Analysis of the influence of 110 kV power line parameters on the probability of their failures

When considering power supply system equipment, it should be noted that most often power supply interruptions occur due to damage to power lines. It is important for power grid companies to promptly diagnose faults and restore power supply in order to minimize losses. Therefore, determining in advance the probability of power outages based on the parameters of the power lines themselves can help power grid companies modernize much more efficiently and build new power grids with less failure rate. Aim of the article is analysis of the influence 110 kV power line parameters on the probability of their failures. This study consists of preparing a data on power outages of 110 kV power lines by by clustering, removing unnecessary information, creating synthetic parameters and processing them by imputing missing values and removing duplicates. Then exploratory data analysis was carried out, including analysis of statistical characteristics for considered parameters, identification of outliers and anomalies, and correlation analysis. The research was carried out in the Python programming language in the Jupyter Notebook development environment, using the Pandas, NumPy, Matplotlib, Seaborn, Phik libraries. The data for analysis was prepared using grouping and merging methods, the least significant parameters were removed, the “power line service life” parameter was standardized, and the target attribute was synthesized - “The fact of power line outage.” The final result was a table containing 10 parameters, including the target feature, and 395 rows. During the analysis of categorical parameters, an imbalance of classes of the target feature, the influence of the type of wire and the transit fact on 110 kV power line failure were identified. An analysis of the distribution of quantitative variables confirmed that a decrease in outage probability is observed with an improvement in technical condition of lines, a decrease in line length, in the number of supports and in the service life of power lines. Correlation analysis allows establishing the absence of a strong correlation with the target feature and the presence of multicollinearity between all parameters reflecting the length of power lines and the number of reinforced concrete supports.