EUBCE 2026 - José A. VELEZ - Identification of Cause–Effect Relationships and ANN-Based Modeling in a 15 MW Biomass Rankine Power Plant Supported by Data Reconciliation

Identification of Cause–Effect Relationships and ANN-Based Modeling in a 15 MW Biomass Rankine Power Plant Supported by Data Reconciliation

Short Introductive summary

One of the most widely used processes for solid biomass conversion is the Rankine cycle. In the case of power plants using solid biomass burned on a grate the achievable steam parameters limit the overall efficiency, which typically ranges from 0.15 to 0.30. For this reason, it is of primary importance to identify any factors that may reduce the nominal efficiency of the biomass plant in order to overcome their effects. This work focuses on two main aspects: • First, a methodology was implemented to identify the main factors responsible for an efficiency reduction of approximately 12%, decreasing from a nominal 25% under site conditions to an annual average of 22%. Using real data collected from the Supervisory Control and Data Acquisition (SCADA) system, and after applying the congruence technique, combustion distortion was identified as the main cause. • Second, a data-driven model based on a Long Short-Term Memory (LSTM) neural network was developed for power and emission prediction as well as plant control. All the data used in the LSTM approach had been previously processed through a reconciliation framework. This model is expected to replace the plant’s current manual control

F. Jimenez-Espadafor, AICIA, Sevilla, SPAIN
J. Velez, Seville University, SPAIN
J. Serrano, Seville University, SPAIN
M. Tagua, Seville University, SPAIN