There are many problems where we understand how a system evolves through time if we know the value of state variables, but those variables may be unobservable. Kalman filtering is a method for refining knowledge about the state using new values of a variable that we can observe. It has numerous applications from navigation to biomedical engineering to electrical power systems, and it's also a basic data analysis tool in econometrics. However because it's so widely applied and has several variants, it can be hard to grasp the fundamentals.

This talk will explain the basics of the Kalman filter and then show how to set up computations using the implementation of state space time-series models in the statsmodels Python library. We'll start with basic examples and progress to an application of forecasting electricity demand. This talk will only assume knowledge of basic Python programming and some background in probability and statistics.