PowerPricePH
Your trusted source of electricity spot price forecasts
What is PowerPricePH
Designed for the UC Berkeley MIDS Capstone course, PowerPricePH is a tool that uses Machine Learning models to forecast electricity spot price based on public data.
Why PowerPricePH
The Philippines has the most expensive electricity in Asia, on top of unfavorable exchange rate and rising general prices. Similar to a stock market, the Wholesale Electricity Spot Market (WESM) started in 2006 in the Luzon region and subsequently extended to the rest of the country. It allows WESM participants to trade electricity. As of August 2022, there were 299 market participants. Half are generation companies with registered capacity of 22 MW with a value of PhP160 million (USD 3 million) per hour at the average PhP7.26 (USD 0.132) per kilowatt-hour.
Forecasting is essential for producers to better plan production and maintenance, and for retailers to better plan consumption and cost savings. Interviews with key stakeholders revealed that the main players currently pay at least $100,000/annual subscription, and some mid-size producer ~$200,000 this year for forecasting services.
PowerPricePH aimed to leverage the last Machine Learning models to compete with the best forecasting services for electric forecast services in Philippines.
​Our interviews show that
-
There is demand for more granular data from
-
both major and middle players
-
both producers and consumers of electricity
-
-
Firms currently pay at least $100,000/annual subscription (in addition to their inhouse forecasting costs, if any)
-
One mid-size producer is paying ~$200,000 this year for forecasting service
-
-
Beating the error threshold would allow to capture market shares
Trading Manager in one of the largest power generation company
“Difficult to plan without price forecasts... For trading, it helps us strategize to optimize margins, and for operations, for planning plant dispatch and maintenance…"
​Assistant Vice President of an independent power producer (IPP) operating coal-fired power plants
"Reliable price forecasts are important not only for our daily spot trading but also for budgeting… when we project revenues or income; assess the viability of bilateral contracts vis-a-vis WESM transactions. Tried in-house development, but we lack skills and computing power… and costly to maintain.”
Head of business intelligence and data science for one of the largest power distribution company
“We use price forecast to determine if we are getting the right price bids for the load we need.”
What we learned from this project?
-
Even with Machine Learning technology and sector expertise, it is difficult to beat the market in this mature sector when access is limited to public data
-
Power industry is complex: access to an electricity price trading expert would have improved the model design
-
Minimal Viable Product and testing were crucial to understand better:
-
If a product is doable (efficacy)
-
Where investment is needed (data, API)
-
How to invest stepwise: first data then API
-
Who we are
We are a multicultural team of UC Berkeley students with diverse backgrounds in economics, logistics and data.
