The mission of this group is to bring together utility professionals in the power industry who are in the thick of the digital utility transformation. 

Post

Unleashing the power of advanced weather forecasting with artificial intelligence

image credit: © BiancoBlue | Dreamstime.com

This item is part of the Special Issue - 2020-12 - Data Analytics & Intelligence, click here for more

Written By: Sean Ratka, Francisco Boshell, Arina Anisie (International Renewable Energy Agency -IRENA-)

Advanced weather forecasting is a powerful tool for the integration of variable renewable energy (VRE), such as solar PV and wind power, into power systems. Accurate forecasting allows for more precise estimates of the amounts of VRE likely to be available in specific time frames, both short- and long-term. This, in turn, improves system stability and system planning, guiding long-term VRE geographical plant placement, among others. There are two main pieces to the advanced weather forecasting puzzle, hardware and software. While meteorological devices capture real-time, site-specific weather data, artificial intelligence (AI) can produce advanced forecasts for solar irradiation and wind speed output based on this data.

Your access to Member Features is limited.

Please or apply for membership to continue reading this post.

Advanced weather forecasting is one of the main applications of AI in facilitating and improving VRE integration by decreasing the generation uncertainty (for more information see IRENA’s Innovation landscape brief: Artificial Intelligence and Big Data, IRENA, 2019a). Driven by an increase in data due in part to the influx of IoT devices, increases in computing power, and refinement of algorithms, the applications for AI in the power sector have expanded in recent years, notably for weather forecasting, key for an accurate renewable generation forecast (Figure 1).

Thanks to the increasing use of AI fuelled by big data, time granularity for short-term predictions has increased as well (IRENA, 2020). When it comes to intraday and day-ahead electricity market trading, accurate weather forecasts can enable renewable power generators to better estimate the generation and bid accordingly, reducing penalties imposed for deviations between actual and scheduled power generation. By using a more advanced forecasting solution with numerical weather prediction models, renewable energy forecasts can be predicted closer to actual generation time. Statistical methods using AI models consider project-specific data along with nearby weather conditions for more accurate short-term forecasts. For example, system operators in the United States, such as the Midcontinent Independent System Operator and Electric Reliability Council of Texas, combine short-term dispatch with very short-term forecasts (within 10 minutes of the actual flow of power), which allows wind power plants to be fully integrated into real-time intraday markets (Orwig et al., 2015).

For more information see IRENA’s Advanced forecasting of variable renewable power generation, IRENA, 2020.

Figure 1: Weather and power generation forecast

Source: IRENA, 2020

The data value chain may seem complex, but with the ocean of granular data being generated by IoT devices and turnkey AI solutions now available, energy sector participants can take advantage of the increased intelligence available and better predict the electricity generation and demand.

System operators usually use centralised forecasting of renewable generation, widely considered a best-practice approach for a cost-effective dispatch. Centralised forecasts provide system-wide forecasts for all VRE generators within a balancing area. Decentralised forecasts, on the other hand, administered by individual VRE asset operators, provide plant-level information to help inform system operators of potential transmission congestion due to a single plant’s output, as well as help position the plant’s bids in the forward or short-term markets (NREL, 2016). System operators can also benefit from decentralised forecasting by aggregating these data and using them for operational procedures and planning purposes.

Accurate weather forecasts, from very short-term to long-term forecasting, are key for effectively integrating VRE generation into the grid and bring valuable contributions for both renewable generators and system operators. Figure 2 summarises the benefits of short-term forecasting (defined here as minutes to one day ahead) and long-term forecasting (weeks to one year ahead). Long-term weather forecasts can be used by system operators to predict extreme weather events and better plan and prepare for such occasions.

Figure 2: Benefits of weather forecasting to system operators and renewable generators

Source: IRENA, 2020

Improving VRE generation forecasts on short-term and long-term timescales engenders a diverse set of benefits for various stakeholders in the power sector. At short timescales, accurate VRE generation forecasting can help asset owners and market players to better bid in the electricity markets, where applicable. Bids based on more accurate forecasts would reduce the risk of incurring penalties for imbalances (i.e. for not complying with the generation offered in the bid). For power system operators, accurate short-term VRE generation forecasting can improve unit commitment (operation scheduling of the generating units) and operational planning, increase dispatch efficiency, reduce reliability issues and, therefore, minimise the amount of operating reserves needed in the system. The Australian Renewable Energy Agency (ARENA) awarded funding of some USD 5.6 million (AUD 9.4 million) to 11 projects to trial short-term forecasting at large-scale wind and solar farms across Australia. The trial covers at least 45% of the National Electricity Market’s registered wind and solar capacity, which collectively provides a total of 3.5 GW of renewable electricity generation (ARENA, 2019).

In the United States, the Sun4Cast solar generation forecasting system combines various forecasting technologies, covering a variety of temporal and spatial scales, to predict local solar irradiance. Forecasts from multiple NWP models are combined via the Dynamic Integrated foreCast System, used for deriving forecasts beyond six hours, and the observation based “nowcasting” technologies, used for short-term forecasts ranging from 0 to 6 hours. These technologies are integrated to derive irradiance forecasts. These irradiance forecasts are converted into expected electricity generation values, which are then provided to industry partners for real-time decision-making (Haupt et al., 2018).

Over longer timescales (e.g. over days or seasons), improved VRE generation forecasting based on accurate weather forecasting brings significant benefits to system operators, especially when planning for extreme weather events. By contributing to the allocation of appropriate balancing reserves, long-term weather forecasting assists in ensuring safe and reliable system operations. It can also help in better planning the long-term expansion of the system, both generation and network transmission capacity, needed to efficiently meet future demand. For instance, the North Atlantic Oscillations, which are seasonal weather phenomena over the North Atlantic and Europe caused by pressure differentials, can cause as much as a 1,020% variation in wind and solar generation (Jerez et al., 2013). Long-term weather forecasts for such weather events can thus improve the resilience of the system, helping system operators in planning for alternative resources to ensure the security of supply.

In China, the Hybrid Renewable Energy Forecasting (HyRef) project installed at a hybrid 670 MW solar and wind generation unit, created by IBM for the Chinese State Grid’s Jibei Electricity Power Company Limited, is using advanced data analytics to improve the forecasting of wind power output. By using advanced tools for weather modelling, cloud-imaging technology and sky-facing cameras, paired with sensors on the wind turbine, HyRef can forecast the power output months ahead, but also up to 15 minutes before actual generation. As a consequence of using these technologies, wind curtailment has been reduced by 10%, which is the equivalent of supplying some additional 14 000 homes with electricity (Cochran et al., 2013).

Figure 3: Generation forecast methods and applications

Source: IRENA, 2020

Besides short- and long-term forecasting, some other specific use cases for AI in the power system today include: power trading, where AI can determine prices for day ahead or intraday auctions; transmission and distribution, where AI can provide electricity loss forecasts for advance purchasing; transmission, distribution and retail, where AI can determine electricity demand forecasts for grids or customers; distribution and retail, where AI can provide heat demand forecasts for district heat portfolios.

Enablers

At its core, AI involves identifying patterns and trends within data sets in order to generate predictive analytics, provide actionable insights, and automate actions based on predictions. The most obvious enabler of powerful, accurate advanced forecasting using AI is data. Models need to be trained with ample clean, time series-relevant data, from which the AI analytics can pick up the patterns and apply them to new scenarios. Fortunately, the abundance of big data, along with the exponential growth in processing power witnessed over the past few decades, has created the ideal setting for AI (IRENA, 2019). Moreover, by 2030, 50 billion internet-connected devices are expected to come online globally in all sectors, double the amount in 2018 (Statista, 2020).

Besides having enough good quality data on which to base your analysis, IRENA’s Advanced forecasting of variable renewable power generation report outlines three distinct enablers:

A market design incentivising accurate short-term VRE power generation forecasting

To enhance the operation of a power system with significant VRE shares, and make use of the accurate generation forecasts, where applicable, electricity markets need to increase time granularity; in other words, the dispatch and scheduling time interval, the pricing of market time units, financial settlement periods, and the time span between gate closure and the real-time delivery of power should be reduced (for more information see IRENA’s Innovation landscape brief: Increasing time granularity in electricity markets, IRENA, 2019b).

Open source systems for weather data collection and sharing

Open sourcing weather data collected by the weather monitoring stations of VRE generators, national meteorological institutes, and information and communication technology developers can foster rapid advances in data analytical techniques and consequently in weather forecasting. For weather data collection, a network of weather stations at national or regional level can be deployed to collect and store long-term meteorological data, which can be used to characterise renewable energy resources.

Advanced meteorological devices

Weather forecasting tools and models that are being experimented with include the use of advanced cloud-imaging technology; sky-facing cameras to track cloud movements; and sensors installed on turbines to monitor wind speed, temperature and direction. Using such advanced meteorological devices, which are connected to the Internet, may help in gathering information of real-time, site-specific weather conditions (for more information on internet-connected devices, see IRENA’s Innovation Landscape Brief: Internet of Things, IRENA, 2019c).

Moving forward

AI is emerging as an enabler of the energy sector of the future. The vast amount of data being generated by, and harvested from, the energy system is ever increasing. AI today could be compared to personal computing three decades ago, in that we are just beginning to explore its potential and those who do not adapt and implement these new technologies will fall behind. Deep forecasting tools are emerging which make it easier for end users to implement these new intelligent solutions. Importantly though, in terms of deployment for advanced forecasting, AI is not a silver bullet, but rather one of the tools needed to navigate an increasingly complex energy sector.


References:

AFRY (2020), Artificial Intelligence in energy forecasting, PowerPoint presentation, AFRY, Stockholm, Sweden, https://afry.com/en/webinars/ai-energy-forecasting              

ARENA (2019), “$9 million funding to enhance short term forecasting of wind and solar farms”, Australian Renewable Energy Agency, https://arena.gov.au/news/9-million-fundingto-enhance-short-term-forecasting-of-wind-andsolar-farms.

Cochran, J. et al. (2013), Market Evolution: Wholesale Electricity Market Design for 21st Century Power Systems, National Renewable Energy Laboratory, Golden, Colorado, www.nrel.gov/docs/fy14osti/57477.pdf.

Haupt, S.E. et al. (2018), “Building the Sun4Cast system: Improvements in solar power forecasting”, Bulletin of the American Meteorological Society, Vol. 99, American Meteorological Society, Boston, pp. 121–136, https://journals.ametsoc.org/doi/abs/10.1175/ BAMS-D-16-0221.1.

IRENA (2020), Innovation landscape brief: Advanced forecasting of variable renewable power generation, International Renewable Energy Agency, Abu Dhabi, https://www.irena.org/-/media/Files/IRENA/Agency/Publication/2020/Jul/IRENA_Advanced_weather_forecasting_2020.pdf?la=en&hash=8384431B56569C0D8786C9A4FDD56864443D10AF

IRENA (2019a), Innovation landscape brief: Artificial intelligence and big data, International Renewable Energy Agency, Abu Dhabi, https://www.irena.org/publications/2019/Sep/Artificial-Intelligence-and-Big-Data

IRENA (2019b), Innovation landscape brief: Increasing time granularity in electricity markets, International Renewable Energy Agency, Abu Dhabi, https://www.irena.org/-/media/Files/IRENA/Agency/Publication/2019/Feb/IRENA_Increasing_time_granularity_2019.pdf?la=en&hash=BAEDCA5116F9380AEB90C219356DA34A5CB0726A

IRENA (2019), Innovation landscape brief: Internet of Things, International Renewable Energy Agency, Abu Dhabi, https://www.irena.org/publications/2019/Sep/Internet-of-Things

Jerez, S. et al. (2013), “The impact of the North Atlantic Oscillation on renewable energy resources in southwestern Europe”, Journal of Applied Meteorology and Climatology, Vol. 52, American Meteorological Society, Boston, pp. 2 204–2 225, https://journals.ametsoc.org/ doi/full/10.1175/JAMC-D-12-0257.1.

NREL (2016), Forecasting wind and solar generation: improving system operations, National Renewable Energy Laboratory, Golden, Colorado, www.greeningthegrid.org/resources/factsheets/copy_of_ForecastingWindandSolarGeneration.pdf

Orwig, K.D. et al. (2015), “Recent trends in variable generation forecasting and its value to the power system”, IEEE Transactions on Sustainable Energy, Vol. 6/3, pp. 924–933, https://ieeexplore.ieee.org/document/6996049.

Statista (2020), Number of internet of things (IoT) connected devices worldwide in 2018, 2025 and 2030(in billions), Statista, https://www.statista.com/statistics/802690/worldwide-connected-devices-by-access-technology/ 

Discussions

Spell checking: Press the CTRL or COMMAND key then click on the underlined misspelled word.

No discussions yet. Start a discussion below.

Get Published - Build a Following

The Energy Central Power Industry Network is based on one core idea - power industry professionals helping each other and advancing the industry by sharing and learning from each other.

If you have an experience or insight to share or have learned something from a conference or seminar, your peers and colleagues on Energy Central want to hear about it. It's also easy to share a link to an article you've liked or an industry resource that you think would be helpful.

                 Learn more about posting on Energy Central »