battery_optimization

This repository contains the code for a battery optimization model that maximizes the profit obtained by charging and discharging the battery and trading the electricity across several electricity markets

Brief summary

For this exercise, I formulated the problem as a linear program. The objective function to maximize is the profit over a period of time (horizon). Profit is defined as revenue from discharging to the markets, minus the cost of charging from the markets and the cost of depreciation of capital as the battery deteriorates with charging and has a lifetime limit. Constraints are placed to ensure that charging and discharging does not exceed the maximum charging/discharging capacity and do not take place simultaneously. Constraints are also placed to ensure that the volume of the battery is not exceeded at any point in time, and that trading on the markets obey the trading time granularity. When objective function or constraints terms are non-linear, appropriate linear approximation is used for simplification. To set the period of time to optimize, I adopted a rolling horizon such that planning decisions are made using the price data for the next 72 hours, and the optimal plan computed is then executed up to the first 24 hours. The model is then re-run at the start of next day using the next 72 hour horizon. Experiments with longer planning horizon length of 144 hours yield negligible improvements but come at greater computational costs and are deemed unnecessary.

Structure of directory

|- data/
    |- battery_specs.csv
    |- daily_data.csv
    |- half_hourly_data.csv
|- output/
    |- output.csv
|- src/
    |- baseline_model.py
    |- battery_model.py
    |- data_processing.py
|- demo_notebook.ipynb
|- model_design.pdf
|- requirements.yml
|- run.py

data/ folder contains the csv files for battery specifications, and electricity prices from 2018-2020 of 2 half-hourly traded markets and 1 daily traded market
output/ folder contains the output of the simulation over the entire period in csv file. The file is generated as output of run.py.
src/ folder contains code for data processing and the battery model class
demo_notebook.ipynb is a jupyter notebook for interactive analysis
model_design.pdf contains information on how the battery model is designed
requirements.yml is the conda requirement file for environment setup
run.py is the main code to run the simulation

Running the code

Clone the repository

git clone https://github.com/zhoutianxun/battery_optimization.git
cd battery_optimization

Set up python environment with conda

conda env create -f requirements.yml
conda activate energy_modeling

To run the model over the time period provided in data/, with default settings:

python run.py

To change the settings of the simulation:

planning horizon length (time unit)
execution horizon length (time unit)
initial charge state of battery (MWh)
save path for simulation output

change the following arguments:

usage: run.py [-h] [--plan_horizon PLAN_HORIZON] [--exec_horizon EXEC_HORIZON] [--s_init S_INIT] [--out_path OUT_PATH]

options:
  -h, --help            show this help message and exit
  --plan_horizon PLAN_HORIZON, -p PLAN_HORIZON
                        planning horizon (default=144)
  --exec_horizon EXEC_HORIZON, -e EXEC_HORIZON
                        execultion horizon (default=48)
  --s_init S_INIT, -s S_INIT
                        initial state of charge of battery (default=0)
  --out_path OUT_PATH, -o OUT_PATH
                        path for saving output (default="output/output.csv")

To run the model in interactive manner, a jupyter notebook "demo_notebook.ipynb" is provided