Online operation optimization based on GA-RBF network in ethylene cracking furnace

Based on Object Linking and Embedding (OLE) for Process Control (OPC) technology, real-time data was collected and preprocessed by data filtering and anomaly detection methods. For ethylene cracking furnace´s Multi-In-Multi-Out (MIMO) process, an online soft measurement model was built based on Radical Basis Functions (RBF) neural network. Meanwhile, an engineering method based on production experience was used to adjust the online model. Finally Genetic Algorithm (GA) optimized the online model by maximizing the sum of the yields of ethylene and propylene to find the optimal operation conditions. The actual industrial applications show that this method can increase the yields of ethylene and propylene, and it has good adaptability and stability and has important operational guiding significance to the actual production process.