Adaptive automatic generation control based on gain scheduling and neural networks

• Main Authors: Dingguo Chen, Lu Wang
• Format: Conference Proceeding
• Language: English
• Subjects: Adaptive Control; Automatic generation control; Automatic Generation Control (AGC); Control Performance Standards (CPS); Frequency control; Gain Scheduling; Generators; Neural Network; Neural networks; Power systems; Scheduling
• ISSN: 1944-9933
• DOI: 10.1109/PESGM.2017.8274457

A neural network based gain scheduling scheme for conventional PI and load frequency controllers has been proposed in this paper. Power systems are highly complicated, dynamic, nonlinear systems. The operating conditions change constantly and move away from the nominal operation conditions at which fixed gain controllers are designed so the set of fixed control parameters are not sufficient to maintain high control performance. The NERC's Control Performance Standards (CPS) set up the minimum CPS1 and CPS2 requirements for power utilities and independent system operators (ISO) to operate respective power systems to meet, which poses even more challenges for the conventional PI controllers from the AGC perspective. To enhance the conventional PI controllers, gain scheduling and neural networks are applied. As this approach is focused on driving the system toward immediate compliance of CPS standards without looking ahead for the ultimate compliance of 12-month CPS1 and monthly CPS2, this represents a less computationally intensive approach compared to a full blown predictive CPS control approach proposed in a previous effort.