Add a model to forecast trends in the data.

[Pmoloney3415]

• Perhaps implement Bayesian analysis to forecast the trends in the price of energy for the next day.
  • Perhaps separately forecast the baseline price from the median of the previous day.
  • Also perhaps try to forecast the height of the spike the next day, perhaps trying to pick out seasonal trend in price spike.
  • This can then be implemented in the reward function.
  • Using a Bayesian analysis should be able to give an uncertainty on the prediction as well, which could also be incorporated into the reward function.