Adjustment formulas for learning causal steady-state models from closed-loop operational data
This addresses the issue of model-based optimization in industrial or engineering settings where historical data is affected by control, though it appears incremental as it builds on existing structural dynamical causal models.
The paper tackles the problem of learning causal steady-state models from closed-loop operational data, which can be confounded by control correlations, and derives an adjustment formula to account for this, enabling estimation from data under fixed control laws including feedforward and feedback control.
Steady-state models which have been learned from historical operational data may be unfit for model-based optimization unless correlations in the training data which are introduced by control are accounted for. Using recent results from work on structural dynamical causal models, we derive a formula for adjusting for this control confounding, enabling the estimation of a causal steady-state model from closed-loop steady-state data. The formula assumes that the available data have been gathered under some fixed control law. It works by estimating and taking into account the disturbance which the controller is trying to counteract, and enables learning from data gathered under both feedforward and feedback control.