We have designed a data recorder that indicates in 15 minute blocks whether a domestic electric storage water heater was used or not. The data is purely yes/no, so it might read 1,1,1,0,0,1,1,0,0,0,... where a one means it was used and a zero means it wasn't.
The heater wastes energy when it is on but not used, so we would like to use this data to forecast the next few values of the data series, perhaps with a confidence level (say the probability it will be used). We would then switch the water heater on only when it is reasonably likely to be used. We want the rules to be implementable in a C program so they would need to be simple.
It is quite likely the data will display fairly clear patterns based on time of day and day of week and it is also likely that there will be long periods periods of constant non-use at night say. So we could perhaps look for regularities based on daily and weekly cycles.
We are looking for a practical solution that works and is easy to implement - it need not be backed up by statistical theory if this is too difficult.
We could simulate some plausible data but don't yet have real data.
19 freelancers are bidding on average $144 for this job
Hello, I am a experienced SAS( Statistical Analysis System software) analyst. Solving of statistical & time series forecasting tasks is part of my professional activity.
Hi! I'm a Mechanical Engineer now studying MSc in Automation. Your project is very interesting, but will be challenging to implement. I would like to work with you till you find a good solution.