應用於三相三階主動式中性點箝位併網換流器及LCL濾波器之FPGA數位控制器研製

Abstract

本文主旨在發展應用於三相三階中性點箝位併網換流器及LCL濾波器之FPGA數位控制器。主動式中性點箝位換流器結合了二階換流器的優點如較低的導通損失，以及傳統中性點箝位架構的優勢如較低的開關損失、輸出電壓諧波以及共模電壓等，因此相當適合分散式發電系統的應用。當換流器以脈寬調變操作時將會產生高頻諧波，必須仰賴濾波器加以抑制，LCL濾波器相較於單電感能提供較佳的諧波抑制能力因此可選用較小的被動元件，不僅可以有效降低濾波器體積亦相當適合高功率、低開關頻率的應用。採用LCL濾波器所面臨的難題為其控制器設計相當困難，若無加入適當阻尼可能導致系統不穩定；而本文採用多迴路控制架構，以三組串級而接的控制器分別調節LCL濾波器的狀態變數，於內電流迴路本文採用預測控制以達到快速的動態響應，然而預測控制對於電流迴授精準度相當敏感，採用LCL濾波器在內迴路將面臨極大的電流漣波，傳統取樣方式無法取得正確平均電流，因此本文採用多重取樣策略解決上述問題，可以有效改善系統雜訊拒斥能力亦能增加訊號雜訊比。除此之外，預測控制對於系統參數變化亦相當敏感，其中電感會因磁飽和現象而產生衰減，因此本文建立在線電感值估測器以修正電感模型錯誤並維持預測控制的動態響應與穩定性。本文以混合訊號FPGA實現數位控制系統，將多迴路控制器、脈寬調變產生器、座標轉換等以數位電路實現，並以軟體進行鎖相迴路、命令產生與資料傳輸等事項。本文以一額定功率9kW之實驗平台進行驗證，實驗驗證本文設計之控制器可以準確追隨電流命令，即使負載發生大幅度變化仍然可以維持系統穩定，且採用LCL濾波器之總電感值相較於單電感濾波器可以大幅降低65.38%而仍然維持相當低的諧波失真。

The purpose of this thesis is to develop an FPGA-based controller in application to a three-phase three-level active NPC grid-connected inverter with LCL filter. The active T-type NPC (AT-NPC) combines the advantages of two-level inverter such as low conduction losses, with positive aspects of conventional NPC inverter such as low switching losses, low common-mode voltage and superior output waveform. For grid filter, LCL filter is adopted since it provides better harmonic attenuation capability with smaller inductance compared with L filter. However, LCL filter may lead to system unstable if no proper damping method is applied. In this thesis multi-loop control structure is utilized to mitigate the LCL resonance. It involves the use of outer loop and inner loop, with the outer loop ensuring steady-state reference tracking performance and the inner loop providing fast dynamic response for system disturbance and sudden reference changes. In this thesis predictive algorithm is applied for inner current loop regulation. Predictive control has the characteristic of dead-beat response therefore it can achieve fast dynamic response with minimum distortion. However, predictive control suffers from its high sensitivity to the accuracy of feedback current. Due to the reduction of inductance in LCL filter, large current ripple will occur on inverter-side. Conventional sampling strategy may not able to retrieve average current under such condition, therefore multi-sampling strategy is applied in this thesis. By multi-sampling technique the noise rejection capability is enhanced and SNR is improved. In addition, predictive control is also sensitive to parameter variations. If magnetic saturation effect occurs the inductance will decrease, which may lead to oscillation or even unstable in predictive control. In this thesis an on-line inductance estimator is developed to identify the inductance and maintain the dynamic response of the system. The designed controllers are implemented on a mixed-signal FPGA device. Experimental verifications are carried out on a 9-kW grid-connected inverter. The results show that the grid current are under well-controlled and even a sudden reference change occurs the designed controllers can still maintain system stable. The total inductance of LCL filter has a 65.38% reduction compared with that of L filter but still remain low harmonic distortion.