Tuesday, November 4, 2014

Reliability (for Planning) and Reliability (for Forecasting)

My first job interview in the US was with Richard Brown. At that time, I knew virtually nothing about the electric power industry. To prepare for the interview, I checked Richard's profile, and found out that he is a Fellow of IEEE for his contribution in power systems reliability. I also browsed through his book Electric Power Distribution Reliability, a must-read book in power systems reliability. Although none of the things I prepared was actually used in the interview, I got to know about reliability before I even heard of load forecasting.

Reliability in the power distribution system typically means the ability the power system delivers power to the end users. There are three frequently used reliability indices in distribution planning:
• System Average Interruption Duration Index (SAIDI) = Sum of all customer interruption durations / Total number of customers served
• System Average Interruption Frequency Index (SAIFI) = Total number of customer interruptions / Total number of customers served
• Customer Average Interruption Duration Index (CAIDI) = Sum of all customer interruption durations / Total number of customer interruptions = SAIDI/SAIFI
There is also a rich literature on reliability planning, including developing other reliability indices, evaluating, benchmarking and enhancing power system reliability. The concept of reliability spreads across generation, transmission and distribution systems.

In load forecasting, or more specifically, probabilistic load forecasting, reliability has a completely different meaning. A probabilistic forecast is reliable (or calibrated) if the empirical coverage is exactly the same as the predicted probability. For instance, a predicted profile at 90th percentile should be on or above 90% of the observations. In other words, we expect 10% of the observations above this predicted profile. The property of reliability in probabilistic forecasting is similar to bias in point forecasting.

There are a few other places where reliability and forecasting are used together:
• A forecasting system is reliable if it functions as specified.
• Load forecasts is a key driver of reliability planning.
• We can forecast future value of reliability indices based on the weather conditions and investment in vegetarian management, etc.