Evaluating solar PV effects on California's hydropower generation with a
hybrid LP-NLP optimization model
A hybrid Linear Programming (LP) and Nonlinear Programming (NLP)
optimization model is developed for California’s hydropower operations.
Built on top of Pyomo library, a high optimization modeling language in
Python, the model can connect to several freely available,
state-of-the-art solvers. In this model, fast evaluation of LP and
detailed model representation of NLP are fully utilized. The hybrid
model solves the same problem with linear approximation (a simplified
objective function representation) and with NLP solver, where no
simplification is made to objective function. Outputs from LP model are
used as initial values (warmstart) for NLP model’s decision variables,
which reduce number of iterations for convergence and so runtime. The
model is capable of representing large network of hydropower plants that
are in serial or parallel, or fixed and variable head plants. The model
is used to evaluate effects of increased solar photovoltaic (PV)
generation in California. California has a goal of generating
electricity from renewable resources at least 33% by 2020, and 50% by
2030, and solar PV generation supplies most of renewable generation
portfolio during daytime. This expanded use of solar PV changes
generation pattern from one daily peak system to two daily peak system.
Due to excess generation of solar PV, negative prices can occur during
daytime. Therefore, evaluating effects of solar PV on hydropower
operations and adapting to new conditions are essential.