Electricity System Planning with Distributed Energy Resources



Electricity System Planning with Distributed Energy Resources
by Jesse Jenkins (PhD Thesis, MIT Institute for Data, Systems, and Society)

Main Points:
  • Who:  Dr. Jesse Jenkins is a recent graduate from the Massachusetts Institute of Technology, where he earned an MS in Technology & Policy, and a PhD in Engineering Systems. He worked as a researcher in the MIT Energy Initiative research consortium studying the evolution of the delivery of electricity services and the role of distributed energy resources in the power system of the Future. He is an NSF graduate Research Fellow and was the former Director of the Breakthrough Institute's Energy and Climate Program. 
  • What: This PhD Thesis offers a new framework for electricity system capacity planning with distributed energy resources and flexible and price-responsive demand. Dr. Jenkins first uses AC power flow simulations to derive equations on resistive losses and net demand reduction. He then uses these equations by plugging them into a capacity planning model that tells him how much of each resource (CCGT, solar, wind) a city needs to have in order to minimize costs, while meeting constraints like thermal ratings/ voltage limits, CO2 caps, transfer capacity, etc. He then simulates this on a fictional environment. Key takeaways? (1) By making 3% of electricity demand price responsive at peak hours, you can eliminate the need for DER’s entirely in new capacity planning (2) carbon prices don’t really impact the locational value of DER’s (3) As more DER’s are located in the same part of the grid their value decreases 2x the rate of power injection. 
  • Where: Enerlandia- a fictional power system modeled on the NYISO. This cased is a mixed integer linear programing problem with 5 transmission zones, 10 distribution zones, 61 eligible resources- 21 at transmission voltage levels of which 12 are thermal units subject to unit commitment constraints and 40 at distribution levels and 8,760 hourly time intervals. 
  • How: In order to solve the linear program for the GenX Capacity planning model, he needed to solve a problem with 2.67 million rows, 3.17 million columns, and 16.63 million nonzero elements, a problem that took 246.54 minutes (approx. 4 hours, 6 mins) using Gurobi’s concurrent solution method! He used the Engaging Supercomputing Cluster located at Massachusetts Green High Performance Computing Center. Each case is run on a single node making use of one Intel XEON E5 Haswell-EP Cpu with 2.1 GHz, 16 cores, and 3.1 GB of memory per core and solved with Gurobi v7.5.0 Solutions within a 1 percent MIP gap. 
  • Limitations of Work: (1) Simulations were performed on network representative of European networks, not those found in the United Satates (2) Case studies use single distribution networks, not a large number of randomly sampled networks across regional power systems (3) Capacity planning model was on Enerlandia which is representative of the smallest ISO in the US, gives good qualitative data, but not full quantitative insights.
  • Main Equations:

Comments

Popular posts from this blog

Consider a Spherical Cow

Should We Risk It?

Read one Book a Week