Background

Photovoltaic (PV) electricity generation, also known as solar farming, is an increasingly popular source of renewable, clean energy. PV farms produce varying amounts of energy based on a number of factors, especially including the local weather and temperature.

Due to the limited amount of grid-level energy storage available, the US electrical grid must closely match the energy being output with the energy being consumed at any given time. If unaccounted for, grid power imbalances can result in substantial lost energy at best, and can be highly damaging to grid systems at worst. Thus, the variable energy generation of PV farms complicates the job of electrical grid overseers.

Accurately forecasting PV power production and the ability to regulate the amount of power the PV farm puts onto the grid would substantially ease job of overseeing the electrical grid and make solar energy a far more attractive energy production option. Unfortunately, available PV forecasting methods suffer from the two major drawbacks of 1) substantial inaccuracy and 2) inability to forecast energy production on any higher resolution than hours at a time. However, to properly maintain grid stability, much finer forecast resolution is needed.

These drawbacks decrease PV farm stability and revenue, ultimately limiting our nation’s implementation of this clean, renewable, and safe source of energy.

Technology Overview

NCSU researchers have developed an improved method of forecasting PV power production with higher accuracy and resolution than the prior art, and further developed an improved method of regulating the power output of PV farms, thus preventing power fluctuations from harming grid infrastructure. Together, these software packages increase PV farm stability and revenue, while also easing the burden of grid maintainers.

The power output forecasting software uses a Temporal Convolutional Network, which examines the power output of neighboring farms in real time and combines this data with the known historical relationship between the neighboring farms to provide highly accurate forecasting with resolution on the minute-by-minute scale. This improved accuracy and resolution will provide a massive increase in controllable energy output and therefore economic value for the PV farmers as compared with available prior art. This system learns as it works, resulting in ever improving performance over time.

The power regulation method is a novel software program that tracks PV power outputs in real time and regulates the power to the desired set point. This software uses a flexible power-point tracking method modified with a novel Perturb and Observe technique to provide power convergence with the chosen set point faster than any other available prior art. This fast acting regulation is critical to safeguard fragile grid infrastructure and minimize energy losses.

Both of these software programs are fully written, validated, and are ready to be licensed to an industry partner. NCSU intends these two software programs to be licensed together as a package, but the university is open to discussing licensing a single program to the right partner.

Further Details

2022-089: https://arxiv.org/abs/2111.08809

2022-096: https://ieeexplore.ieee.org/abstract/document/9560073?casa_token=8A1qjpCkDOoAAAAA:fAUKkjtHxNLhSrQt2NbLWw471R–cN4r5HndYBcNWN95YnQhplueB2BqFALd_dPF_jTydoEZ4Q

Benefits

Solar farm improved stability.
PV farm improved power output forecast accuracy.
PV farm improved power output forecast resolution, down to the minute-by-minute scale.
Improved PV farm power output regulation.
Improved grid stability.
Reduced energy loss.
Improving attractiveness of clean and renewable energy.

Applications

PV farm administration.
PV farm forecasting.
Grid monitoring.
Grid maintenance.

Photovoltaic Farm Power Output Prediction and Regulation Software Package