Abstract:Model predictive control (MPC) has proven its applicability in power conversion control with its fast dynamic response to reference changes while ensuring critical system constraints are satisfied. Even then, the computational burden still remains a challenge for many MPC variants. In this regard, this paper formulates an indirect MPC scheme for grid-side wind converters. A quadratic program with linear constraints is solved in a receding horizon fashion with a subsequent PWM modulator. To facilitate its solution within a few hundreds of microseconds, its decision variables (modulating signals) are restricted to a specific frequency content. This approach limits the increase in problem size due to horizon length. In case studies, the proposed MPC exhibits fast response in faults and operates the converter within its safety limits.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to efficiently control the grid - side operation in wind power conversion systems, especially in the case of low device switching frequencies. Specifically, the paper proposes an indirect model predictive control (MPC) scheme, aiming to reduce the number of decision variables, so as to solve the optimization problem within a few hundred microseconds, while ensuring a rapid response to faults and operating the converter within a safe range.
### Main problems
1. **Computational burden**: Although traditional MPC methods can quickly respond to reference changes and ensure that key system constraints are met, the computational burden remains a challenge.
2. **Number of decision variables**: In order to solve the optimization problem within a limited time, the number of decision variables needs to be reduced.
3. **Frequency limitation**: By limiting the frequency content of the modulation signal, the problem size does not increase significantly with the length of the prediction range.
### Solutions
The paper proposes an indirect MPC scheme, with the following main features:
- **Frequency limitation**: By limiting the frequency content of the decision variables (modulation signals), the number of decision variables is reduced.
- **Linear growth**: The problem size is only linearly related to the number of included frequency components, rather than the length of the prediction range.
- **Rapid response**: In the case study, the proposed MPC scheme shows a rapid response to faults and operates the converter within a safe range.
### Formula
The optimization problem in the paper can be expressed as a quadratic programming problem (QP) with linear constraints, in the following specific form:
\[
\begin{aligned}
& \text{minimize} & & \sum_{t \in T_h} \left( \lambda_{\text{dc,sav}} (v_{\text{dc,sav}} - V^*_{\text{dc,sav}})^2 + \lambda_{\text{dc,sdf}} v^2_{\text{dc,sdf}} + \lambda_{Q,\text{ps}} (2 v^\top_{\text{pcc,ps}} J i_{\text{av,ps}} - Q^*_{\text{ps}})^2 \right. \\
& & & \left. + \lambda_{P,\text{ns}} (2 v^\top_{\text{pcc,ns}} i_{\text{av,ns}} - P^*_{\text{ns}})^2 + \lambda_{Q,\text{ns}} (2 v^\top_{\text{pcc,ns}} J i_{\text{av,ns}} - Q^*_{\text{ns}})^2 \right) \\
& \text{subject to} & & v_{\text{av}} = v_{\text{av,ps}} + v_{\text{av,ns}}, \quad \frac{d v_{\text{av,ps}}}{dt} = \omega J v_{\text{av,ps}}, \quad \frac{d v_{\text{av,ns}}}{dt} = -\omega J v_{\text{av,ns}}, \\
& & & v_{\text{df}} = v_{\text{df,ps}}, \quad \frac{d v_{\text{df,ps}}}{dt} = \omega J v_{\text{df,ps}}, \\
& & & v_{\text{pcc}} = v_{\text{pcc,ps}} + v_{\text{pcc,ns}}, \quad \frac{d v_{\text{pcc,ps}}}{dt} = \omega J v_{\text{pcc,ps}}, \quad \frac{d v_{\text{pcc,ns}}}{dt} = -\omega J v_{\text{pcc,ns}}
\end{aligned}
\]