| 摘要 |
A recurrent distribution network provides an embedded, cost-based decision mechanism for a real-time, closed-loop process control system that outputs to multiple controllable output devices while observing numerous input constraints. A global incremental distribution request to increase or decrease a controlled process variable is accepted by a recurrent neural network. The network iteratively solves and applies a distribution of a global incremental request to multiple controllable output devices based on individual incremental unit costs, output ranges, output scaling and process input constraint limits.
|
