Efficiency Increase of 1 in Coal-Fired Power Plants with Mathematical Optimization

摘要

The entire process of a combined-heat-and-power (CHP) coal-fired power plant from coal delivery to electricity and heat generation can be modeled using machine learning methods that generate a single set of equations that describe the entire plant. The basis for the model is the historical data from the data historian. No engineering or human input is required. This model is used to compute the optimal operational point at any moment in time, taking into account the current status of the plant as well as relevant external and internal factors. The criterion for optimality is the overall equipment efficiency over the entire plant. This method requires no changes to be made to the plant and requires nominal human effort to implement. The achievements are based on operating changes that are computed. No human judgment is involved, thus the model excludes a potential bias or hu man error. Having a uniform operational strategy 24 hours a day combined with the fact that this model can incorporate the full complexity of the plant is the novel element that lies at the heart of the improvement possibility in the range of one percent. This has been demonstrated in the Reuter-West CHP plant of Vattenfall in Berlin, Germany.