Remmy195/PriceTaker-StochasticDualDynamicProgramming

Introduction

Companion code to Remi Akinwonmi's 2023 Master's Thesis "A stochastic dual dynamic programming approach to operational scheduling of long duration energy storage systems." This is a price-taker model that incorporates stochastic elements into the price variable.

READ SDDP.jl documentation BEFORE RUNNING THE SCRIPTS!

Getting Started

Prerequisites:
- These simulations were developed in Julia. Download Julia.
- These packages are needed to be installed in your Julia environment:
  - CSV
  - DataFrames
  - Gurobi # Note: Gurobi is a commercial solver and may require a license.
  - Plots
  - SDDP
  - Clustering
  - StatsPlots
  - Distributions

Scripts Overview

Stochastic_Dual_Dynamic_Programming.jl: Implements SDDP; performs in-sample and out-of-sample policy evaluations.
Stochastic_Dual_Dynamic_Programming_RH.jl: Implements a rolling horizon strategy with different levels of foresight.
Monte_Carlo_Optimal_Contol.jl: Monte Carlo simulation of the deterministic optimal control model.
SDDP.bat: This batch file runs the Monte Carlo simulation and the SDDP script sequentially.
SDDP.sh: This shell script runs the Monte Carlo simulation and the SDDP script sequentially.

NOTE:

To run Stochastic_Dual_Dynamic_Programming.jl, set parameters and the distribution of the random variable.
The initial Bound on the objective value is estimated by running Monte_Carlo_Simulation.jl for many scenarios and using the mean and confidence interval of the objective value to calibrate the bound.
Stochastic_Dual_Dynamic_Programming.jl performs In-Sample and Out-of-Sample evaluations of the SDDP policy.
Set the type of policy evaluation.
For out_of_sample_historic simulation, provide a vector of prices.
All simulation scripts draw data from SDDP_Hourly.csv.

Running the Batch Script

On Windows, run the SDDP.bat file
On Linux/Mac, run the SDDP.sh file
- Either will execute the Monte Carlo simulation, which estimates the objective function bound (The bound is stochastic). Then, it runs the SDDP script with the estimated bound.
Manual Computation of Bounds
If you prefer to set the bound manually instead of computing the Monte Carlo simulation, you can do so by modifying the SDDP script:
- Find the section where the script reads the bounds from file and comment out this section to prevent the script from reading the file like below.
```
# Read the bounds from the file
#bounds_file = "calculated_bounds.txt"
#bounds = Float64[]
#if isfile(bounds_file)
#    for line in readlines(bounds_file)
#        push!(bounds, parse(Float64, line))
#    end
#end
```
- Uncomment the line where bounds are set manually. Adjust the values as deemed fit.
```
bounds = [35_000_000, 40_000_000, 50_000_000, 50_000_000, 70_000_000]
```
- Save the changes and run only the SDDP script.

Documentation

Detailed information can be found in the SDDP.jl documentation.

License

This project is under the MIT License.

Contact

[Remi Akinwonmi / reak4480@colorado.edu]

Remmy195 / PriceTaker-StochasticDualDynamicProgramming