Article Preview
TopIntroduction
As the power system evolves, relay protection plays a crucial role in ensuring system stability and safe operation. Among them, the relay protection pressure plate, a key component of the protection device, is of crucial importance in maintaining the power system’s stability (Abbassi et al., 2022). However, the growing diversity and quantity of relay protection pressure plates highlights the demand for efficient and accurate identification and verification methods within their operational processes.
Traditional verification methods are often inefficient and present challenges related to quality. Therefore, studying an efficient and accurate method for identifying the status of relay protection pressure plates has become crucial. Status recognition of relay protection pressure plate involves detecting whether the pressure plate is open or closed to determine its operational status. Commonly used state recognition methods include:
- •
Direct Observation Method: This includes observing the appearance of the relay protection pressing plate to determine whether its metal sheets are connected or disconnected.
- •
Sound Method: This involves listening for auditory cues. For instance, the presence of a “da” sound can indicate the status of the spring plate within the relay protection pressure plate. When in the input state, the spring plate will be pressed down and will remain silent. When disconnected state, it will make a “da” sound.
- •
Manual Method: This method entails the manual operation of the relay protection pressure plate and observing its state changes after the operation to identify the input status.
- •
Positioner Method: This involves installing a locator on the relay protection pressure plate and reading position information on the locator to determine the plate’s open or closed status.
- •
Current Method: By detecting the current on the relay protection pressing plate, we can determine the input status. When the current reaches a certain value, it indicates an input state, whereas a current reading of zero indicates a disconnected state.
Depending on the verification challenges, the most appropriate state recognition methods can be selected. For example, for large-scale testing of the relay protection pressure plate status, automated equipment can be used for rapid detection. Conversely, for inspecting a smaller number of pressure plates, manual methods may be a more practical solution.
In similar fields, the following are common self-similarity datasets and proof of effectiveness for selecting appropriate machine learning models:
- •
Weather Forecast Dataset: This dataset contains historical weather data and forecast data. The use of time series analysis methods like random forest time series, ARIMA, and LSTM can predict future weather patterns. By comparing the root mean square error (RMSE) values between predicted results and actual data, one can gauge the effectiveness of these models.
- •
Natural Language Processing Dataset: This dataset contains text data and corresponding labels. Deep learning algorithms, including recurrent neural networks (RNN), convolutional neural networks (CNN), and short-term memory networks (LSTM), are used to develop language models for tasks like text classification or sentiment analysis. The effectiveness of these models can be assessed by calculating the accuracy of prediction results and actual data, F1 scores, and other indicators.
- •
Image Classification Dataset: This dataset contains image data and corresponding labels. Deep learning algorithms like CNN can develop classifiers for image classification tasks. The efficacy of these models is assessed by calculating metrics like accuracy, precision, recall, and other indicators. These are then compared against actual data to demonstrate effectiveness.