Nonlinear loads such as diode and thyristor rectifiers draw distorted current from the electrical supply. The resulting harmonics can affect power quality and power factor. A shunt active power filter (SAPF) addresses these effects by injecting compensating current at the point of common coupling. In a simulated three-phase system, combining direct power control with particle swarm optimization (PSO) to tune the DC-link PI controller improved the reported harmonic and transient performance compared with conventional PI tuning.
Why use a shunt active power filter?
Passive LC filters have traditionally been used to reduce current harmonics, but active filtering offers a different approach. A SAPF uses a power-electronic inverter, a coupling filter and an energy source, together with a controller that determines the inverter’s switching states. Its purpose is to produce filter current that counteracts unwanted load-current components, so the source supplies the required fundamental current more closely.
The current relationship can be expressed as the sum of source current and filter current equaling load current, with the sign depending on the chosen current directions. In practical control terms, the filter is commanded to compensate for the load components that should not be drawn from the grid, including harmonics and reactive current. The intended result is improved source-current quality and reduced reactive power at the supply.
Direct power control for a three-phase SAPF
Direct power control (DPC) regulates instantaneous active power and reactive power rather than relying on inner current-control loops and a separate pulse-width modulation stage. The controller estimates active and reactive power from the measured three-phase voltages and currents, then compares those estimates with their reference values.
Hysteresis comparators convert the power errors into discrete control signals. The controller also identifies the sector occupied by the supply-voltage vector in the alpha-beta reference plane, which is divided into 12 sectors. A switching table uses the sector and comparator outputs to select inverter switching states intended to move active and reactive power toward their references.
This direct selection of switching states is a defining feature of the strategy described. It avoids the inner current loops and PWM modulator used in some other converter-control approaches. The reactive-power reference can be set by the control system, while the active-power reference is produced by the DC-link voltage controller.
Why tune the DC-link PI controller with PSO?
The DC-link capacitor voltage must be regulated as the SAPF operates and the load changes. A PI controller compares measured DC-link voltage with its reference and uses the error to generate the active-power command. Its proportional and integral gains affect the voltage response, including overshoot and how quickly the system responds to changes.
Instead of selecting those gains only through conventional tuning, the described method treats them as optimization variables. PSO searches for candidate gain pairs by evaluating a fitness function. Each particle represents a candidate solution; its search is influenced by the best position it has found and the best position identified by the swarm.
The optimization objective is based on the integral time absolute error (ITAE), with maximum overshoot also considered in the stated formulation. In general, an error measure that weights error over time encourages the search to favor controller settings that reduce persistent tracking error. The authors used MATLAB/Simulink simulations to compare the conventional PI controller with the PSO-tuned PI controller in the DPC-based SAPF.
What the simulation results indicate
Before compensation, the simulated supply current had a reported total harmonic distortion (THD) of 28.50%. After compensation, the reported THD was 1.71% with the conventional PI controller and 1.03% with PI-PSO. These figures describe the study’s simulation results; they should not be interpreted as guaranteed outcomes for other installations, where performance depends on the system, loads and implementation.
The comparison also reported a substantial difference in DC-link voltage overshoot and ITAE:
- Conventional PI: proportional gain of 0.2, integral gain of 10, overshoot of 14.258%, and ITAE of 0.1048.
- PI-PSO: proportional gain of 0.512, integral gain of 20.92, overshoot of 1.579%, and ITAE of 0.0151.
During a simulated load step at 0.4 seconds, source-current amplitude changed as load demand changed, while the current waveform remained sinusoidal in the reported results. The active power changed with the load, and reactive power continued to oscillate around zero. Together, the reported measures suggest that PSO tuning improved the modeled controller’s transient response as well as its harmonic-compensation result.
Practical considerations for engineering teams
For engineers evaluating SAPF control, the key distinction is between the filter’s power stage, the DPC switching logic and the DC-link voltage regulator. The PSO method tunes the PI regulator; it does not replace the DPC switching table or the filter hardware. Keeping these functions clear helps teams assess which part of a control design is responsible for a particular result.
The reported findings are based on a numerical model, not field testing. Before applying a tuning approach in an operating plant, system integrators and engineering teams would need to validate it against their own electrical network, load profile, converter implementation and operating conditions. Simulation can help compare controller designs, but it does not establish performance for every installation.
Conclusion
A DPC-controlled shunt active power filter can compensate for load-generated harmonic and reactive current by selecting inverter switching states from instantaneous power errors. Tuning the DC-link PI controller with PSO offers a structured way to search for gain values, and the cited simulation reported lower current THD, overshoot and ITAE than conventional PI tuning. For plant and system-design decisions, these results provide a comparison point—not a substitute for application-specific validation.