Study On Adaline Network Based Control Strategy For HVDC Hybrid Active DC Filter

During the past several decades, HVDC has become the dominating technology for long distance transmission of bulk power. However, along with the growth of voltage grade and transmission capacity of HVDC systems, increasingly serious harmonic contamination has been brought into DC transmission lines, which causes deterioration in system operation and interference to communication. This situation poses a quite serve challenge for harmonics suppression in consideration of the drawbacks of traditional passive DC filters and strict interference criteria.With the development of power electronic technology and digital control technology, hybrid active DC filters (HADFs), a combination of active filters and passive DC filters, have been applied in several HVDC projects for harmonic currents suppression. Comparing with passive DC filters, HADFs have wide frequency range, fast dynamic response, excellent compensation effects and high cost-efficiency, while take less space. As the soul of the HADF, the control strategy plays a crucial role in its performance, so a thorough study on the control strategy is of great necessity.In this dissertation, the typical topology of the HADF has been introduced, key factors for the control strategy of HADF, and both advantages and disadvantages of existing control strategies have been analyzed in detail. An adaptive linear neural (Adaline) network based self-tuning control strategy for HADF has been proposed. This control strategy, consisting of two Adaline networks as controller and identifier respectively, is an open-loop tracking control strategy, which has been proved simple, robust and accurate. It ensures optimal controlling performance regardless of the uncertainty and time-variability of the plant. In order to make the control strategy feasible for hardware, the control algorithm has been improved, and an Adaline network based predictor added. Besides, the selection of initial values of weight vectors of predictor and identifier, as well as their leaning rate has been also studied in detail. Furthermore, as an important aspect of the control strategy, analog and digital integrators for harmonic currents measuring method based on Rogowski coil have been studied. In order to keep the stability of digital integrators under non-ideal conditions, a quasi-integration algorithm has been proposed, which works properly even with a DC component in the integrated signal, and a unproper initial value for integration.Finally, the mathematical model of the controlled plant has been deducted according to the control strategy. Simulations on the MATLAB/Simulink software platform of the harmonic currents measuring method with the digital quasi-integrator and the Adaline network based control strategy for HADF under both steady states and transient states have been implemented. For better steady accuracy and dynamic response of the control strategy, an improvement with increased sampling frequency has been made. The simulation results have verified the feasibility and effectiveness of the mentioned harmonic currents measuring method with digital quasi-integrator and the Adaline network based control strategy for HADF in DC line harmonic currents measurement and suppression.