Abstract
Distributed energy resources (DERs) can serve as non-wire alternatives to capacity expansion by managing peak load to avoid or defer traditional expansion projects. In this paper, we study a planning problem that co-optimizes DERs investment and operation (e.g., energy efficiency, energy storage, demand response, solar photovoltaic) and the timing of capacity expansion. We formulate the problem as a large scale (in the order of millions of variables because we model the operation of DERs over a period of decades) non-convex optimization problem. Despite its non-convexities, we find its optimal solution by decomposing it using the Dantzig-Wolfe Decomposition Algorithm and solving a series of small linear problems. Finally, we present a real planning problem at the University of Washington Seattle Campus.
The Challenge of Peak Load
Electric distribution systems are traditionally designed to handle "peak load"—the rare hours each year when power demand reaches its maximum. As urban centers and campuses grow, the standard response is to reinforce physical infrastructure, such as feeders or substations. However, these "wires" solutions are capital-intensive, often costing tens of millions of dollars, and face significant hurdles including long lead times, legal challenges like eminent domain, and public resistance.
Non-Wire Alternatives (NWAs) offer a different path. By deploying Distributed Energy Resources (DERs)—including energy storage, solar photovoltaics (PV), energy efficiency (EE), and demand response (DR)—utilities can "shave" peak demand. The economic benefit lies in the time-value of money: by spending a smaller amount on DERs today, a utility can defer a massive capital expenditure for several years, allowing those funds to be utilized elsewhere in the interim.
A Scalable Optimization Framework
Determining the optimal mix and operation of NWAs is a complex mathematical challenge. It requires modeling investment decisions alongside hourly operations over a 20-year horizon, resulting in a non-convex optimization problem with millions of variables. This research addresses this "curse of dimensionality" through a three-part methodology:
- Holistic Formulation: A model that simultaneously optimizes asset procurement, hourly dispatch, and the specific timing of traditional capacity expansion.
- Convexification: The researchers proved that the non-convex costs associated with capacity expansion could be reformulated into a series of manageable convex sub-problems.
- Dantzig-Wolfe Decomposition (DWDA): The problem is broken into smaller chunks. Each technology (e.g., battery storage) is solved as a sub-problem, while a "Master Problem" coordinates these resources to meet grid requirements. This allows the use of off-the-shelf linear solvers for problems that would otherwise be computationally prohibitive.
Case Study: University of Washington Seattle Campus
The framework was validated using real-world data from the University of Washington (UW) Seattle Campus. The study used 2016 load data, a projected 3.5% annual growth rate (amounting to a 17 MW increase over a decade), and industry-standard cost parameters for various DERs.
Key Results: The Impact of Deferral
The optimization model demonstrated that a strategic mix of NWAs could significantly delay the need for physical infrastructure upgrades. For the UW campus, the model identified an optimal portfolio consisting of 8.5 MW of Solar PV, 6.7 MW of Demand Response, and 2 MW of Energy Efficiency.
As shown in Figure 1, the "business as usual" peak load would exceed the substation's 60 MW limit by year 9. By implementing the NWA portfolio, the peak load is managed such that the limit is not reached until year 14. This five-year deferral of a $60 million substation upgrade provides substantial economic relief. Beyond the primary goal of deferral, the NWAs reduced overall energy costs by 6% ($32 million) and demand charges by 15% ($5.8 million) over the 20-year study period.
Insights and Future Directions
The study highlights that the economic viability of NWAs is highly sensitive to technology costs and forecasts. For instance, lithium-ion storage only became cost-effective for capacity deferral in the UW case when prices dropped below $200/kWh. Furthermore, the optimal plan is more sensitive to solar irradiation and load growth projections than to discount rates or DER capital costs.
While the current model is deterministic, future work aims to incorporate uncertainty through stochastic optimization to account for the variability of renewable generation and load growth. Additionally, expanding the scope to include electric vehicle (EV) fleet orchestration could further enhance the flexibility and value of non-wire alternatives.
Summary updated