Creating Value for Rural Electric Communities in Michigan

Team: Zihao Zhang (NE), Liwei Chen (CEE), Elizabeth Foster (ME)

Advisors: Seth Hoedl (Post Road Foundation), Gabriel Gomes (ME)

Small electric utilities often lack the sensor infrastructure to monitor their load effectively. This lack of data hinders their ability to take advantage of new distributed energy resources that can improve grid resiliency and reduce peak demand. The Post Road Foundation team developed an electrical load forecasting model that uses limited data sets to accurately predict load across a year for rural electric utilities, which can eventually be used to optimize the deployment of distributed energy resources.

Motivation

The Post Road Foundation brings clean energy to rural communities by incentivizing them with widespread broadband connection, which enhances a community’s ability to be economically viable in today’s digital world. We partnered with Presque Isle Electric & Gas, a rural co-op in Michigan to deliver a plan for distributed energy resource implementation.

Rural Electric Cooperatives

Electric Cooperatives (“Co-Ops”) were founded to serve the needs of rural communities. Generally, they have limited resources and aging infrastructure, which makes it difficult to accommodate new forms of energy generation and storage.

Solar Energy Analysis

Solar is only viable for Commercial or Industrial customers based on similar installations and Department of Energy analysis tools. Residential solar is not viable because of MI’s low electricity rates and low solar irradiance.

Wind Energy Analysis

Wind is not viable at a distributed scale for rural Mi due to inconsistent wind patterns and low electricity buyback rates.

Electric Load Forecasting with Machine Learning

Load forecasting is key for Co-Ops to plan future projects and optimize distributed energy resources. Presque Isle Electric & Gas lacked sufficient data for optimization, so the team sought to develop a synthetic load profile through forecasting.

A Long Short Term Memory (LSTM) Recurrent Neural Network (RNN) was built to synthesize the load profile using a very limited, but high fidelity data set (3 months). The model showed promise in forecasting short-term demand, but still needs to be modified for long term predicting.

Next Steps

In the future, the Post Road Foundation can utilize the predicted load data to optimize solar arrays and battery storage for commercial and industrial customers. Not only will the customers benefit from reduced peak demand, access to dedicated backup power, and increased electric grid utilization, but the community as a whole will also benefit from cleaner energy and infrastructure investments.

Related News: