Enterprise AI Analysis
Cascaded adaptive model predictive and PID control for integrated LFC-AVR enhancement
By Mohamed Ayman, Mahmoud A. Attia & Ahmed M. Asim • Scientific Reports | (2026) 16:12734
This paper addresses the challenge of degraded dynamic performance in power systems due to inherent coupling between Load Frequency Control (LFC) and Automatic Voltage Regulation (AVR) dynamics, exacerbated by load disturbances and parameter uncertainties. It proposes a novel coordinated LFC-AVR control scheme based on a cascaded Adaptive Model Predictive Controller (AMPC) in the outer loop and a PID controller in the inner loop. The AMPC uses online Recursive Least Squares (RLS) identification and a time-varying Kalman filter (TVKF) for real-time adaptation to system dynamics. Extensive simulations on single-area and two-area power system models demonstrate that the proposed AMPC-PID strategy significantly improves damping, reduces under/overshoot, and shortens settling times for both frequency and voltage deviations, outperforming recent EDO-PID and HS-optimized PID approaches under various operating conditions and parameter variations. This robust and effective solution enhances power system stability and reliability.
Executive Impact: Quantifiable Results
Implementing this cascaded AI control system delivers significant operational improvements, boosting stability and efficiency across various power system scenarios.
0
LFC Undershoot Reduction (Single-Area)
0
LFC Settling Time Improvement (Single-Area)
0
AVR Overshoot Reduction (Single-Area)
0
LFC Settling Time Improvement (Two-Area)
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Power systems face challenges in maintaining stability due to inherent coupling between Load Frequency Control (LFC) and Automatic Voltage Regulation (AVR) dynamics. Conventional approaches often treat these as independent, leading to suboptimal performance under disturbances and parameter uncertainties. Fixed-gain controllers lack the adaptability for real-time dynamic changes. Model Predictive Control (MPC) offers advanced capabilities but can be computationally intensive. This paper addresses these limitations by proposing an adaptive, coordinated cascaded control scheme.
The core innovation is a cascaded control architecture: an Adaptive Model Predictive Controller (AMPC) in the outer loop and a PID controller in the inner loop. The AMPC continuously updates its prediction model online using Recursive Least Squares (RLS) identification and a time-varying Kalman filter (TVKF) for state estimation. This allows it to adapt to time-varying dynamics and manage constraints. The inner PID loop ensures fast tracking and practical implementability, combining adaptive optimization with rapid local regulation. The control objective is to minimize a finite-horizon cost function, subject to operational constraints.
The study employs detailed single-area and two-area interconnected power system models. These models explicitly account for the dynamic coupling between LFC and AVR loops, which is often ignored in simpler designs. The LFC loop aims to regulate frequency and tie-line power, while the AVR loop maintains terminal voltage through excitation control. The models include various components like amplifier, exciter, generator, sensor, governor, turbine, and inertia/load, represented by their respective transfer functions and parameters, allowing for accurate simulation of their interactions.
Simulations on both single-area and two-area systems under step load perturbations and parameter variations demonstrate superior performance of the proposed AMPC-PID. For a single-area LFC, frequency undershoot is reduced by 33% and settling time by ~33%. For single-area AVR, voltage overshoot is reduced from ~2.5% to ~1.5%, with faster settling. In two-area systems, LFC undershoot improves by 35-40% and settling time by ~60%, while AVR overshoot is reduced from 7-8% to 4-5% and settling time improved significantly. Robustness is confirmed under varying load disturbances and parameter uncertainties (e.g., ±50% time constant variations).
The cascaded AMPC-PID controller offers significant improvements in coordinated LFC-AVR performance, demonstrating enhanced damping, reduced under/overshoot, and shorter settling times compared to optimized EDO-PID and HS-optimized PID controllers. The strategy proves robust under severe operating variations. Future work includes validation through OPAL-RT hardware-in-the-loop experiments, exploration of nonlinear power system models, and development of computationally efficient optimization techniques to reduce online burden.
33%
Reduction in LFC Frequency Undershoot (Single-Area)
AMPC Adaptive Control Strategy Steps
Initialization
→
State Measurement
→
Online Identification (RLS)
→
State Estimation (TVKF)
→
MPC Optimization
→
Control Application
→
Model Update
| Metric | AMPC-PID (Proposed) | EDO-PID (ITSE) | EDO-PID (ITAE) |
|---|---|---|---|
| Overshoot | 0.3 x 10-3 Hz | 0.8 x 10-3 Hz | 0.7 x 10-3 Hz |
| Undershoot | -4 x 10-3 Hz | -6 x 10-3 Hz | -6.2 x 10-3 Hz |
| Settling Time | ≈4 s | 6 s | 7 s |
Two-Area LFC Enhancement (0.03 p.u. load)
Under a 0.03 p.u. step load disturbance, the proposed AMPC-PID controller significantly outperforms the EDO-PID. In Area-1, the maximum frequency undershoot is reduced from -0.12 Hz to approximately -0.075 Hz, representing a 35-40% improvement. Settling time is also significantly faster, achieving recovery within 2-2.5 s compared to 6-7 s for EDO-PID. Similar improvements are observed in Area-2, confirming superior damping and enhanced coordination.
60%
Settling Time Reduction (Two-Area LFC)
| Metric | HS-EDO-PID (Proposed) | EDO-ITSE-PID | EDO-ITAE-PID |
|---|---|---|---|
| Peak Overshoot | 1-1.5% | 2% | 2.5% |
| Settling Time | ~1.5 s | ~2 s | ~2.5 s |
Robustness to Load Time Constant Variations (LFC)
The AMPC-PID controller demonstrates strong robustness to ±50% variations in the load time constant. Maximum frequency undershoot remains bounded (e.g., -4.2 x 10-3 Hz for 50% case, -3.8 x 10-3 Hz for 100% nominal, -3.6 x 10-3 Hz for 150% case). The system maintains stable and consistent frequency regulation performance, converging within 8-10 s despite initial transient oscillations, showcasing its reliability under uncertainty.
Calculate Your Potential AI ROI
Estimate the potential annual operational savings and reclaimed hours by implementing an Enterprise AI solution in your power system operations.
Annual Estimated Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A structured approach to integrating advanced AI control in your power systems, from model identification to real-world deployment.
Phase 1: Model Definition & Identification
Accurately define and identify the dynamic models of LFC and AVR subsystems. This involves gathering system parameters and applying online recursive least squares (RLS) and Kalman filter techniques for real-time model adaptation.
Phase 2: Controller Design & Tuning
Design the cascaded AMPC-PID controller, carefully tuning the AMPC's prediction horizons, weighting matrices (Q, R), and the inner PID gains. Focus on balancing transient performance, control effort, and robustness, ensuring proper coordination between the loops.
Phase 3: Simulation & Validation
Extensively simulate the proposed control strategy on single-area and multi-area power system models. Validate its effectiveness and robustness under various operating conditions, including step load perturbations, parameter uncertainties, and different load change scenarios.
Phase 4: Hardware-in-the-Loop (HIL) Implementation
Conduct Hardware-in-the-Loop (HIL) experiments using platforms like OPAL-RT. This crucial step strengthens the real-time feasibility assessment and bridges the gap between simulation and practical deployment, testing the controller under more realistic conditions.
Phase 5: Real-World Deployment & Monitoring
Implement the cascaded AMPC-PID scheme in a pilot real-world power system. Continuously monitor its performance, make necessary adjustments, and explore further enhancements like integrating nonlinear models or optimizing computational efficiency for long-term operational benefits.
Ready to Transform Your Operations?
Connect with our AI specialists to explore how these advanced control strategies can be tailored to your enterprise's unique needs and infrastructure.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat