-load tap changer, with a resolution of 18 bits.

[0013] Optionally, the vibration acceleration sensor is fixed by a magnetic base and further secured with beeswax. The Hall current sensor's wires are centered; The concentricity error of the angular displacement sensor is less than 0.05 mm; All signal lines in the multi-sensor acquisition network are double-shielded cables, and the grounding system is set up separately.

[0014] Optionally, the signal conditioning unit includes a multi-channel independent conditioning circuit; wherein the multi-channel includes at least a vibration signal channel and a current signal channel; The vibration signal channel is equipped with an ICP constant current source and an 8th-order Butterworth low-pass filter with a cutoff frequency of 25kHz; the current signal channel uses a 4th-order Bessel low-pass filter with a cutoff frequency of 50kHz. The conditioning circuit for the vibration signal channel includes: an amplifier circuit built with an amplifier with a gain of 100, and a fourth-order anti-aliasing filter network composed of RC passive filters with a cutoff frequency of 10kHz; the conditioning circuit for the current signal channel includes: an amplifier circuit built with an amplifier with a gain of 20. Each conditioning circuit is powered by an isolated power supply, which is a constant voltage isolated DC-DC module with an input voltage of 24VDC and outputs of ±15V and +5V.

[0015] Optionally, the data processing system includes: a processor and an analog-to-digital conversion module; The processor has a clock speed of 480MHz, supports double-precision floating-point operations, and has 2MB of built-in Flash and 1MB of RAM. The highest sampling rate of the analog-to-digital conversion module is 500kSPS.

[0016] This application provides a method and system for detecting the switching position of an on-load tap changer. The method includes: acquiring vibration signals, current signals, and angular displacement signals during the switching process of the on-load tap changer; conditioning the vibration signals, current signals, and angular displacement signals to obtain conditioned vibration signals, conditioned current signals, and conditioned angular displacement signals; preprocessing the conditioned vibration signals, conditioned current signals, and conditioned angular displacement signals to obtain processed vibration signals, processed current signals, and processed angular displacement signals; extracting key features based on the processed vibration signals, processed current signals, and processed angular displacement signals; identifying the tap position and switching process of the on-load tap changer based on the key features; and triggering an alarm based on the tap position and switching process of the on-load tap changer. This method extracts key features from the vibration signals, current signals, and angular displacement signals during the switching process of the on-load tap changer, and identifies the tap position and switching process of the on-load tap changer based on the key features, achieving real-time and accurate detection of the tap position and switching process of the on-load tap changer. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a method for detecting the tap changer position of an on-load tap changer provided in an embodiment of this application; Figure 2This is a schematic diagram of the structure of an on-load tap changer position detection system provided in an embodiment of this application. Detailed Implementation

[0018] To make the technical solutions and advantages of the embodiments of this application clearer, the exemplary embodiments of this application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0019] In the process of developing this application, the inventors discovered that current traditional tap position detection methods for on-load tap changer position detection and status monitoring mostly employ simple mechanical indicators or position sensors, which can only provide limited tap position information and are difficult to accurately identify the real-time position and switching process of the switch. When it is necessary to detect single-to-double or double-to-single tap position switching, traditional methods cannot provide sufficient time accuracy, and detection systems based on a single sensor are easily interfered with in complex electromagnetic environments, leading to signal distortion. Especially when extracting vibration signal features, traditional linear analysis methods are difficult to effectively capture the nonlinear and complex dynamic characteristics of the signal. Existing detection methods usually need to be performed under power outage conditions or require the tap changer to be separated from the transformer windings, which is a complex and time-consuming process. Some online monitoring methods suffer from computational complexity and poor real-time performance, making it difficult to achieve efficient operation on embedded devices. With the development of smart grid technology, higher requirements are placed on the fast and accurate voltage regulation characteristics of on-load tap changers, urgently requiring a method and system that can detect the tap position switching position in real time and with high accuracy.

[0020] To address the aforementioned problems, this application provides a method and system for detecting the on-load tap changer's position switching. The method includes: acquiring vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process; conditioning the vibration, current, and angular displacement signals to obtain conditioned vibration, current, and angular displacement signals; preprocessing the conditioned vibration, current, and angular displacement signals to obtain processed vibration, current, and angular displacement signals; extracting key features based on the processed vibration, current, and angular displacement signals; identifying the on-load tap changer's position and switching process based on the key features; and triggering an alarm based on the on-load tap changer's position and switching process. This method extracts key features from the vibration, current, and angular displacement signals during the on-load tap changer switching process and identifies the on-load tap changer's position and switching process based on these key features, achieving real-time and accurate detection of the on-load tap changer's position and switching process.

[0021] See Figure 1 This embodiment provides a method for detecting the tap changer position of an on-load tap changer. The implementation process of this method is as follows: 101. Collect vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process.

[0022] It can simultaneously collect vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process.

[0023] 102. The vibration signal, current signal and angular displacement signal are conditioned to obtain the conditioned vibration signal, conditioned current signal and conditioned angular displacement signal.

[0024] 103. The conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal are preprocessed to obtain the processed vibration signal, processed current signal, and processed angular displacement signal.

[0025] The preprocessing process in step 103 includes, but is not limited to, signal integrity checks, signal-to-noise ratio calculations, and outlier detection. The vibration signal is preprocessed in segments to compress the data volume and highlight state characteristics.

[0026] The implementation process of step 103 is as follows: 103-1. The integrity of the conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal is checked. Based on the check results, the signal is processed for the first time to obtain the signal after the first processing.

[0027] The first processing step completes the signal integrity check, and the data loss problem caused by transmission interference is identified by analyzing the continuity and integrity of the data.

[0028] 103-2. The sliding window method is used to evaluate the signal quality of the signal after the first processing at each time period. Based on the signal quality, a second processing is performed to obtain the signal after the second processing.

[0029] The signal-to-noise ratio (SNR) can be calculated through a second processing step to achieve quality detection. The SNR calculation uses a sliding window method to dynamically evaluate the signal quality at each time period, providing a basis for subsequent processing.

[0030] 103-3, perform outlier detection on the signal after the second processing, remove outlier data, and obtain the signal after the third processing.

[0031] The outlier detection process can be based on statistical principles, using the 3σ criterion to identify and remove outlier data points that significantly deviate from the normal range.

[0032] The quality assessment was completed after executing steps 103-1 to 103-3.

[0033] 103-4. Based on the wavelet basis function, the signal after the third processing is decomposed into multiple layers to separate the signal and noise at multiple scales. The noise is then denoised by threshold rules to obtain the processed vibration signal, processed current signal and processed angular displacement signal.

[0034] Steps 103-4 involve signal enhancement, such as using a wavelet thresholding denoising algorithm. First, a wavelet basis function (e.g., db8) is selected to decompose the signal into eight levels, separating the signal from noise at different scales. Then, a preset threshold rule (e.g., the SUREShrink threshold rule) is used to adaptively determine the threshold size based on the noise characteristics of each sub-band. Finally, a semi-soft thresholding function is used to complete the denoising process. This method effectively suppresses noise while preserving the useful impulse components in the signal to the maximum extent.

[0035] In addition, standardization will be carried out to ensure that the subsequent data are consistent in units and comparable.

[0036] In addition, the vibration signal will be preprocessed in segments to compress the data volume and highlight the segmentation of state characteristic signals.

[0037] 104. Based on the processed vibration signal, processed current signal, and processed angular displacement signal, key features are extracted.

[0038] The implementation process of step 104 is as follows: 104-1, Spatial features of processed vibration signals are extracted based on a deep separable convolutional neural network (DSCNN).

[0039] Among them, the Deeply Separable Convolutional Neural Network (DSCNN) has a 1×3 convolutional kernel with a stride of 1, zero padding, and a batch normalization layer and ReLU activation function added after each convolutional layer.

[0040] Deeply Separable Convolutional Neural Networks (DSCNNs) first fuse information between channels through pointwise convolutions, and then use depthwise convolutions to extract local spatial features from each channel. This decompositional structure significantly reduces the number of network parameters, making it more suitable for deployment on embedded devices. The convolutional operations use a 1×3 kernel size with a stride of 1, and zero-padding is used to maintain the feature map size. Each convolutional layer is followed by a batch normalization layer and a ReLU activation function to accelerate network convergence and improve generalization ability. After spatial feature extraction, the feature sequence is fed into a Long Short-Term Memory (LSTM) network for temporal modeling.

[0041] 104-2, Time series features of processed vibration signals, processed current signals and processed angular displacement signals are extracted based on Long Short-Term Memory (LSTM) network.

[0042] The Long Short-Term Memory (LSTM) network contains 128 hidden units. It captures long-range dependencies in signals through a carefully designed gating mechanism. Its bidirectional structure allows the network to utilize information from both preceding and following time points simultaneously, enabling a better understanding of the complete switching process.

[0043] Signal features can be extracted based on a combination of Deeply Separable Convolutional Neural Network (DSCNN) and Long Short-Term Memory Network (LSTM) through steps 104-1 and 104-2.

[0044] Deeply separable convolutional neural networks (DSCNN) are responsible for extracting spatial features from vibration signals, while long short-term memory networks (LSTM) capture time-series features in the signals.

[0045] In addition, to further optimize the feature extraction effect in the specific implementation, a self-attention (SA) mechanism can be introduced when performing steps 104-1 and 104-2 to adaptively assign weights to different features.

[0046] In other words, a self-attention (SA) mechanism is introduced on top of deep learning networks to further enhance feature representation capabilities. This mechanism processes the feature sequence output by the Long Short-Term Memory (LSTM) network in parallel through eight independent attention heads, each responsible for evaluating the importance of features from different perspectives. Specifically, each attention head first linearly maps the input features to three spaces: query, key, and value. Then, it calculates the association weights between time steps using scaled dot product attention. Finally, it performs a weighted summation of the value vectors based on these weights. The outputs of multiple heads are concatenated and linearly transformed to form the final feature representation. This design allows the model to adaptively focus on feature fragments most relevant to the current classification task, effectively improving the model's discriminative ability.

[0047] In step 104, a deep separable convolutional neural network (DSCNN) can be used to extract local features of the vibration signal, a long short-term memory network (LSTM) can be used to capture the temporal dependence of the signal, and a self-attention (SA) mechanism can be used to select the key features that are most effective for classification.

[0048] 105. Identify the tap position and switching process of the on-load tap changer based on key features.

[0049] The implementation process of step 105 is as follows: 105-1, Identify the peak position of the vibration signal based on key features.

[0050] 105-2, determine the time interval between consecutive peaks.

[0051] 105-3, extract the statistical characteristics of each time interval.

[0052] The statistical characteristics include mean, variance, skewness, and kurtosis.

[0053] 105-4. Identify the tap position and switching process of the on-load tap changer based on statistical characteristics.

[0054] For example, in step 101, vibration, current, and angular displacement signals are collected during each switching process of the on-load tap changer. In step 102, the vibration, current, and angular displacement signals are conditioned. In step 103, the conditioned vibration, current, and angular displacement signals are preprocessed. In step 104, key features are extracted based on the processed vibration, current, and angular displacement signals, such as the time interval between vibration signals under the two switching modes. In step 105-1, the peak position of the vibration signal is identified based on the key features (such as the time interval between vibration signals under the two switching modes). In step 105-2, the time interval between consecutive peaks is determined. In step 105-3, the mean, variance, skewness, and kurtosis of each time interval are extracted. In step 105-4, the time interval with the largest difference is selected as the feature value based on statistical characteristics. According to the preset time criteria, the feature value is compared with the time criteria to automatically identify the tap position of the on-load tap changer and the switching mode of single-double and double-single.

[0055] Specifically, step 105 can be implemented based on a classifier. The classifier is optimized using the Particle Swarm Optimization (PSO) algorithm, which can identify the gear position status of the key features extracted in step 104. The PSO algorithm is used to find the optimal parameters of the classifier, improving recognition accuracy and noise resistance. In step 105, the gear position status of the on-load tap changer can be identified by analyzing the time interval characteristics between vibration signals and setting time criteria, thereby distinguishing between single-to-double and double-to-single gear switching modes.

[0056] For example, a Support Vector Machine (SVM) classifier optimized using the Particle Swarm Optimization (PSO) algorithm is employed. The PSO optimization process first randomly initializes 100 particles, each representing a set of SVM penalty factors and kernel function parameters. During iteration, particles update their velocity and position based on their individual and group historical best positions. The fitness function comprehensively considers classification accuracy and model complexity, ensuring performance while avoiding overfitting. The optimization process continues for 200 generations, ultimately yielding the optimal parameter combination. For the identification of single-to-double and double-to-single gear shifting, a discrimination logic based on time interval analysis is also employed. This logic first accurately detects the peak position of the vibration signal, calculates the time interval between consecutive peaks, and then extracts the statistical features of these time intervals, including mean, variance, skewness, and kurtosis. By analyzing the distribution patterns of these features under different switching modes, the two switching methods are effectively distinguished.

[0057] 106. An abnormal alarm will be triggered based on the tap position and switching process of the on-load tap changer.

[0058] In step 106, the on-load tap changer position and switching process obtained in step 105 can be transmitted to the host computer system, and an alarm can be triggered when an abnormal state is detected based on the on-load tap changer position and switching process.

[0059] The process of triggering the alarm when an abnormal state is detected based on the tap position and switching process of the on-load tap changer can be as follows: by combining the current signal and the angular displacement signal, the mechanical state of the on-load tap changer is comprehensively evaluated, and an alarm signal is generated when an abnormal state is detected.

[0060] In comprehensive evaluation, a multi-feature fusion analysis strategy can be adopted. For example, the following features can be extracted: tap position and contact movement trajectory characteristics during the switching process of the on-load tap changer (which can be obtained based on the processed angular displacement signal, including travel time, velocity, and acceleration curves); electrical characteristic characteristics (which can be obtained based on the processed current signal, including current amplitude, rise time, and energy distribution); and mechanical impact characteristics (which can be obtained based on the processed vibration signal, including impact intensity, frequency components, and energy attenuation characteristics). These features can be extracted in step 104.

[0061] In step 106, these features can be comprehensively judged by a decision-level fusion algorithm, in which a random forest-based classifier independently evaluates each feature subset, and then a weighted voting mechanism is used to obtain the final diagnostic conclusion.

[0062] In addition, a historical database of equipment status can be established, and by analyzing the long-term changing trends of characteristic values, early warning of equipment performance degradation and prediction of remaining life can be achieved.

[0063] The communication interface with the host computer system supports Ethernet, Wi-Fi, or 4G / 5G wireless transmission.

[0064] In practical implementation, the on-load tap changer position detection method provided in this embodiment can be triggered in three ways: current threshold trigger mode, which automatically starts data acquisition when the detected circuit current exceeds a set value; vibration energy trigger mode, which monitors mechanical actions in real time based on a vibration sensor; and manual trigger mode, which supports remote command control. Each trigger mode can be independently configured with a pre-trigger buffer, and the buffer depth can be configured from 1 to 10 seconds to ensure complete recording of the entire switching action process. The acquired parameters are saved in an EEPROM memory and support automatic recovery upon power-on, avoiding repeated configuration.

[0065] The on-load tap changer position detection method provided in this embodiment operates efficiently on an embedded platform through a task scheduling mechanism. Tasks such as data acquisition, preprocessing, feature extraction, and state recognition are assigned to different processing cores, and data sharing and pipelined processing are achieved through a circular buffer. The system also features online learning capabilities, dynamically adjusting model parameters based on newly acquired sample data to gradually improve recognition accuracy.

[0066] The on-load tap changer position detection method provided in this embodiment integrates multi-sensor information (vibration, current, and position) and combines it with a deep learning algorithm to accurately identify the tap changer position and switching process. Experiments show that the PSO-optimized DSCNN-LSTM-SA method achieves higher tap changer recognition accuracy compared to mainstream classification algorithms. Employing depthwise separable convolution and self-attention mechanisms effectively extracts key features from the signal, suppresses noise interference, and ensures reliable operation even in complex electromagnetic environments. Through signal segmentation preprocessing and embedded edge computing technology, the computational complexity is significantly reduced while maintaining recognition accuracy, making it suitable for online monitoring and real-time diagnosis. It can detect not only normal tap changers but also various fault states such as mechanical jamming and contact wear. For example, by analyzing angular displacement signals, the contact opening distance and overtravel can be calculated, allowing for timely detection of mechanical faults caused by wear.

[0067] The on-load tap changer position detection method provided in this embodiment can solve the problems of insufficient detection accuracy, poor anti-interference ability and low fault identification efficiency in the prior art.

[0068] This embodiment provides a method and system for detecting the switching position of an on-load tap changer. The method includes: acquiring vibration signals, current signals, and angular displacement signals during the switching process of the on-load tap changer; conditioning the vibration signals, current signals, and angular displacement signals to obtain conditioned vibration signals, conditioned current signals, and conditioned angular displacement signals; preprocessing the conditioned vibration signals, conditioned current signals, and conditioned angular displacement signals to obtain processed vibration signals, processed current signals, and processed angular displacement signals; extracting key features based on the processed vibration signals, processed current signals, and processed angular displacement signals; identifying the tap position and switching process of the on-load tap changer based on the key features; and triggering an alarm based on the tap position and switching process of the on-load tap changer. This embodiment's method extracts key features from the vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process, and identifies the tap position and switching process of the on-load tap changer based on these key features, achieving real-time and accurate detection of the tap position and switching process of the on-load tap changer.

[0069] Based on the same inventive concept as the above-mentioned on-load tap changer position detection method, this embodiment provides an on-load tap changer position detection system, see [link to relevant documentation]. Figure 2 The system includes: a multi-sensor acquisition network, a signal conditioning unit, a data processing system, and an edge computing platform.

[0070] 1. Multi-sensor acquisition network A multi-sensor acquisition network is used to collect vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process. For example, it can simultaneously collect vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process.

[0071] All signal lines in the multi-sensor acquisition network are double-shielded cables, and the grounding system is set up separately to minimize interference during signal transmission.

[0072] The multi-sensor acquisition network includes: a vibration sensing module, a current sensing module, and a position sensing module.

[0073] 1) Vibration sensing module The vibration sensing module is used to collect vibration signals during the on-load tap changer switching process.

[0074] The vibration sensing module includes three vibration acceleration sensors, which are respectively arranged on the drive mechanism base of the on-load tap changer, the connecting flange of the selector chamber, and the top cover of the transition resistance chamber, forming a three-dimensional monitoring network.

[0075] Among them, the vibration acceleration sensor can be an ICP (Integrated Circuit Piezoelectric) acceleration sensor.

[0076] The response range of any vibration acceleration sensor is 0.5Hz (Hertz) - 10kHz, and the sensitivity is 100mV / g (millivolts / gravitational acceleration). The high sensitivity ensures complete capture of the high-frequency vibration signal generated by the switch action.

[0077] The vibration acceleration sensor can be fixed with a magnetic base and further secured with beeswax to ensure close contact with the measurement surface.

[0078] 2) Current sensing module The current sensing module is used to collect the current signal of the circuit during the on-load tap changer switching process.

[0079] The current sensing module is a Hall current sensor, which is installed on the main circuit of the on-load tap changer. It has an accuracy of ±0.5% and a bandwidth of 200kHz, ensuring accurate measurement of the current signal.

[0080] For example, a Hall current sensor is a closed-loop Hall current sensor.

[0081] The leads of the Hall current sensor should be centered. In other words, special care must be taken to center the leads when installing the Hall current sensor to avoid measurement errors.

[0082] 3) Position sensing module The position sensing module is used to detect the action information of the switching disconnector, and then collect the angular displacement signal during the on-load tap changer switching process.

[0083] The position sensing module is an angular displacement sensor, which is installed on the transmission component of the on-load tap changer (e.g., directly connected to the transmission shaft via a special coupling). Its 18-bit resolution can accurately reflect changes in the contact position.

[0084] For example, the angular displacement sensor is an ATM60 absolute encoder.

[0085] The concentricity error of the angular displacement sensor is less than 0.05 mm. In other words, a laser alignment instrument can be used to ensure that the concentricity error is less than 0.05 mm when installing the angular displacement sensor.

[0086] 2. Signal conditioning unit The signal conditioning unit is used to condition vibration signals, current signals, and angular displacement signals to obtain conditioned vibration signals, conditioned current signals, and conditioned angular displacement signals.

[0087] The signal conditioning unit includes multi-channel independent conditioning circuits, meaning that each sensor channel is equipped with an independent signal conditioning circuit.

[0088] Among them, the multi-channel includes at least a vibration signal channel and a current signal channel.

[0089] For vibration signal channels The vibration signal channel is equipped with an ICP constant current source and an 8th-order Butterworth low-pass filter with a cutoff frequency of 25kHz to ensure signal quality while maintaining phase characteristics.

[0090] The conditioning circuit of the vibration signal channel includes: an amplifier circuit built with an amplifier with a gain of 100 times, and a fourth-order anti-aliasing filter network composed of RC (Resistor-Capacitance) passive filters with a cutoff frequency of 10kHz.

[0091] For current signal channel The current signal channel uses a fourth-order Bessel low-pass filter with a cutoff frequency of 50kHz.

[0092] The conditioning circuit for the current signal channel includes an amplification circuit built with an amplifier with a gain of 20 (such as a precision operational amplifier). It can also integrate electromagnetic isolation protection circuitry to prevent high-voltage crosstalk in the field.

[0093] All conditioning circuits can use analog switches for channel selection, and the core microcontroller enables digital control of gain and filter parameters.

[0094] In addition, each conditioning circuit can be powered by an isolated power supply, effectively suppressing electromagnetic interference on site.

[0095] The power supply uses a constant voltage isolated DC (Direct Current)-DC module with an input voltage of 24VDC and outputs ±15V and +5V to supply power to analog and digital circuits respectively, effectively suppressing power supply noise.

[0096] The signal conditioning unit amplifies, filters, and digitizes vibration, current, and angular displacement signals acquired by a multi-sensor network. Specifically, this module performs segmented preprocessing on the vibration, current, and angular displacement signals to compress them and highlight their state characteristics. The conditioned vibration, current, and angular displacement signals obtained by the signal conditioning unit... Figure 2 The original data signal in the data.

[0097] In practical implementation, the multi-sensor acquisition network and signal conditioning unit can adopt a hierarchical distributed architecture to optimize cost configuration while ensuring performance. It can be flexibly adjusted to meet the monitoring needs of on-load tap changers.

[0098] The on-load tap changer position detection system provided in this embodiment achieves 8-channel time-division multiplexing acquisition through a multi-channel analog switch with a conditioning circuit, and ensures synchronization with a sample-and-hold circuit. This on-load tap changer position detection system can be equipped with 128MB of SDRAM as a data cache and uses a 4GB industrial-grade TF card for local data storage, supporting continuous storage of more than 240 hours of monitoring data. The clock circuit uses a DS3231 high-precision real-time clock chip with a temperature compensation accuracy of ±2ppm, ensuring the accuracy of the timestamp.

[0099] 3. Data Processing System The data processing system is used to preprocess the conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal to obtain the processed vibration signal, processed current signal, and processed angular displacement signal.

[0100] The data processing system includes: a processor and an analog-to-digital converter module.

[0101] The processor has a clock speed of 480MHz, supports double-precision floating-point operations, and has 2MB of Flash memory and 1MB of RAM.

[0102] The highest sampling rate of the analog-to-digital converter module is 500kSPS (thousands of samples per second), and in practical applications, the working sampling rate can be 100kSPS.

[0103] In practical implementation, the preprocessing process of the data processing system includes, but is not limited to, signal integrity checks, signal-to-noise ratio calculations, and outlier detection. It also involves segmenting the vibration signal for preprocessing, compressing the data volume, and highlighting state characteristics.

[0104] The implementation process of a data processing system can be as follows: 1) Perform integrity checks on the conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal. Based on the check results, perform the first processing to obtain the signal after the first processing.

[0105] The first processing step completes the signal integrity check, and the data loss problem caused by transmission interference is identified by analyzing the continuity and integrity of the data.

[0106] 2) The sliding window method is used to evaluate the signal quality of the signal after the first processing at each time period. Based on the signal quality, a second processing is performed to obtain the signal after the second processing.

[0107] The signal-to-noise ratio (SNR) can be calculated through a second processing step to achieve quality detection. The SNR calculation uses a sliding window method to dynamically evaluate the signal quality at each time period, providing a basis for subsequent processing.

[0108] 3) Perform outlier detection on the signal after the second processing, remove outlier data, and obtain the signal after the third processing.

[0109] The outlier detection process can be based on statistical principles, using the 3σ criterion to identify and remove outlier data points that significantly deviate from the normal range.

[0110] The quality assessment was completed after executing steps 1)-3).

[0111] 4) The signal after the third processing is decomposed into multiple layers based on the wavelet basis function, and the signal and noise are separated at multiple scales. The noise is then denoised by threshold rules to obtain the processed vibration signal, processed current signal and processed angular displacement signal.

[0112] 4) This refers to the process of signal enhancement, such as denoising using wavelet thresholding algorithms. First, a wavelet basis function (e.g., db8) is selected to decompose the signal into eight levels, separating the signal from noise at different scales. Then, a preset threshold rule (e.g., the SUREShrink threshold rule) is used to adaptively determine the threshold size based on the noise characteristics of each sub-band. Finally, a semi-soft thresholding function is used to complete the denoising process. This method effectively suppresses noise while preserving the useful impulse components in the signal to the maximum extent.

[0113] In addition, standardization will be carried out to ensure that the subsequent data are consistent in units and comparable.

[0114] In addition, the vibration signal will be preprocessed in segments to compress the data volume and highlight the segmentation of state characteristic signals.

[0115] 4. Edge computing platform The edge computing platform extracts key features from processed vibration, current, and angular displacement signals. Based on these features, it identifies the tap position and switching process of the on-load tap changer. Anomaly alarms are then triggered based on the tap position and switching process of the on-load tap changer.

[0116] An edge computing platform may include: a feature extraction module, a gear position recognition module, and a communication alarm module.

[0117] 1) Feature extraction module The feature extraction module is used to extract key features based on the processed vibration signal, processed current signal, and processed angular displacement signal.

[0118] Specifically, the feature extraction module can extract signal features based on a combination of Deeply Separable Convolutional Neural Network (DSCNN) and Long Short-Term Memory Network (LSTM).

[0119] Deeply separable convolutional neural networks (DSCNN) are responsible for extracting spatial features from vibration signals, while long short-term memory networks (LSTM) capture time-series features in the signals.

[0120] Among them, the Deeply Separable Convolutional Neural Network (DSCNN) has a 1×3 convolutional kernel with a stride of 1, zero padding, and a batch normalization layer and ReLU activation function added after each convolutional layer.

[0121] The Deeply Separable Convolutional Neural Network (DSCNN) first achieves information fusion between channels through pointwise convolution, and then uses depthwise convolution to extract local spatial features of each channel. This decompositional structure significantly reduces the number of network parameters, making the on-load tap changer position detection system provided in this embodiment more suitable for deployment on embedded devices. The convolution operation uses a 1×3 kernel size with a stride of 1, and zero padding is used to keep the feature map size unchanged. A batch normalization layer and a ReLU activation function are added after each convolutional layer to accelerate network convergence and improve generalization ability. After spatial feature extraction, the feature sequence is fed into a Long Short-Term Memory (LSTM) network for temporal modeling.

[0122] The Long Short-Term Memory (LSTM) network contains 128 hidden units. It captures long-range dependencies in signals through a carefully designed gating mechanism. Its bidirectional structure allows the network to utilize information from both preceding and following time points simultaneously, enabling a better understanding of the complete switching process.

[0123] Therefore, the process by which the feature extraction module extracts key features based on the processed vibration signal, processed current signal, and processed angular displacement signal can be described as follows: 1) Extract spatial features of the processed vibration signal based on a deep separable convolutional neural network (DSCNN).

[0124] 2) Extract time-series features of the processed vibration signal, processed current signal, and processed angular displacement signal based on Long Short-Term Memory (LSTM) network.

[0125] Signal features can be extracted using a combination of Deeply Separable Convolutional Neural Networks (DSCNN) and Long Short-Term Memory Networks (LSTM) structures, 1) and 2).

[0126] The feature extraction module uses a deep separable convolutional neural network (DSCNN) to extract local features of vibration signals, uses a long short-term memory network (LSTM) to capture the temporal dependence of signals, and uses a self-attention (SA) mechanism to select the key features most effective for classification.

[0127] In addition, to further optimize the feature extraction effect in the specific implementation, the feature extraction module can also introduce a self-attention (SA) mechanism to adaptively assign weights to different features.

[0128] In other words, a self-attention (SA) mechanism is introduced on top of deep learning networks to further enhance feature representation capabilities. This mechanism processes the feature sequence output by the Long Short-Term Memory (LSTM) network in parallel through eight independent attention heads, each responsible for evaluating the importance of features from different perspectives. Specifically, each attention head first linearly maps the input features to three spaces: query, key, and value. Then, it calculates the association weights between time steps using scaled dot product attention. Finally, it performs a weighted summation of the value vectors based on these weights. The outputs of multiple heads are concatenated and linearly transformed to form the final feature representation. This design allows the model to adaptively focus on feature fragments most relevant to the current classification task, effectively improving the model's discriminative ability.

[0129] 2) Gear Position Recognition Module The gear position recognition module is used to identify the gear position and switching process of the on-load tap changer based on key features.

[0130] The gear position recognition module can identify the gear position and switching process of the on-load tap changer based on key features as follows: 1) Identify the peak position of the vibration signal based on key features.

[0131] 2) Determine the time interval between consecutive peaks.

[0132] 3) Extract the statistical characteristics of each time interval.

[0133] The statistical characteristics include mean, variance, skewness, and kurtosis.

[0134] 4) Identify the tap position and switching process of the on-load tap changer based on statistical characteristics.

[0135] For example, a multi-sensor acquisition network collects vibration, current, and angular displacement signals during each switching process of an on-load tap changer. A signal conditioning unit conditions these signals. A data processing system preprocesses the conditioned vibration, current, and angular displacement signals. The edge computing platform's feature extraction module extracts key features based on the processed vibration, current, and angular displacement signals, such as the time interval between vibration signals under the two switching modes. In step 1), the gear position recognition module identifies the peak position of the vibration signal based on key features (such as the time interval between vibration signals under the two switching modes). In step 2), the module determines the time interval between consecutive peaks. In step 3), the module extracts the mean, variance, skewness, and kurtosis of each time interval. In section 4), the gear position recognition module will select the time interval with the largest difference as the feature value based on statistical characteristics. According to the preset time criteria, the feature value will be compared with the time criteria to automatically identify the gear position of the on-load tap changer and the switching mode of single-double and double-single.

[0136] Specifically, the gear position recognition module utilizes a classifier optimized using the Particle Swarm Optimization (PSO) algorithm. The classifier identifies the gear position status based on key features extracted by the feature extraction module. The PSO algorithm is used to find the optimal parameters for the classifier, improving recognition accuracy and noise resistance. The gear position recognition module can distinguish between single-to-double and double-to-single gear shifting modes by analyzing the time interval characteristics between vibration signals and setting time criteria.

[0137] For example, the gear position recognition module uses a support vector machine classifier optimized by the particle swarm optimization algorithm. The particle swarm optimization process first randomly initializes 100 particles, each representing a set of SVM penalty factors and kernel function parameters. During iteration, particles update their velocity and position based on their individual historical best position and the group's historical best position. The fitness function comprehensively considers classification accuracy and model complexity, ensuring performance while avoiding overfitting. The optimization process continues for 200 generations, ultimately obtaining the optimal parameter combination. For the recognition of single-to-double and double-to-single gear shifting, the on-load tap changer gear shifting position detection system provided in this embodiment also employs a discrimination logic based on time interval analysis. This logic first accurately detects the peak position of the vibration signal, calculates the time interval between consecutive peaks, and then extracts the statistical features of these time intervals, including mean, variance, skewness, and kurtosis. By analyzing the distribution patterns of these features under different switching modes, the two switching methods are effectively distinguished.

[0138] 3) Communication alarm module The communication alarm module is used to generate alarms based on the tap position and switching process of the on-load tap changer.

[0139] The communication alarm module can transmit the on-load tap changer position and switching process obtained by the position recognition module to the host computer system, and trigger the alarm when an abnormal state is detected based on the on-load tap changer position and switching process.

[0140] The communication alarm module can combine current signals and angular displacement signals to comprehensively evaluate the mechanical state of the on-load tap changer and generate an alarm signal when an abnormal state is detected.

[0141] In comprehensive evaluation, a multi-feature fusion analysis strategy can be adopted. For example, the following features can be extracted: tap position and contact movement trajectory characteristics during the switching process of the on-load tap changer (which can be obtained based on the processed angular displacement signal, including travel time, velocity, and acceleration curves); electrical characteristic characteristics (which can be obtained based on the processed current signal, including current amplitude, rise time, and energy distribution); and mechanical impact characteristics (which can be obtained based on the processed vibration signal, including impact intensity, frequency components, and energy attenuation characteristics). These features can be extracted in step 104.

[0142] The communication alarm module also uses a decision-level fusion algorithm to make a comprehensive judgment on these features. The random forest-based classifier evaluates each feature subset independently, and then a weighted voting mechanism is used to arrive at the final diagnostic conclusion.

[0143] In addition, a historical database of equipment status can be established, and by analyzing the long-term changing trends of characteristic values, early warning of equipment performance degradation and prediction of remaining life can be achieved.

[0144] The communication alarm module and the host computer system support Ethernet, Wi-Fi or 4G / 5G wireless transmission via their communication interface.

[0145] In practical implementation, the on-load tap changer position detection system provided in this embodiment can be triggered in three ways: current threshold trigger mode, which automatically starts data acquisition when the detected circuit current exceeds a set value; vibration energy trigger mode, which monitors mechanical actions in real time based on vibration sensors; and manual trigger mode, which supports remote command control. Each trigger mode can be independently configured with a pre-trigger buffer, and the buffer depth can be configured from 1 to 10 seconds to ensure complete recording of the entire switching action process. The acquired parameters are saved in an EEPROM memory and support automatic recovery upon power-on, avoiding repeated configuration.

[0146] The on-load tap changer position detection system provided in this embodiment can adopt various standard industrial interfaces. For example, the RS485 interface uses the MAX3485 chip, supports the Modbus RTU protocol, and has a maximum baud rate of 115200bps (bits per second), for communication with field intelligent devices. The Ethernet interface uses the DM9000A chip, supports the TCP / IP protocol stack, and can achieve 10 / 100M (megabit) adaptive network connectivity. A 4G module interface is also reserved, using a Quectel EC20 series module, supporting the LTE Cat4 standard for wireless data transmission. All communication interfaces are equipped with opto-isolation protection with an isolation voltage of 2500Vrms (effective voltage value), ensuring stable operation of the system in strong electromagnetic environments.

[0147] The on-load tap changer position detection system provided in this embodiment features a mechanical structure with an aluminum alloy protective housing, achieving an IP54 protection rating. The internal circuit board is mounted via a DIN rail for easy maintenance. The power input design offers a wide voltage range (18-36VDC) and supports reverse connection and overvoltage protection. The overall power consumption of this on-load tap changer position detection system is controlled below 15W, making it suitable for long-term continuous operation in industrial settings. On the software side, the system utilizes the FreeRTOS real-time operating system with a task scheduling cycle of 1ms, ensuring real-time data acquisition and accuracy. This on-load tap changer position detection system significantly reduces costs while meeting performance requirements, providing feasibility for large-scale engineering applications.

[0148] The on-load tap changer position detection system provided in this embodiment integrates multi-sensor information (vibration, current, and position) and combines it with deep learning algorithms to accurately identify the tap changer's position and switching process. Experiments show that the PSO-optimized DSCNN-LSTM-SA method achieves higher tap changer recognition accuracy compared to mainstream classification algorithms. Employing depthwise separable convolution and self-attention mechanisms effectively extracts key features from the signal, suppresses noise interference, and ensures reliable operation even in complex electromagnetic environments. Through signal segmentation preprocessing and embedded edge computing technology, the system significantly reduces computational complexity while maintaining recognition accuracy, making it suitable for online monitoring and real-time diagnosis. It can detect not only normal tap changers but also various fault states such as mechanical jamming and contact wear. For example, by analyzing angular displacement signals, the contact opening distance and overtravel can be calculated, allowing for timely detection of mechanical faults caused by wear.

[0149] The on-load tap changer position detection system provided in this embodiment can solve the problems of insufficient detection accuracy, poor anti-interference ability and low fault identification efficiency in the prior art.

[0150] The system provided in this embodiment extracts key features from vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process. Based on these key features, it identifies the tap position and switching process of the on-load tap changer, achieving real-time and accurate detection of the tap position and switching process of the on-load tap changer.

[0151] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0152] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0153] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0154] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0155] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0156] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0157] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting the tap changer position of an on-load tap changer, characterized in that, The method includes: Vibration signals, current signals, and angular displacement signals were collected during the on-load tap changer switching process; The vibration signal, current signal, and angular displacement signal are conditioned to obtain conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal; The conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal are preprocessed to obtain the processed vibration signal, processed current signal, and processed angular displacement signal. Key features are extracted based on the processed vibration signal, processed current signal, and processed angular displacement signal; Identify the tap position and switching process of the on-load tap changer based on the aforementioned key features; An alarm will be triggered based on the tap position and switching process of the on-load tap changer.

2. The method according to claim 1, characterized in that, The preprocessing of the conditioned vibration signal, conditioned current signal, and conditioned angular displacement signal to obtain the processed vibration signal, processed current signal, and processed angular displacement signal includes: The integrity of the conditioned vibration signal, conditioned current signal and conditioned angular displacement signal is checked, and the first processing is performed based on the check results to obtain the first processed signal. The sliding window method is used to evaluate the signal quality of the signal after the first processing at each time period. Based on the signal quality, a second processing is performed to obtain the signal after the second processing. Anomaly detection is performed on the signal after the second processing to remove abnormal data, resulting in the signal after the third processing. The signal after the third processing is decomposed into multiple layers based on the wavelet basis function, and the signal and noise are separated at multiple scales. Noise is then denoised using threshold rules to obtain the processed vibration signal, processed current signal, and processed angular displacement signal.

3. The method according to claim 1, characterized in that, The key features extracted based on the processed vibration signal, processed current signal, and processed angular displacement signal include: Spatial features of the processed vibration signal are extracted based on a deep separable convolutional neural network. The deep separable convolutional neural network has a 1×3 convolutional kernel, a stride of 1, zero padding, and a batch normalization layer and a ReLU activation function are added after each convolutional layer. The time-series features of the processed vibration signal, processed current signal, and processed angular displacement signal are extracted based on the Long Short-Term Memory (LSTM) network; the LSM network contains 128 hidden units.

4. The method according to claim 1, characterized in that, The process of identifying the tap position and switching process of the on-load tap changer based on the key features includes: Identify the peak position of the vibration signal based on the key features; Determine the time interval between consecutive peaks; Extract the statistical characteristics of each time interval; the statistical characteristics include mean, variance, skewness and kurtosis; Identify the tap position and switching process of on-load tap changers based on statistical characteristics.

5. A system for detecting the position switching of an on-load tap changer, characterized in that, The system includes: a multi-sensor acquisition network, a signal conditioning unit, a data processing system, and an edge computing platform; The multi-sensor acquisition network is used to acquire vibration signals, current signals, and angular displacement signals during the on-load tap changer switching process; The signal conditioning unit is used to condition the vibration signal, current signal and angular displacement signal to obtain the conditioned vibration signal, the conditioned current signal and the conditioned angular displacement signal. The data processing system is used to preprocess the conditioned vibration signal, the conditioned current signal, and the conditioned angular displacement signal to obtain the processed vibration signal, the processed current signal, and the processed angular displacement signal. The edge computing platform is used to extract key features based on the processed vibration signal, processed current signal, and processed angular displacement signal; identify the tap position and switching process of the on-load tap changer based on the key features; and generate an alarm for abnormality based on the tap position and switching process of the on-load tap changer.

6. The system according to claim 5, characterized in that, The multi-sensor acquisition network includes: a vibration sensing module, a current sensing module, and a position sensing module; The vibration sensing module is used to collect vibration signals during the on-load tap changer switching process; The current sensing module is used to collect the current signal of the circuit during the on-load tap changer switching process. The position sensing module is used to collect angular displacement signals during the on-load tap changer switching process.

7. The system according to claim 6, characterized in that, The vibration sensing module includes three vibration acceleration sensors, which are respectively arranged on the drive mechanism base of the on-load tap changer, the selector chamber connection flange, and the top cover of the transition resistance chamber; the response range of any vibration acceleration sensor is 0.5Hz-10kHz, and the sensitivity is 100mV / g. The current sensing module is a Hall current sensor, which is installed on the main circuit of the on-load tap changer, with an accuracy of ±0.5% and a bandwidth of 200kHz. The position sensing module is an angular displacement sensor, which is installed on the transmission component of the on-load tap changer and has a resolution of 18 bits.

8. The system according to claim 7, characterized in that, The vibration acceleration sensor is fixed by a magnetic base and further attached with beeswax. The wires of the Hall current sensor are centered; The concentricity error of the angular displacement sensor is less than 0.05 mm; All signal lines in the multi-sensor acquisition network are double-shielded cables, and the grounding system is set up separately.

9. The system according to claim 5, characterized in that, The signal conditioning unit includes multi-channel independent conditioning circuitry; wherein, the multi-channel includes at least a vibration signal channel and a current signal channel; The vibration signal channel is equipped with an ICP constant current source and an 8th-order Butterworth low-pass filter with a cutoff frequency of 25kHz; the current signal channel uses a 4th-order Bessel low-pass filter with a cutoff frequency of 50kHz. The conditioning circuit for the vibration signal channel includes: an amplifier circuit built with an amplifier with a gain of 100, and a fourth-order anti-aliasing filter network composed of RC passive filters with a cutoff frequency of 10kHz; the conditioning circuit for the current signal channel includes: an amplifier circuit built with an amplifier with a gain of 20. Each conditioning circuit is powered by an isolated power supply, which is a constant voltage isolated DC-DC module with an input voltage of 24VDC and outputs of ±15V and +5V.

10. The system according to claim 5, characterized in that, The data processing system includes: a processor and an analog-to-digital conversion module; The processor has a main frequency of 480MHz, supports double-precision floating-point operations, and has 2MB of built-in Flash and 1MB of RAM. The highest sampling rate of the analog-to-digital conversion module is 500kSPS.

Images