Growth of Solar on PJM’s Horizon

4634

Consumers in the region served by PJM are taking an increasingly active role in their electricity purchase decisions. This trend is reflected in the rise of distributed energy resources, or DER – generation or energy storage resources connected at the local distribution level, often behind a customer’s electricity meter.

Nowhere is this growing popularity more evident than in the expansion of rooftop solar. Over the past decade, the biggest increase in solar generation in the PJM region has been in customer-side installations – generally, photovoltaic panels placed on the roofs of homes and businesses.

In 2009, there were virtually no solar resources on the PJM grid. Today, customer-side solar has the capacity to generate well over 4,000 MW of electricity, and grid-connected installations (such as the larger arrays you might find off a country road) provide more than 1,500 MW. (For context, 1 MW can power about 800 homes.)

Solar on the Rise

Solar generation represents a tiny fraction of the approximately 180,000 MW of capacity committed to serve the needs of 65 million people in the PJM footprint.

Expected growth of behind-the-meter solar in PJM.

But the demand for renewables is growing. PJM predicts that power from rooftop solar installations will swell to nearly 12,000 MW over the next 15 years. New wholesale solar projects planned through 2023 represent an additional 3,376 MW.

Of the top-10 solar states in the country, two are served by PJM: North Carolina (second) and New Jersey (seventh). The largest amount of planned wholesale solar generation in the PJM region is destined for Virginia (1,817 MW), North Carolina (634 MW) and Ohio (470 MW).

Ready for the Future

PJM is technology and fuel neutral. Its job is to maintain a reliable grid at the lowest reasonable cost, without favoring any one type of energy resource. In doing so, PJM is continually evolving to meet the changing needs of its customers and the industry.

“We don’t promote any one kind of generation over another,” said Joseph Mulhern, a senior engineer in the Generation Department. “Our mission is: Whatever mix of generation we end up having, we need to meet our responsibility of operating the grid reliably. In the case of increasing solar, that means taking steps such as forecasting output and enhancing visibility.”

Currently, PJM’s generating fuel portfolio is more diverse than it has ever been – in 2018, the grid was powered by about 29 percent coal, 34 percent nuclear and 31 percent natural gas. Recent studies of this breakdown, with an eye toward fuel security, indicate that the grid is as reliable as ever, and will remain so into the future.

The studies also show that there is room for many more renewables to come online, with significant growth expected in distributed energy resources.

Among the country’s transmission regions, PJM is uniquely large and geographically diverse. It spans two time zones, varying load patterns, and different areas rich in solar, onshore wind and offshore wind resources. This makes it easy to reliably and cheaply integrate large quantities of wind and solar resources, since variability in one part of PJM is offset by a different kind of variability in another area.

Seeing Behind the Meter

While some DER participate in wholesale markets, their output visible to PJM operators, the majority of which is solar, is not.

Non-wholesale generation is just that: Its output isn’t traded in the energy markets, nor can it be tracked in real time. But it’s critical to include these resources in the short- and long-term forecasts of energy demand.

When the sun sets or is shrouded by clouds, the gap in power can lead to steep ramp-up times for the PJM-dispatched resources that must step in and meet the demand.

Conversely, in the middle of the day, when the sun is shining brightest, the system could one day experience a glut of energy that otherwise would have value to the system.

The largest amount of planned wholesale solar generation in the PJM region is destined for Virginia (1,817 MW), North Carolina (634 MW) and Ohio (470 MW).

Ahead of the Curve

Mulhern said that the PJM system is just starting to experience the impact of changing patterns in demand, known as “load” in the industry.

“We’re ahead of the curve in terms of our planning,” he said.

Seeing the trend on the horizon, PJM created long-term and short-term forecasts for solar power usage that is not directly visible to PJM operators.

The long-term solar forecast looks at trends in solar power pricing and policies to predict how many solar power systems will be installed annually, extending out 15 years.

PJM’s short-term solar forecast predicts how much output distributed solar resources will generate every five minutes for the next six hours, and hourly for the next six days.

Last year, PJM tested three mathematical methods for integrating the short-term solar forecast into the short-term load forecast to account for the impact that non-wholesale solar has on load. The one that performed best has been implemented. It’s called the Reconstituted Load method, and it is similar to a method implemented previously in the long-term forecast processes.

Harnessing ‘Reconstituted Load’ for Better Forecasting

Traditionally, load forecasts have been largely temperature dependent, because so much of the demand for energy is related to the temperature.

Load that includes non-wholesale solar, however, doesn’t follow the same patterns as traditional load; it has its unique properties.

While PJM operators can’t see the real-time operations of customer-side solar systems that influence the load, they know where they are located, when they began operating and their installed capacity.

This public information lives in the Generation Attribute Tracking System (GATS) operated by a PJM subsidiary, Environmental Information Services. GATS tracks and records data from registered generators and creates an electronic certificate with a unique serial number. This system allows renewable generators to create and sell Renewable Energy Credits, which many states in PJM and the country require utilities to procure to meet renewable energy targets.

Instead of using historical, measured load to train PJM’s neural network load forecast models, as is done in the conventional process, the Reconstituted Load method uses estimates of historical total power used – what PJM’s utility-scale resources produce, plus estimates for behind-the-meter resources. This method then adjusts the resulting forecasts of total power used by subtracting forecasted non-wholesale solar output.

The estimates of historical behind-the-meter solar output come from “backcasts” that measure years of data about cloud cover and other weather factors.

The Reconstituted Load method is applied to several load forecast models generated for PJM dispatchers, who then adjust those automated smart composites using additional data and experience. “That process has a combination of manual and automated elements,” Mulhern said. The next step, he said, is evaluating the performance of Reconstituted Load models with the goal of automating more of that for the dispatcher.