There are two main goals to this project: 1) develop a risk index based on historical data to qualify the meteorological events by intensity according to the characteristics of the infrastructure; 2) use this index to improve the management on daily operations to mitigate the potential outages caused by natural disasters (meteorological causes).
The methodology considers evaluate the historical data of the last 5 years of many variables: operational information, outage (cause and duration), impacted area, distribution lines infrastructure, meteorological conditions and characteristics management. The CPFL serves 9.1 millions of users and 691 municipalities distributed on South and Southeast of Brazil.
This area is exposed to many several meteorological impacts caused by cold fronts, cyclones, thunderstorms, squall lines, downbursts and tornados. In the other hand, this area doesn’t have an adequate monitoring system. Only in São Paulo state, the area contains a radar and a weather stations network, but in the South area (Rio Grande do Sul), the conditions are precarious and strongly dependent on the satellite information (30 minutes delay) and lightning information (5 minutes delay). Because of that, is as huge challenge improving the nowcasting and the forecast to enhance the reliability and the resilience of the power system.
In Brazilian law, the causes of outages are classified according rules defined by ANEEL. This project will consider outages associated with meteorological causes: wind (tree falling), floods, fires, lightings, heavy rain, etc. Analyzing the data sample to the north area of Rio Grande do Sul, it is possible to verify the main cause of outage is associated with vegetation and wind.
To improve the forecast quality using WRF model (5.0 km resolution) will be applied a post processing technique to adjust the wind speed distribution based on the CDF matching. This technique was adapted from Zhu1 (2015) and adjusts the wind speed distribution based in a recent historical distribution. This technique was applied to both: direction and speed. In the Figure 3 is possible observe one example of this technical application.
To correlate all variables and develop the meteorological risk index this research uses two approaches: vector similarity and rough sets to develop classifiers. The initial results indicated the importance to include the characteristics of the infrastructure to classify the risk index associated to a several meteorological events.