AML hydrogen production automation and Control
The ultimate goal of the AML system is for intelligent AI process control and optimize
Step 1: Design the process and identify important variables
Step 2: Build AML machine learning model
Step 3: Clean data for online control
in some cases, conventional machine learning algorithms allow for “over training”, to minimize the error between the predicted and actual values.
The outcome predictive model in such cases model the data – not the process. This leads to errors in control commands and produce sub-optimal process operations.
To eliminate such cases, AML’s algorithm does not allow for overtraining, and by cleaning the raw data AML produces a more accurate process predictive data.
AML model on raw data
AML model on clean data
It is important to note that the weight of each process variable on the predictive value in the case of raw data is different from that of clean data.
Wrong weight leads to a wrong control command, and
Advanced online & real-time Artificial Intelligence computer modules