Utilities are integrating renewables such as wind and solar power into their production systems, in combination with traditional generation – but managing these complex networks efficiently requires sophisticated predictive modeling and optimization.
Red Eléctrica de España completed a proof of concept using IBM Watson Studio with IBM Decision Optimization (CPLEX) to forecast electricity demand on the Canary Islands and explore how the company could streamline its modeling and optimization processes.
Acceleratesthe development of sophisticated supply and demand forecasting models
Unlocksnew techniques such as machine learning to help improve forecast accuracy
Reducesthe burden of code maintenance and deployment, potentially saving hours of effort
Business challenge story
Need for self-sufficiency
When demand for electricity peaks or subsides, utilities typically have the option of trading electricity with the grid to maintain a reliable service. That’s not an option on the Canary Islands, which lie far out at sea and are isolated from mainland Spain’s electric transmission network.
Unable to trade capacity, the islands must meet their entire electricity demand through local production. If they schedule more electricity than required, the excess capacity is spilled and money is wasted. Worse still, if demand exceeds production, black-outs and power-cuts can follow. To maintain a reliable electricity supply for Canarian homes and businesses while minimizing cost, Red Eléctrica de España must accurately forecast demand and schedule sufficient generation to match it.
As if that weren’t challenging enough, there’s another layer of complexity. Along with other utilities, Red Eléctrica de España is attempting to reduce the area’s carbon footprint by integrating renewable energy sources. Compared to fossil fuels, wind and solar power are sustainable and highly efficient. However, they are also unreliable, because no-one can say for certain when the wind will blow or the sun will shine. To provide a reliable supply, Red Eléctrica de España must therefore strike the right balance between renewable energy sources and fossil fuels.
To predict energy consumption and optimize the energy mix, Red Eléctrica de España performs highly sophisticated predictive and prescriptive analytics. The company currently conducts these analytics processes using a custom-built solution that it developed in-house many years ago. While the process and models are fairly accurate, computations are heavy and the software is difficult to maintain.
For example, each time Red Eléctrica de España develops a new prescriptive optimization model, the update must be installed individually on each of its 19 machines – a time-consuming process. This limitation makes it difficult for the company to develop and roll out enhancements for the solution to refine its models and improve its reporting capabilities.
To escape these shackles, the company was eager to test a more modern data science toolset that would be easier to manage.
In a recent proof of concept, Red Eléctrica de España trialed moving its analytics models to IBM Watson Studio Local. Watson Studio is a collaborative platform for data scientists that combines proprietary and open-source data science tools into a coherent, integrated and controlled environment. Within the IBM Watson Studio Local portfolio, Red Eléctrica de España used IBM Decision Optimization, including IBM CPLEX Optimizer, and IBM SPSS® Modeler Stream Canvas.
Mustafa Pezic, Senior Project Manager at Red Eléctrica de España, explains: “IBM Watson Studio Local puts a powerful set of data-science tools at our fingertips that we can use in an efficient, synchronized manner across our team. For example, one team member can write code in Python in a Jupyter Notebook, while another can use SPSS to solve problems in a more visual way. We can collaborate effectively while giving each person freedom to choose how they work.”
For help on the proof of concept, Red Eléctrica de España turned to the IBM Data Science Elite Team, which provided six weeks of on-site support to help the company’s data science team learn how to get the most value out of the platform.
“The IBM Data Science Elite Team provided outstanding support during the proof of concept,” comments Mustafa Pezic. “They showed us how to use the IBM Watson Studio Local solution and switch between the different tools, and with their help we successfully built three complex models within just six weeks. Throughout the engagement, the IBM team were professional and enthusiastic and demonstrated their immense technical expertise.”
At the end of the proof of concept, Red Eléctrica de España provided a strong recommendation for other companies to work with the IBM Data Science Elite team: an end-of-project survey gave the IBM team a net promoter score of 9.5 out of 10.
Putting operations under the microscope
In the first stage of the project, IBM helped Red Eléctrica de España port its existing model for forecasting electricity demand to IBM Watson Studio Local. Next, the team transferred the energy generation model, which optimizes how to produce enough energy each hour to meet demand while maximizing use of renewable energy and minimizing use of fossil fuels, and which is written and solved with IBM CPLEX Optimizer.
The energy generation model is highly detailed, showing when to turn particular generators on and off, and how much reserve capacity the company should maintain in its transmission network to accommodate the uncertainty of wind and minimize the risk of blackouts. The outputs of the forecasting and energy generation models enable Red Eléctrica de España to plan and simulate generation one year ahead.
In a second phase of the project, IBM and Red Eléctrica de España used IBM Watson Studio Local to examine its long-term demand forecasting model, which looks at factors such as climate change and population growth. An updated version of this model could potentially help Red Eléctrica de España decide where and when the system needs new capacity, and the impact of building new wind turbines, solar panels or fuel-powered plants.
For Red Eléctrica de España, the proof of concept demonstrated that IBM Watson Studio Local would enable closer connections between its demand forecasting and energy generation environments.
“At the moment, we operate separate environments for predictive analysis of demand forecasting, and the prescriptive analysis that helps us set the right balance between renewables and fossil fuels,” explains Mustafa Pezic. “IBM Watson Studio Local, with the embedded IBM Decision Optimization (CPLEX), lets us connect those two environments in a combined platform, which helps us better account for uncertainties in how much electricity we can generate using renewable sources.”
Streamlining model development
The proof of concept demonstrated the viability of implementing Red Eléctrica de España’s existing models in IBM Watson Studio Local – reducing the burden of system management.
“Our current data science platform is installed individually on each user’s laptop and there is a significant risk of inconsistencies,” continues Mustafa Pezic. “For example, if I am using the older version of the model then my results are not reflected in the newer version being used by my colleagues. To reduce this risk, each time we update our model we need to manually update all 19 machines across our test and development environments. It takes a day to update, and if we hit any system errors we need to restart from scratch.
“By contrast, with IBM Watson Studio Local all users would run on a centralized application, so if we publish an update then immediately everyone is live on it. As a result, collaboration is much easier and we can avoid inconsistencies arising from using different versions. Because the update process is so much easier with IBM Watson Studio Local, we would be able to refine our models more frequently.”
Looking to the future
In the future, Red Eléctrica de España could achieve even greater benefits from IBM Watson Studio Local by taking further advantage of its machine learning capabilities.
For example, imagine that a certain combination of weather patterns across the country means that the optimal result is always to use certain generators and a particular fuel type. Today, Red Eléctrica de España would need to re-run the optimization model each time that situation arises. But by using a machine learning model to identify and encode that rule, the company would be able to identify the correct solution much faster, and save the cost of computation.
Mustafa Pezic concludes: “We have seen how IBM Watson Studio Local can help us collaborate more efficiently and reduce the burden of system management. We are very interested in exploring how machine learning can deliver further value for our business.”
Red Eléctrica de España
Red Eléctrica de España is responsible for the supply of electricity to people and corporations across Spain. Founded in 1985, the company employs more than 1,800 people and its network stretches across 44,000 kilometers (27,000 miles).
Take the next step
IBM Analytics offers one of the world's deepest and broadest analytics platform, domain and industry solutions that deliver new value to businesses, governments and individuals. For more information about how IBM Analytics helps to transform industries and professions with data, visit ibm.com/analytics. Follow us on Twitter at @IBMAnalytics, on our blog at ibmbigdatahub.com and join the conversation #IBMAnalytics.