facebookresearch / carbonexplorer Goto Github PK

Make the extra capacity range starting from 0%.

CAS algorithm is shared across the cells in section 4. Extract out the algorithm and deduplicate the code.

Data download and pre-process sections are duplicated in the two notebooks

Extract out the battery charge/discharge algorithm out from the notebook into a separate file to deduplicate the code and make it easier for it to be configured.

In the final total carbon footprint section, expand the output to include renewable coverage for all solutions and produce a better report.

Add time zone label (UTC) to plots (EIA_Energy_Data_Analysis yearly analysis plots) to make it more intuitive

Instead of sweeping a fixed range or parameters for battery capacity and scheduling capacity, implement a search algorithm (i.e. binary search) to find the optimal point faster.

Current CAS algorithm simply uses a flexible workload ratio as a % of workloads at every hour. We can improve the model using more realistic workload traces with SLAs, durations etc.

Current renewable energy generation projections (from a specified wind farm i.e. 500MW) are done by scaling EIA's BA (balancing authority) level data into the corresponding amount. However BAs are spanning a wide ranges and on a localized level the fluctuations might be different than the BA level fluctuations (at the BA level fluctuations might be getting averaged out).

The projections can be improved to make it more localized or adding some realistic perturbation.