Compact Temporal Causal Algorithm for Modeling Prediction in the Green Ammonia Production Process

Abstract Because of the complexity of green ammonia production, identifying useful information from massive data to construct a streamlined and interpretable prediction model is a challenge. This paper proposed an improved transfer entropy method and enhanced the model in capturing causal relationships between variables. Furthermore, a compact temporal causal identification algorithm was introduced, combining transfer entropy and direct causal identification. This algorithm integrates the advantages of both methods, enabling rapid identification of clear causal pathways in high‐dimensional variable spaces. In industrial validation of the actual green ammonia process, the average number of predictive variables based on compact temporal‐dependent variables and temporal Markov blankets were 3.341 and 10.171, respectively, yielding average prediction accuracies ( R 2 ) of 0.887 and 0.905. The method proposed provides new solutions for modeling prediction in green ammonia processes.