I use the Kalman filter to estimate the stator and rotor currents for a 2MW doubly fed induction generator, I used the state model in the alpha beta reference, How to calculate the initial matrix P_0, Q_0, R_0 for a extended Kalman filter?
Hi,
I suggest you see the links and attched file the topic.
http://www.anuncommonlab.com/articles/how-kalman-filters-work/part2.html
https://hal.science/tel-01247723/document
http://eceweb1.rutgers.edu/~orfanidi/aosp/aosp-ch13.pdf
https://homes.cs.washington.edu/~todorov/courses/cseP590/readings/tutorialEKF.pdf
https://journals.sagepub.com/doi/abs/10.1177/0954410016641708?journalCode=piga
Best regards
Mohammed Abbas In an extended Kalman filter (EKF), the starting matrices P 0, Q 0, and R 0 play a significant role in the estimating process. They are, in order, the starting values of the state covariance matrix P, the process noise covariance matrix Q, and the measurement noise covariance matrix R. These matrices are used to represent the uncertainty in the state estimation, process dynamics, and measurement noise, in that order.
You must consider the following factors while calculating these matrices:
1. P 0 state covariance matrix: This matrix captures the state estimate's starting uncertainty. To represent the fact that the initial state estimate is unknown, it is commonly set to a big number for the diagonal elements and zeros for the off-diagonal elements.
2. Q 0: Process noise covariance matrix This matrix illustrates the process dynamics' uncertainty. It is often determined based on a prior understanding of process dynamics and how they evolve over time.
3. R 0 is the measurement noise covariance matrix. This matrix displays the measurement uncertainty. It is often set based on measurement noise characteristics such as sensor accuracy.
P 0, Q 0, and R 0 values can be derived by tuning or previous knowledge of the system. As the EKF converges to a more accurate estimate, the values can be changed over time. Finding the proper values for these matrices, on the other hand, may be difficult, as estimation success is largely dependent on the choice of these.
I agree with Qamar's response.
In addition, you will need an initial estimate of the state vector, x^(0) ( '^' for estimate), also. A proper way of initializing x^(0) and its error covariance is to use enough number of initial measurements to estimate them both. For instance, to estimate initial position and velocity, we may use first two position measurements (assuming that there is no velocity sensor). This process of estimation for both x^(0) and its variance is given in detail in
Bar-Shalom, Y., Rong Li, X., and Kirubarajan, T., Estimation with Applications to Tracking and Navigation, Wiley 2001
Hari Hablani
IIT Indore
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix