Optimising operations: Artificial Intelligence for Renewables
Power production from renewable sources has never been a simple business. Trying to account for weather conditions or seasonal water levels means that there are always multiple challenges standing in the way of getting the most value out of your project. This blogpost explains how Montel AI uses technology to make power outputs more predictable.
With so many variables in play, forecasting can prove tricky. However, Artificial Intelligence (AI) is already finding new ways to predict power output and solve these problems.
One such method of AI able to do this is machine learning. Put simply, this uses existing data to run a series of simulations, which produce new forecasts based on previous results.
As underlying fundamentals are already included in the production data, machine learning only requires historical production outputs and upcoming weather data (provided by SMHI) to begin making new predictions.
More specifically, Montel AI uses a method of machine learning known as Artificial Neural Networks (ANNs) that sends inputs (historical and real-time weather data and power production levels) through multiple layers of mathematical functions in our advanced model to produce a forecast.
ANNs are very good at finding patterns, so once those functions have tested the input data across multiple parameters, they can begin to recognise the expected outputs in each scenario.
This ultimately helps renewable plant owners to reduce their exposure to imbalance prices because actual production outturns closer to forecasted levels.
Increasing efficiencies in areas such as this are important to help plant owners ensure they are maximising the value of projects as renewable energy support schemes fall away, not to mention the way in which imbalance pricing seems likely to become more punitive as renewables begin to dominate energy grids across Europe.
Montel AI also helps to lower administrative costs where the software is integrated with third-party reporting platforms. Where Day-Ahead predictions are required for spot exchanges for example, forecasts can be automatically fed-in, saving you both time and money.
Forecasts can be provided up to seven days ahead of time, as often as every hour. Using a Secure File Transfer Protocol (SFTP), automatically renewed forecasts are available when you upload new data, making your predictions ever more useful.
You can even see regular performance reports which help you assess how our predictions perform against the actual figures.
All the model requires is historical production and weather data, so whether your renewable generation is hydro, wind or solar, it's simple to start. Contact us and we'll build you a free trial model for your project - find out how on our renewable forecasts page.
This article by Montel first appeared in the British Hydropower Association's newsletter, Spotlight.
Blogpost by
Simon White
Content manager