PyUC: Stochastic unit commitment in python

Overview

This model includes:

• Clustered unit commitment for numerical efficiency [1]. Instead of representing eight coal units with eight binary variables, a single integer variable is used. [1]
• Ability to include thermal generation, variable renewables and storage/batteries.
• Comprehensive set of unit commitment constraints:
  • Minimum stable generation
  • Minimum up and down times
  • Ramp constraints
  • Reserve and inertia requirements
• Ability to include a look-ahead period to better optimise decisions.
• Stocashtic model allows simulating different scenarios of wind, solar and demand (or forecast errors). Commitment decisions (made ahead of time must be the same in all scenarios, but dispatch and storage profiles can differ).
• Policy costs (carbon price, renewable energy certificate price, etc).
• Determines energy and reserve prices (using constraint shadow prices).

This model was original written GAMS in support of a doctoral thesis to model energy-only markets with increasing renewable generation [2].

Planned features

• Generation of multiple uncertain wind, solar and demand scenarios using an auto-regressive moving average approach.
• Ability to run multiple days in series, using the end state (commitment status, stored energy, etc) of day n as the initial condition for day n+1.
• Generation expansion.

Requirements

• PuLP with CBC solver.
• Pandas
• Matplotlib
• Numpy

Getting started

Demonstration problems are available in the demo_problem directory, containing:

• half-hourly demand (demand.csv),
• wind and solar traces (variable_traces.csv),
• a range of unit types (unit_data.csv), including fuel costs and thermal efficiency, unit commitment constraints, battery properties, etc.

Constraints can be turned on/off using constraint_list.csv, and initial conditions (such as initial energy stored in the battery and number of units on or offline) is in initial_state.csv.

Deterministic problem

Use the following snippet, or run demo.py from the pyuc main directory.

from pyuc import pyuc
from visual import plots

input_data_path = "path/to/pyuc/demo_problem/deterministic"
output_data_path = "path/to/pyuc/demo_problem/deterministic"
name = "PyUC Demo"

pyuc.run_opt_problem(input_data_path, output_data_path, name=name)
plots.plot_dispatch(output_data_path, name=name, plot_price=True)

Results will be available under demo_problem/deterministic/results, and the dispatch chart (above) under demo_problem/deterministic/charts

Stochastic problem

To run a stochastic problem with three different demand traces (high, medium and low), replace "deterministic" with "stochastic" in the input_data_path and output_data_path in the snippet above, or use demo_stochastic.py.

There will now be three dispatch charts for the different demand scenarios. The commitment schedule is the same in all, but units can dispatch differently to meet the prevailing level of demand.

References

[1] Palmintier, Bryan Stephen. "Incorporating operational flexibility into electric generation planning: Impacts and methods for system design and policy analysis." PhD diss., Massachusetts Institute of Technology, 2013.

[2] Marshman, Daniel. "Performance of electricity markets & power plant investments in the transition to a low-carbon power system." The University of Melbourne (2018).

License

This project is licensed under the GNU General Public License. See the LICENSE file for details.