Weather-Informed Forecasting for Time Series Optimal Power Flow of Transmission Systems With Large Renewable Share

Abstract

Planning and operational systemwide analyses relying on a deterministic optimal power flow (OPF) are subject to the uncertainties associated with the stochasticity of non-conventional renewables such as solar and wind. In this context, forecasting techniques will be essential for future power grids. An accurate and reliable forecast can complement a conventional time series optimal power flow (TSOPF) formulation by tackling uncertainty via accurate estimates of the intermittent generation resources. By doing so, the need for complex forecasting models is bypassed, and the computational tractability of the TSOPF problem is significantly reduced. Our work introduces a method to produce realistic load profiles, together with solar and wind generation forecasts. The generated data records are then input to a conventional TSOPF problem to determine the impact of an increasing share of non-conventional renewables into the grid over five years; we analyze increased penetration levels of 12%, 20%, 25%, and 30%. Our case study corresponds to an adapted version of the IEEE 39 test case, tailored to match the ISONE region transmission system. Forecasts are made using a data-driven neural network technique that accounts for the weather conditions in distinct locations of the ISONE footprint. Our results show several benefits of increased penetration of renewables, such as reducing line overloading and preventing undervoltage scenarios. However, it also points out potential problems that system planners might have to address by expanding the system or commissioning new equipment in particular locations.

Keywords

• Convolutional neural networks
• time series optimal power flow
• weather-informed renewable energy forecasting
• long short-term memory units
• synthetic load profiles
• seasonal profiles