Skip to main content
Electrification Software
Customer story

Exelon Optimizes Wind Forecasting Accuracy

18 seconds

Forecasting 4 wind farms in 18 seconds

70 percent

Performance increase

Introduction

Products

Proficy Manufacturing Data Cloud

Background

As a leading utility company with more than $31 billion in global revenues in 2016 and over 32 gigawatts (GW) of total generation, Exelon was developing a strategic plan to enable digital transformation across its business lines. Specifically, Exelon was developing strategies for managing its various generation assets across nuclear, fossil fuels, wind, hydro, and solar power. In addition, the company was also determining how it would leverage the enormous amount of data those assets would generate going forward.

 

Challenges

In evaluating its strategies, Exelon reviewed its current on-premises IT/OT infrastructure across its entire energy portfolio. The company’s business leaders looked at the system administration challenges and costs they would face to maintain the current infrastructure. This assessment made digital transformation an even greater imperative and inspired discussions about how Exelon could leverage a combination of edge and cloud technology, rather than expanding or upgrading its existing on-premises infrastructure.

 

Additionally, Exelon had accumulated decades of operating and organizational data that it had been unable to leverage because it was either siloed or the technology, processes, and expertise to analyze it were not available. The company realized that they needed to bring data, people, and analytics together to improve business outcomes. Further, they needed to solve these challenges at scale.

 

Exelon knew they needed a shared, centralized approach to managing data, analytics, and solutions so that they could implement their strategy across the company. The individual power generation companies within Exelon each have separate IT departments, and the company had already experienced pain points from the mix of architectures and technology stacks in place at each generation company’s data centers. Going forward, Exelon saw a need to centralize, standardize, and lower costs.

 

Solutions

In reviewing its challenges and goals, Exelon found that GE offered tremendous synergy with Exelon’s technology and business imperatives. Further, Exelon found that GE’s renewables and power solutions, built on Predix Platform, address the company’s business imperatives in the areas of asset management, outage excellence, equipment reliability, and Operational Excellence. Exelon quickly realized that GE’s co-innovation approach and Predix Platform’s roadmap were aligned with its long-term needs.

 

Exelon and GE began working together on ways to achieve the company’s goals using Predix Platform. The companies held a series of technical and visionary collaboration sessions that resulted in selecting five projects to quickly demonstrate the value of using Predix Platform to leverage the right data to drive business outcomes.

 

One of these projects was finding a new wind forecasting solution. A key challenge with monetizing wind power is that it is difficult to forecast ramp events—when there is excess capacity to sell for the power that will be generated. Forecasts are used by dispatchers, traders, and operations and maintenance planners in regulated and deregulated markets. An accurate forecast is key for all of these stakeholders. Furthermore, accurate forecasting requires the use of detailed meteorological data. This type of data can be expensive if it comes from private data providers.

 

Along with accurate forecasts, leveraging free data sources such as NOAA in the US can increase the profitability of wind farms. In its current state, Exelon’s wind forecasts were not responsive enough to predict wind ramp events. When wind farms had more capacity, they couldn’t dispatch the additional power in the Midwest Independent System Operator (MISO) market because they could not anticipate quickly enough when power was going to be available.

 

Exelon and GE took a co-innovation approach to creating a solution. Exelon had access to many of GE’s services—such as data science, architecture, and software engineering—during the solution development process. Exelon gave GE a year of historical data, enabling its data scientists to understand any diurnal and seasonal effects that would impact the forecasts, as well as specific operating characteristics for the four initial wind farms. With that information in hand, GE’s data scientists were tasked with reducing under-forecasting.

 

GE’s data science team created a net new model—a physical and statistical wind power forecast model based on historical data provided by Exelon. They incorporated diverse data sources and took into account seasonal or time-of-day effects. The result was an industry-leading forecasting solution as measured by a substantial reduction in under-forecasting.

 

The team was able to increase their time-to-market by using the powerful analytics capabilities and higher-level analytics languages available on Predix Platform. For wind forecasting, the analytics were coded in Python, including all the forecasts and data quality analytics. Data quality is essential for forecasting applications, so the application needed to clean and validate the data prior to use, particularly because upstream systems, such as data historians, do not typically address data quality.

 

Results

Since Predix Platform’s microservice architecture enables horizontal scalability, the wind forecasting application was able to support wider rollout, including additional wind farms. Not only did the application perform substantially better than Exelon’s stated SLA, it also supported the forecasts for four wind farms, consumed the data from the data historian, ran the analytics in Predix Cloud, and wrote back the results in just 18 seconds—creating a 70% performance increase.

 

While the application is tuned to reduce under-forecasting in real time, it also improved other forecasts as well. Day-ahead generation relative forecast accuracy improved by 9%, which better enabled outage planning so that maintenance can be scheduled when there is little to no wind, minimizing revenue loss.

 

While the wind forecasting project was just one of five initial projects across Exelon, the success of these projects have resulted in a long-term enterprise agreement with GE so that Exelon continue to partner with GE to develop robust industrial applications on Predix Platform.

 

Download the technical story.

Explore GE Digital's Predix Platform

Industrial app developer using Predix Platform from GE Digital on mobile device
IIoT Platform

Predix Platform

Secure and scalable IoT cloud foundation for GE Digital applications