I’m currently in my final semester at BITS and this semester is dedicated to a dissertation. I have to present a dissertation on any data science topic but it should be a valid use case from my employing organisation (Itron). The professors and management at BITS say that this degree is done while working at an organization so the dissertation should help the organization.

So I talked to my manager and I was given a theme around electrical load forecasting and I’m part of the same team. I talked to a person who is deep into research and development of load forecasting in Itron. He gave me an interesting use case to apply load forecasting.

Solar photovoltaic cells have gained popularity in consumer markets and many households are installing such solar PV cell on their rooftops. These solar PV cells are connected behind-the-meter (BTM), which means the power generated by the solar PV cells is not recorded. The only way we can detect the presence of solar power is by observing fluctuations in load demand during the day or by observing an unusual increase in load demand during day-time. This happens because the amount of electricity generated by the solar PV cells is dependent on the amount and duration of sunlight, which changes because of meteorological factors like cloud cover, temperature etc.

The main idea behind my dissertation is to use real-time meter load data (which only records the amount of electric load delivered by the grid) and predict the short-term load demand. This project is unique as we need to forecast the load demand while there is a hidden (latent) variable, amount of electricity generated by solar PV cells.

I have made some prelimnary research on load forecasting. There are many articles and papers surrounding load forecasting even in the presence of BTM solar PV. University students are tackling the same problem at many geographical locations, and each location presents a new set of challenges. I hope I’m able to contribute to Itron and complete the project.

Diary Log Entries