AI’s literally game-changing potential was amply demonstrated in March 2016 when Google’s AlphaGo software managed to beat the game’s South Korean grandmaster in four out of five matches. Siemens Corporate Technology (CT) researchers were similarly jolted when they confirmed that an artificial intelligence system they had developed could contribute to a system’s optimization. “As it turns out, we’re finding some interesting alternatives using our methods,” comments Volkmar Sterzing, a machine learning expert in CT’s Business Analytics & Monitoring Technology Field. “Any system where operating performance is based on the experience of experts can also be optimized with artificial intelligence.”
Brains in Every Burner
Thanks to neural network-based artificial intelligence (AI) developed by Siemens Corporate Technology (CT), the combustion processes in the company’s flagship gas turbine are being steadily optimized. The processes learn how to continuously adjust fuel valves, resulting in optimized combustion and reduced emissions and wear. Siemens’ Power Generation Services Division is now using the jointly developed technology for the first time in a customer application for the largest and most modern stationary Siemens gas turbine. Given the large number of such complex systems run by Siemens customers whose operations could be improved by AI, the potential for improved efficiency is enormous.
As a system increases in structural complexity and reacts to a wide range of factors, a human, even an expert, will have to make compromises when it comes to adjustments. Specialists simply cannot be on hand around the clock. This is where artificial intelligence, which monitors a system continuously, offers clear advantages.
Siemens has been researching neural networks for about 30 years and has made significant progress in applying this technology to artificial intelligence. For example, the company’s Software Environment for Neural Networks (SENN) is being continuously refined and adapted to new and evolving applications, including the optimization of gas turbines and wind turbines. “We hold something like 50 patents for learning processes,” notes Sterzing.
Siemens Power Generation Services and CT have developed a system that continuously optimizes the operation and control of combustion in gas turbines. Based on AI from CT, the system, which is known as a Gas Turbine Autonomous Control Optimizer (GT-ACO), is currently being installed at a top customer in Asia. It will be tested extensively on Siemens’ flagship H classgas turbine fleet. Testing will begin before the end of February 2017. Improvements in overall gas turbine operation can be difficult to achieve because lower emissions characteristically result in shorter service life. The reason for this is that high-energy combustion oscillations, which cause material fatigue, can intensify and cause increased wear.
Tests on a number of different gas turbine types have already demonstrated that GT-ACO works. After an expert set the turbine manually to minimum emission of nitrogen oxides, artificial intelligence then took over control of the combustion unit. “Two minutes after switching on, the value had dropped by 20 percent,” says Hans-Gerd Brummel, who is responsible for GT-ACO development at Power Generation Services and is a pioneer in turbine remote diagnostics and maintenance at Siemens. In December he received the company’s Inventor of the Year Award in honor of his life’s work.
The primary goal of using AI in turbines is to minimize emission of nitrogen oxides. To achieve this, GT-ACO’s neural model alters the distribution of fuel in a turbine’s burners. However, the settings for each burner vary depending on factors such as location, gas composition and local weather conditions. That’s why GT-ACO needs a few weeks of learning on each turbine before it can autonomously make beneficial changes to the controls.
“Customers are showing great interest in our technology,” says Brummel. “Given the high proportion of renewable energy in the supply grid, gas turbines often have to step in to maintain grid frequency.” In this constantly changing operating environment, there is growing risk of increased oscillation amplitudes, which increases wear. Brummel is confident, however “that GT-ACO can help in this case by focusing the optimization effect on damping vibrations.”
GT-ACO can also be used to partially compensate for gas turbine ageing. The reason for this is that the technology combines the collective knowledge of gas turbine thermodynamics in the form of physical models with machine learning.
Sterzing is convinced that the technology has vast potential at Siemens – in power distribution, production automation, and process industry applications. Brummel does not need any convincing in this respect. He is already in negotiations with customers in the power generation industry about GT-ACO and is developing additional optimization applications.