Digitization-supported power generation forecasts:

More precise data for safer operation at a lower cost

The use of technological tools is crucial to develop more accurate predictions, helping guide better decisions in the operation of the electrical system.

The variability of the intermittence of renewable energies sets several challenges to the operation of the National Electric System (SEN), which must be addressed in the best possible way to ensure a safe, quality and lower-cost supply. Among the tools for controlling this variability are the energy forecasts made by generating companies, which predict the amount of energy expected to be produced in a given period.

The CTO of the startup Suncast, Pedro Correa, explained that the non-conventional renewable energy forecasts (NCRE) are forecasts that estimate the amount of energy, which is generated in the following hours or days. These projections allow operators to make intelligent decisions regarding the performance of operation and maintenance activities. "For example, if the forecast indicates that there will be low energy production on a certain day, it is convenient to schedule maintenance for that day, thus minimizing the opportunity cost if production has to be stopped during maintenance activities," explains the specialist.

On this matter, Claudio Seebach, dean of the Faculty of Engineering and Sciences of the UAI, highlights the importance of these forecasts: "For example, as opposed to fossil fuels, where one can store them and/or contract them in the long term, that is, they are manageable, variable renewables, solar photovoltaic and wind are not manageable, so it is necessary to predict when there is going to be sun and when there is going to be wind."

“Data technologies have enabled the successful incorporation of NCRE into the energy matrix in both Chile and abroad, and have without a doubt become key elements in the operation of electricity systems worldwide”, Pedro Correa, CTO of Suncast.

According to the scholar, the factors needed to prepare forecasts for solar energy have to do with the presence of clouds, seasonality, geographic location, and topography, depending on the power plant's location. In wind power generation, they are linked to pressure variations, seasonality, correlation with scarcity or presence or absence of rainfall, cyclical factors such as El Niño/La Niña and time of day, among others.

From the coordinates to the Coordinador

The Coordinador Eléctrico Nacional reports that currently the projection of the energy production of each plant is reported by the companies with an hourly detail and, depending on their use, they consider different time frames. "The Coordinador incorporates this information in the programming and operation models of the interconnected system, intending to have a system operating safely and at minimum cost, as it allows anticipating the availability and variability of this type of sources," the agency explains.

To process this information appropriately, the Coordinador designed a "Centralized Forecasting System", which integrates the different production forecasts from each power plant and compares, combines and mixes this information, achieving a result that seeks to minimize deviations compared to its actual operation.

This system uses several technological tools, including machine learning developments and analytical and visualization tools, to obtain more accurate production forecasts for each plant.

Smart forecasts

According to experts, forecast accuracy is sustained by using digital tools such as machine learning, artificial intelligence (AI), big data and IoT, as well as, other systems that ease the processing of different variables.

As Claudio Seebach explains, the systems gather data from numerous sensors, which measure humidity, wind speed, temperature, and solar radiation, and are complemented by aerospace or climate modelling information. "All this supplies machine learning systems and, through AI mechanisms, they can generate these forecasts to improve the prediction of the availability of variable renewable energy sources and, therefore, help system operators to securely and optimally operate, making the best economical use of these energy sources," says the dean of UAI Engineering.

As an example, Pedro Correa explained that to calculate generation forecasts, Suncast is based on machine learning models that combine various meteorological variables of satellite origin with historical generation data from the plant. "These models learn to predict the performance of power generation in response to different weather conditions, also considering operational aspects such as nominal and minimum technical power, as well as the availability of the plant, among other factors," says the expert.

New methods

Correa highlights that the use of data technology has increased globally across all industries, including that of energy: "These tools are currently being used to improve the accuracy and efficiency of the generation forecasts, as well as to optimize the management and operation of energy systems."

According to him, big data allows the analysis of large volumes of data from various sources, whereas the integration of intelligent systems and machine learning makes it possible to develop more sophisticated and adaptable models. "Additionally, IoT facilitates real-time collection of data from sensors and connected devices, improving monitoring and response to changes in the environment," adds the specialist.

"Considering the large volume of information generated by the operation and used in a series of processes (...) we aim to move towards a 'Data Driven' company, which implies having an internal governance of the data generated, official information sources with known holders, standardization and simplification in a common language like the one used by the electricity industry," Coordinador Eléctrico Nacional.

Currently, the Coordinador is working on several pilot projects to address, among other aspects, new methods for forecasting variable renewable generation behavior as well as electricity consumption. “We have a machine learning pilot that seeks to predict the behavior of PV generation at dawn, related to ramping control in real-time operation and control of grid frequency. The model was tested, and currently, we are at the stage of testing the performance on A real online history database.”

The technical organization assures that it is engaged in digitizing processes, seeking and developing tools to help the different areas of the company and that it is closely watching what other independent network operators (ISOs) are doing in other countries.

B2B Media Group (2024). Digitization-supported power generation forecasts: More precise data for safer operation at a lower cost. In Electricidad 290 magazine, pages 20-23. Retrieved from [https://issuu.com/csa2020/docs/elec_290?fr=xKAE9_zU1NQ]