Deep Learning Method for Fault Diagnosis in High Voltage Transmission Lines: A Case of the Vietnam 220kV Transmission Line

Transmission Line Fault Classification Based on the Combination of Scaled Wavelet Scalograms and CNNs Using a One-Side Sensor for Data Collection

“…In addition, no particular variable has proven to be adequate to be considered the best variable for fault categorization, despite the high accuracy in some of them, and the accuracy varied between studies [28][29][30][31]. Part of the challenge is considering most of the parameters that can affect the fault classification in Table 2 that were not presented in [28,34], or considering some of them, as in [18,25,27,30,31,38], while the others they did not even mention, as in [22,26]. For the latter two references, they achieved 100% accuracy.…”

“…For the latter two references, they achieved 100% accuracy. In [34], they used only a small amount of data (210) with a very simple network with no testing accuracy for their model. In [18], despite the fact they reached 100% in validation, the testing shows a different accuracy, 98.61%, which represents the real classification accuracy because the new data will be classified according to the testing accuracy, not the validation, which is considered the true standard for evaluating system performance.…”

“…While the use of machine learning in different fields gained this reputation for categorization, many researchers raced to use different algorithms in the field of deep learning prior to preparing the input for categorization [32][33][34]. Even though they only used deep learning in different articles because of its ability to extract features internally by manipulating the system hyperparameters or even tuning them.…”

Development of intelligent protection and automation control systems using fuzzy logic elements

“…A literature review has shown that combining multiple functions in one microprocessor terminal dramatically reduces the reliability of protection, since if this terminal fails, too many functions will be lost at once compared to the case when these functions are distributed among several terminals [6]- [9], the need to limit the number of functions implemented in one terminal was also discussed in their reports by scientists at the international conference "modern trends in the development of protection systems and automation of power systems" [10]- [12], a sharp increase in the level of complexity of the work of personnel servicing protection with the transition to the MPD, as the cause of severe accidents in power systems [13], [14], the complexity of the software interface and the need to introduce an excessive number of set points during the programming of the MPR [15]- [17], the unsatisfactory state of the electromagnetic environment at most old substations, which were designed and built for electromechanical protection [18]- [20], and not for microprocessor protection [21], and about the numerous failures in the operation of the MPD arising from this [22], the lack of universal strict requirements for the hardware of the MPD and software, and as a result [23], too many programs and algorithms embedded in the MPD used in one power system, which leads to problems during operation and to increase the probability of false operation of these devices. Most of these devices must be installed in heated switchgear, since at temperatures below -25 degrees, the liquid crystal display with which microprocessor devices are equipped becomes unreadable.…”