Power generation gas turbines carry the risk of significant equipment damage caused by combustion pressure fluctuations under high-temperature combustion for high-efficiency operation. The technology in this report allows the AI to learn combustion pressure fluctuation characteristics in real time and makes oscillation predictions that account for the AI's prediction confidence, thereby automatically adjusting operating conditions cautiously in low-prediction-confidence regions and proactively in high-prediction-confidence regions. By utilizing this technology, combustion pressure fluctuations have been successfully and rapidly suppressed while reducing the risk of incorrect adjustments. This report describes an overview of this technology and the results of verification at Mitsubishi Heavy Industries, Ltd.'s demonstration power generation facility.
