A PI+R Control Scheme Based on Multi-agent Systems for Economic Dispatch in Isolated BESSs

What the math gives to ML

The paper offers a nonstandard hybrid PI+R mechanism for distributed economic dispatch: an integral state accumulates mismatch during continuous flow, then is reset to zero when the proportional error crosses zero, with a dwell-time guard preventing repeated instantaneous resets. Its transferable asset is not ordinary PI control alone, but event-triggered removal of accumulated integral memory precisely at a predicted sign change to reduce overshoot and settling time. A practical neural-network analogue is a reset-integral optimizer whose accumulated gradient or residual state is cleared at zero crossings while enforcing a minimum number of updates between resets.