The challenge of accurate renewable energy load forecasting may be helped by utilising Artificial Intelligence, writes Sid Sachdeva.
The shift from fossil fuels to renewable energy sources comes as welcome news for the environment.
While plants powered by oil, coal and gas hasten climate change with their carbon emissions, wind turbines and solar panels have proven a reliable source of non-polluting electricity.
This article was originally published in Smart Energy International issue 1 – 2020. Read the full digimag here or subscribe to receive a print copy here.
But the shift to renewables has also made it that much harder for utilities to predict how much energy they need to supply the grid at any given time, as increases in supply from renewable sources fail to line up with periods of high demand. The result: billions of dollars wasted on generating and purchasing unnecessary power.
The time has come for utilities and retail energy providers (REPs) to update their load forecasting practices to reflect today’s rapidly changing energy landscape. By drawing on the customer usage data supplied by millions of smart meters — and then analysing it with artificial intelligence and machine learning — power planners can generate forecasts that save money while ensuring that customers get the electricity they need for their homes, businesses and electric vehicles (EVs).
Renewables now a mainstay
In the push to end the pollution that speeds climate change, wind and solar power have graduated from their status as energy alternatives to become mainstays of electrical generation.
The burning of oil, coal and natural gas to produce electricity and heat is the largest single source of global greenhouse gas emissions, according to the federal Environmental Protection Agency. In an effort to reduce that impact, government officials, industry leaders and concerned consumers have begun a massive reset of the energy mix.
Take Texas, for instance. The Lone Star State both produces and consumes more energy than any other state, according to federal statistics. But it is also leading the way toward a renewables-first power mix, adding more wind and solar capacity than any state last year, according to Fred Beach of the University of Texas-Austin’s Energy Institute.
In California, meanwhile, a state building code taking effect in 2020 will require all new homes and apartment buildings to be “net zero,” meaning their on-site solar panels produce enough power to offset their annual electricity use for cooling, appliances and lighting. And while about one-third of the state’s electricity now comes from renewable sources, that percentage is required to increase to 60% in 2030 and 100% by 2045.
All over America, roadways will soon be full of electric vehicles, which will in turn create a huge demand for power at home and public charging stations. According to the Edison Institute, EVs will account for 20% of all new vehicles sold by 2030.
Such shifts — happening across the country and around the world — are testing the ability of power suppliers to adjust their plans to accommodate the intermittent quality of renewable generation, and to avoid wasting money on producing power their customers end up not needing.
Poor planning could cost power suppliers and their customers billions, according to the Rocky Mountain Institute. A decade-long analysis found that planners overestimated electricity demand by one percentage point per year – so forecasts 10 years out were on average 10% too high.
Using AI to build better forecasts
When it comes to renewable energy, time matters. Time of day, that is.
Solar power is produced only during daylight hours and, depending on their location, many wind turbines also run more strongly when the sun is up. That means that while renewable power is abundant — and therefore, inexpensive — during the day, it can be in limited supply during the evening hours when people return home and crank up their air conditioners, electronics and lights.
Nowhere is the financial impact of that supply/ demand imbalance greater than in Texas, the nation’s first fully deregulated energy market.
According to a recent federal report on the impact of renewables on wholesale energy costs, prices throughout ERCOT (which covers most of the state) ranged from $10 per MWh during high-supply periods to $80 during the evening — an eightfold variation.
Making planning even more difficult, the growth of “behind-the-meter,” customer installed solar has introduced more unpredictability into the power mix. While efforts are proceeding to improve storage methods, individual customers are still routinely producing more electricity than they — or their retail providers — need during the day. But at night, utilities are still firing up expensive and polluting fossil fuel plants.
The time has come to trade traditional forecasting models — which predicted energy demand based on a top-down look at decades of usage information — for a bottom-up approach that more accurately reflects the dramatic changes brought by renewables. Doing so would enable power suppliers to plan effective ways to store renewable power for high-demand periods, offer customers conservation plans attuned to their particular needs, and minimise the use of fossil fuels.
Writing in Forbes, sustainability advocate Georg Kell called for using data to speed an Earth-friendly energy transformation. Artificial intelligence “is poised to revolutionize the way we produce, transmit and consume energy,” added Harvard’s Franklin Wolfe.
In more than 700 million homes and businesses around the world, smart meters provide a steady supply of data that provides an unprecedented and detailed look at how customers are using — and generating — power today.
In our experience, companies that combine a bottom-up analysis of that data with AI and machine learning develop load forecasts that are 40% to 60% more accurate than older models.
That has the power to help save not only billions of dollars, but also the planet.
About Sid Sachdeva
Sid Sachdeva is the founder and CEO of Innowatts, a leading provider of AMI-enabled predictive analytics and AI-based solutions for utilities, energy retailers and smart energy communities. Sid was previously the director of new technology at NRG Energy and Reliant Energy.
Source: Smart Energy International
