Accurate load forecasting could make or break the resiliency of solar energy in times of war and crisis

As the unexpected happens, accurate forecasting becomes a vital tool for energy providers and stakeholders to sustain grid operations. Amidst evolving demand dynamics, advanced AI and machine learning technologies can automatically forecast electricity generation and demand, turning mere data into actionable information.

As the disruptions of the Covid-19 lockdown and the conflicts in Ukraine and Israel demonstrated, predictive modelling helped solar energy providers and energy retailers swiftly adapt to drastic and sudden changes in demand.

Predictive modelling in practice

During these three distinct crisis situations, predictive software was instrumental in managing grid stability in times of extreme and sudden fluctuations in energy demand and supply. The data that follows was compiled and reported by Tigo Energy through the Predict+ platform, which is designed to provide accurate, scalable, and robust energy forecasts by augmenting consumption data at the smart meter level with external data sources to forecast grid demand accurately.

In 2020, the Covid-19 lockdowns had a severe impact on the energy grid in Israel, with unprecedented changes in energy consumption patterns and forecast deviations as high as 12%. With the bottom-up approach of at-the-meter monitoring and automated recalibration, the predictive modelling software quickly pinpointed hotels as the primary source of deviation. Without guests, the hospitality sector scaled energy use back to the bare minimum.

The industrial sectors, however, remained comparatively stable, highlighting the uneven impact of lockdowns across industry sectors. Energy providers used predictive modelling technology to adapt forecasts, creating a new ‘lockdown normal’ baseline maintaining model accuracy amidst fluctuating consumption dynamics with constant updates.

The sudden invasion of Ukraine by Russia provided predictive energy modelling yet another opportunity to help stabilise what quickly became an embattled grid. The AI-driven forecasting system played a vital role in keeping the nation’s grid stabilised, particularly during drastic disruptions from infrastructure bomb strikes.

By quickly adjusting output in response to sudden drops in electricity demand, the software helped prevent power outages and established it as an essential tool for grid management during wartime.

Also faced with the disruptions of armed conflict, the energy market in Israel experienced a significant shift in energy consumption patterns as of 7 October 2023. Nationally, a more than 10% decline in energy consumption could be attributed solely to the impact of the war.

At the time, the predictive modeling technology served seven out of nine Independent Power Producers (IPPs) and three major virtual energy suppliers in Israel. During the crisis, the team behind the software kept up a rigorous process to maintain the reliability of their forecasts, including pinpointing energy consumer clusters, introducing an ‘expert mode’ to facilitate new operator interventions, and characterising distinct ‘war-affected’ consumption patterns.

Forecast accuracy was restored to pre-war levels in just three days.

Combining AI and the bottom-up approach

The performance of this software in these three crises scenarios underscores the value of predictive platforms in their ability to quickly adapt to sudden and extreme changes in energy consumption behaviour through robust AI and machine learning technologies.

Particularly compelling, however, is the combination of predictive modeling and the bottom-up approach, starting at the smart meter level with automated recalibration. These features allow the system to quickly identify the primary sources of deviation. While the Tel Aviv airport shutdown and hospitality industry restrictions were known, the resulting changes in energy consumption were not. The system had to quickly adapt to a hotel industry that was suddenly limiting itself to only essential services, like refrigeration.

Beyond excelling in times of volatility, advanced predictive energy platforms also prove highly effective during significant consumption shifts in peacetime. For instance, when large industrial users undergo maintenance without prior notification, predictive modelling can swiftly adapt to maintain demand forecast accuracy.

The adaptability of this system underscores the important role of energy prediction systems in improving operational efficiency for energy stakeholders, driving sustainability across sectors, and their potential to save millions in imbalance costs.

The long-term prediction

Rapidly ingesting, integrating, and analysing extensive datasets, ranging from weather forecasts to historical consumption patterns, establishes predictive modelling software as essential enablers for solar generation generally, and handling unpredictable energy demand fluctuations during emergencies in particular.

This capability allows energy providers to make informed decisions to maintain grid stability, even amidst the most challenging of circumstances. The combination of precision, velocity, and adaptability is indispensable in today’s dynamic energy market. But more than that, predictive technology is now positioned as the ultimate tool for ensuring the viability of solar power resiliency on a global scale for years to come.

Evgeny Finkel is an entrepreneur and software engineer by trade, but currently serves as the Vice President of Business Development at solar technology manufacturer Tigo Energy, and prior to the acquisition by Tigo served as the CEO of FSIGHT, which delivers artificial intelligence solutions for distributed energy.