I'm a beginner in the Data Science field and I'm trying to find a way to solve this problem, but I'm not sure where to start.

Here's the situation: I have a battery that stores and discharges energy into my house from the power grid. Energy is produced by different generation techniques throughout the day and so I have data on the emission impact of using electricity at different hours. I also have data on the energy I've used in the past year. My goal is to charge and discharge this battery in the most effective way for reducing the emission impact of my home.

If anyone can point me in the right direction for what I should know and use to solve this problem I would be greatly appreciative!

  • $\begingroup$ Welcome to DataScienceSE. This looks like an interesting problem to me, but could you please give more detail about it? For example, if you were trying to find the solution manually, what would you do? what does a solution look like? how would you check that things have improved after trying a "solution"? $\endgroup$ – Erwan Jan 11 at 22:22
  • $\begingroup$ Hey thanks for the response! Trying to think about how I'd do this manually is something I've been struggling with. One Idea I had would be to set a certain carbon emission value as a border between buying and selling, set that as a function over my energy data, and then brute force the optimal values. At it's heart I believe it's a problem in optimization, but I'm not sure if there would also be some way to train the computer to find patterns in energy usage and then optimize daily charge/discharge times (since carbon emission data is given). $\endgroup$ – Jakob Koblinsky Jan 12 at 3:22
  • $\begingroup$ My apologies if I come off as very uninformed. I don't know anyone experienced to talk about this with, but given some direction I'm generally good at figuring things out. $\endgroup$ – Jakob Koblinsky Jan 12 at 3:24
  • $\begingroup$ No worries, I tried to propose some directions in my answer, hope it helps. I think at this stage you're the only one who can refine the formulation of the problem, like decomposing it into simpler pieces. Note that this question might get closed as unclear, but once you have reached a more specific design of the problem don't hesitate to ask a new question, it might be possible to give you more precise directions. $\endgroup$ – Erwan Jan 12 at 10:35
  • $\begingroup$ 1) Your problem can be formulated as constrained optimization problem: minimize the carbon footprint F for the particular day and ensure that the battery capacity C(t) is greater than your need N(t) for all t for this day. $\endgroup$ – aivanov Jan 12 at 22:34

I agree that it looks like an optimization problem, but the parameters are not completely clear. Here are some vague ideas I have about the design of such a system:

  • Let's assume a unit of time for which you have data and by which you attempt to predict things, for instance by the hour.
  • I think the first part needs to be about predicting energy needs by unit of time, based on past data and using any relevant factors as features (e.g. temperature, week day or not, etc.). I think this would be a reasonably standard time series regression problem.
  • The second part would be the optimization of energy consumption or carbon footprint: assuming we know the energy needs for a given period of time by unit of time, what are the optimal times for storing/discharging the battery? I don't know enough about the technical factors here, but I guess it's possible to simulate the target (say carbon footprint) based on different scenarios, and then select the optimal scenario. It seems to me that the problem at this stage could be simple enough to simulate all the scenarios and find the exact optimal times, that would probably correspond to a grid search. but in case this is too complex there are ML techniques which could estimate them, for instance genetic algorithms.

Not the answer you're looking for? Browse other questions tagged or ask your own question.