Design and Implementation of Kalman Filter-Based MPC-MPPT Algorithm for PV DC-DC Converter Systems
Open Access
- Author:
- Capuano, Matthew
- Graduate Program:
- Electrical Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- November 04, 2021
- Committee Members:
- Rafic Bachnak, Program Head/Chair
Xinwei Niu, Thesis Advisor/Co-Advisor
Javad Khazaei, Committee Member
Kiana Karami, Committee Member
Nashwa Nabil Elaraby, Committee Member - Keywords:
- Maximum Power Point Tracking
MPPT
Kalman Filter
Field Programmable Gate Array
FPGA
Solar Energy
Photovoltaics
Incremental Conductance
INC
DC-DC Converter
Boost Converter
Model Predictive Control
MPC - Abstract:
- Abstract— DC-DC converters and their respective control systems are commonly used in photovoltaic (PV) energy systems in order to maximize the power that can be extracted from a PV source and supply a steady DC signal to a load while providing a desired amount of gain. Since PV cells have low power efficiency and contain variable I-V and P-V characteristics, a maximum power point tracking-based (MPPT) control system for the converter must be designed and implemented in order for the converter to consistently draw maximum possible power from the PV source and thus apply maximum possible power to a load. However, noise present in the DC-DC converter and its sensors can lead to tracking failure for many of the common MPPT algorithms in use today. In this thesis, a MPPT algorithm is proposed where a Kalman filter is combined with the Incremental Conductance (INC) algorithm in order to track maximum PV power. Moreover, the control of a custom topology DC-DC boost converter is performed in an optimal control scheme comparable to that of Model Predictive Control (MPC) by using the tracked value. The design utilizes an averaged state space model of the DC-DC converter that a Kalman filter uses to estimate system states, filter out noise from existing sensors, and predict future states of the system given a small amount of change in duty cycle, thus allowing for a reduction in sensor count and an improvement in tracking ability given the presence of noise. The Incremental Conductance algorithm then generates the desired reference signal that is compared to the predicted signals generated from the Kalman Filter to control the converter’s duty cycle as needed. The proposed system in its entirety is designed and simulated in MATLAB/Simulink software, and the results show that the proposed algorithm can not only reduce sensor count, but also achieve higher accuracy and efficiency in the presence of noise. Specifically, accurate tracking is seen to be maintained when sensor noise power levels exceed beyond 1*10-10, which, from results, is a threshold value where other tracking algorithms are seen to begin to lose tracking. Additionally, an FPGA-based hardware-software co-simulation platform is implemented, verified, and analyzed. The results show that under real-world noise situations, the proposed design can still achieve high efficiency results under the reduced sensor count conditions. Meanwhile, the FPGA maintains low power levels, with thermal power estimates as low as 468.87 mW. The functionality of this design is compared to that of the Incremental Conductance and Model Predictive Control – Incremental Conductance (MPC-INC) algorithms, and analysis of transient response, steady state oscillations, and power efficiency is conducted under various levels of PV solar irradiance, PV temperature, and sensor noise.