In the modern climate of global warming, rising energy costs and carbon trading, measurement and optimisation of resource usage is becoming increasingly important. One example is the management of building energy usage. Buildings account for approximately 1/5 of the world's energy use, but there is very little control or optimisation to minimise that use.
Recently, I have been working on machine learning techniques to predict building energy usage from dependent parameters such as the weather. This is very useful for predictive control and diagnostics. Studies have shown that simple feedback of energy consumption typically results in significant savings (~5-15%). With intelligent analysis of that consumption, this figure could be even higher.