As a critical component in photovoltaic systems, solar inverters play a pivotal role in converting direct current from solar panels into grid-compatible alternating current. The conversion efficiency of these solar inverters directly impacts the overall energy yield of photovoltaic power plants, making it a paramount concern for researchers and engineers. In this paper, I delve into the study of modulation strategies aimed at enhancing the efficiency of three-level solar inverters, with a particular focus on Discontinuous Pulse Width Modulation (DPWM) algorithms. Through detailed analysis and experimental validation, I demonstrate how DPWM can significantly reduce switching losses and improve performance compared to conventional Space Vector Pulse Width Modulation (SVPWM). The findings presented here are based on extensive simulations and tests conducted on a 50kW string solar inverter platform, underscoring the practical applicability of DPWM in real-world solar inverter applications.
The topology of a typical two-stage solar inverter, as employed in many string solar inverters, consists of a front-end BOOST circuit for DC voltage elevation and a rear-end three-level inverter bridge. This structure ensures efficient power conversion, but its performance is heavily influenced by the modulation technique used. In this context, I explore two prominent modulation strategies: SVPWM and DPWM. Both methods have distinct characteristics that affect the switching behavior, loss distribution, and overall efficiency of solar inverters. By understanding these nuances, we can optimize solar inverter designs for higher energy harvest.
Space Vector Pulse Width Modulation (SVPWM) is a continuous modulation technique widely adopted in three-level solar inverters due to its high DC-link voltage utilization and ease of digital implementation. In SVPWM, the reference voltage vector is synthesized using the nearest three vectors principle, resulting in a seven-segment switching sequence per carrier cycle. From a carrier-based perspective, SVPWM can be equivalently realized by injecting a zero-sequence voltage component into the sinusoidal modulation waves. Specifically, for a three-phase system with modulation waves \(V_x\) (where \(x = a, b, c\)), the zero-sequence voltage \(V_o\) is given by:
$$V_o = -0.5 \times (V_{\text{max}} + V_{\text{min}})$$
where \(V_{\text{max}}\) and \(V_{\text{min}}\) are the maximum and minimum phase voltages in each sector. The modified modulation waves \(V_x’ = V_x + V_o\) transform from pure sinusoids to saddle-shaped waveforms, as illustrated in modulation diagrams. This approach helps in centering the modulation waves, but it also introduces continuous switching actions throughout the fundamental period, leading to higher switching losses in solar inverters.
Neutral-point potential balancing is a crucial issue in diode-clamped three-level solar inverters, as voltage imbalances can cause waveform distortion and damage components. For SVPWM, a common control method involves injecting a zero-sequence voltage based on the instantaneous phase currents and voltage deviation. The adjustment term \(\Delta d\) is calculated as:
$$\Delta d = \frac{C \times f_{\text{sw}} \times \Delta V}{V_{\text{dc}} \times I_x}$$
where \(\Delta V\) is the voltage difference between the upper and lower DC-link capacitors, \(f_{\text{sw}}\) is the switching frequency, \(V_{\text{dc}}\) is the total DC-link voltage, \(C\) is the capacitance, and \(I_x\) are the phase currents. When \(\Delta V = 0\), the formula simplifies to:
$$\Delta d = \frac{C \times f_{\text{sw}}}{V_{\text{dc}} \times I_x}$$
This ensures that the average neutral-point current over a switching cycle is zero, mitigating low-frequency fluctuations. The final modulation waves are then decomposed into positive and negative components for driving the inverter switches.
In contrast, Discontinuous Pulse Width Modulation (DPWM) is a discontinuous modulation strategy that reduces switching actions by clamping the modulation waves to the carrier boundaries for a portion of the fundamental period. Among various DPWM types, DPWMA is particularly advantageous for solar inverters due to its low common-mode voltage and balanced switching loss reduction. The essence of DPWM lies in superimposing a zero-sequence component that forces the modified modulation waves to saturate at ±1 or 0 levels. Define \(V_{hx}\) and \(V_{lx}\) as the distances from the original modulation waves \(V_x\) to the upper and lower carrier boundaries, respectively:
$$V_{hx} = \begin{cases}
1 – V_x & \text{if } V_x \geq 0 \\
1 + V_x & \text{if } V_x < 0
\end{cases}$$
$$V_{lx} = \begin{cases}
V_x & \text{if } V_x \geq 0 \\
-V_x & \text{if } V_x < 0
\end{cases}$$
For DPWMA, the zero-sequence voltage \(V_{o,\text{DPWMA}}\) is computed as:
$$V_{o,\text{DPWMA}} = \text{sign} \left( \min\left( \min(V_{hx}), \min(V_{lx}) \right) \right) \times \min\left( \min(V_{hx}), \min(V_{lx}) \right)$$
where the sign is positive if \(V_o\) is derived from \(V_{hx}\) and negative if from \(V_{lx}\). After injection, the modulation waves exhibit periods of no modulation—approximately one-third of the grid cycle—thereby cutting switching actions by one-third compared to SVPWM. This directly translates to lower switching losses in solar inverters, enhancing overall efficiency.
Neutral-point potential control under DPWM follows a similar principle but with a simplified algorithm. By monitoring the upper and lower DC-link voltages, different zero-sequence injection formulas are applied based on the voltage deviation. A flowchart-based approach is often used, where the control logic switches between modulation strategies to maintain balance. This method is computationally efficient and well-suited for digital implementation in solar inverters.
To quantitatively compare the two modulation strategies, I analyze the loss distribution in solar inverters, focusing on power semiconductor devices and filters. The total losses in a solar inverter arise from conduction losses, switching losses, and passive component losses. For a three-level T-type inverter, the losses can be broken down per phase. Assume the phase voltage and current are sinusoidal:
$$v(t) = V \sin(\omega t)$$
$$i(t) = I \sin(\omega t – \phi)$$
where \(V\) and \(I\) are peak values, and \(\phi\) is the power factor angle. The modulation index \(M\) is defined as:
$$M = \frac{2\sqrt{2} V_{\text{rms}}}{V_{\text{dc}}}$$
where \(V_{\text{rms}}\) is the RMS output voltage and \(V_{\text{dc}}\) is the DC-link voltage.
The conduction losses in IGBTs and diodes depend on their on-state characteristics. For an IGBT, the collector-emitter voltage \(V_{ce}\) can be approximated as:
$$V_{ce} = V_{ce0} + r_{ce} \cdot I_c$$
where \(V_{ce0}\) is the threshold voltage and \(r_{ce}\) is the slope resistance. Similarly, for a diode, the forward voltage is:
$$V_f = V_{f0} + r_f \cdot I_f$$
where \(V_{f0}\) is the diode threshold voltage and \(r_f\) is its slope resistance. Based on the switching sequences in different power factor intervals, the conduction losses for each device can be integrated over the fundamental period. For instance, the conduction loss of IGBT T1 in phase A is given by:
$$P_{\text{cond,T1}} = \frac{1}{2\pi} \int_0^{\pi – \phi} \left[ V_{ce0} + r_{ce} I \sin(\omega t – \phi) \right] \cdot I \sin(\omega t – \phi) \cdot M \sin(\omega t) \, d(\omega t)$$
Solving this integral yields:
$$P_{\text{cond,T1}} = \frac{I}{2\pi} \left\{ V_{ce0} \left[ (\pi – \phi) \cos \phi + \sin \phi \right] + 2 r_{ce} I (1 + \cos \phi)^2 \right\}$$
Similar expressions can be derived for other devices. To summarize the conduction losses across different operating points, I present the following table based on calculations for a 50kW solar inverter using Vincotech M669-F28 IGBT modules at 700V DC-link and 16kHz switching frequency:
| Device | Conduction Loss Expression (General Form) | Typical Value at Rated Power |
|---|---|---|
| T1, T4 | \(P_{\text{cond}} = \frac{I}{2\pi} \left[ V_{ce0} A(\phi) + r_{ce} I B(\phi) \right]\) | 45 W |
| T2, T3 | \(P_{\text{cond}} = \frac{I}{2\pi} \left[ V_{ce0} C(\phi) + r_{ce} I D(\phi) \right]\) | 38 W |
| D1, D4 | \(P_{\text{cond}} = \frac{I}{2\pi} \left[ V_{f0} E(\phi) + r_f I F(\phi) \right]\) | 22 W |
| D2, D3 | \(P_{\text{cond}} = \frac{I}{2\pi} \left[ V_{f0} G(\phi) + r_f I H(\phi) \right]\) | 18 W |
Switching losses, however, are more significant in high-frequency solar inverters. The switching energy per cycle for an IGBT is typically obtained from datasheets as a function of current and voltage. The total switching loss \(P_{\text{sw}}\) for a device can be expressed as:
$$P_{\text{sw}} = f_{\text{sw}} \cdot \left( E_{\text{on}} + E_{\text{off}} \right)$$
where \(E_{\text{on}}\) and \(E_{\text{off}}\) are the turn-on and turn-off energies. These energies are often corrected using coefficients to account for real operating conditions. The correction factors are as follows:
| Component | Current Coefficient \(K_i\) | Voltage Coefficient \(K_u\) | Temperature Coefficient \(C_t\) |
|---|---|---|---|
| IGBT | 1.0 | 1.4 | 1.00 |
| Diode | 1.2 | 1.1 | 0.95 |
The switching loss for a phase can be generalized using the Switching Loss Function (SLF). For continuous modulation like SVPWM, the SLF is constant over the fundamental period, whereas for DPWM, it varies. Specifically, for DPWMA, the switching loss relative to CPWM is given by:
$$\text{SLF}_{\text{DPWMA}} = \frac{0.577}{M} \quad \text{for} \quad 0.7 < M < 1.15$$
This implies that at high modulation indices, DPWM reduces switching losses by nearly 50% compared to SVPWM in solar inverters. To illustrate, I compute the switching losses for both strategies across different power levels. The results are summarized below:
| Power Level (kW) | SVPWM Switching Loss (W) | DPWM Switching Loss (W) | Reduction Percentage |
|---|---|---|---|
| 10 | 120 | 65 | 45.8% |
| 25 | 280 | 150 | 46.4% |
| 50 | 520 | 270 | 48.1% |
| 75 | 750 | 390 | 48.0% |
Filter losses in solar inverters primarily come from inductors and capacitors. The inductor losses consist of copper losses (due to winding resistance) and core losses (due to magnetic hysteresis and eddy currents). For a 50kW solar inverter with 0.5mH filter inductors per phase, the losses at rated power are approximately 146.7W total, with 19.7W copper loss and 29.2W core loss per phase. These values scale with current and frequency, and they remain largely unaffected by the modulation strategy, as the current harmonics are similar in both cases. However, DPWM may slightly alter harmonic spectra, influencing additional losses in capacitors and other passive components.
To validate the theoretical analysis, I conducted experiments on a 50kW three-level string solar inverter testbed. The setup includes a photovoltaic simulator providing DC input, with grid voltage of 480V, DC-link voltage of 750V, DC-link capacitance of 1.4mF, filter inductance of 0.5mH, and filter capacitance of 15µF. Both SVPWM and DPWMA algorithms were implemented on a digital signal processor, and performance metrics such as efficiency and total harmonic distortion (THD) were measured under various load conditions.
The output voltage and current waveforms for SVPWM show continuous switching patterns, while DPWM exhibits clear periods of no modulation, confirming the reduced switching activity. The grid current THD for DPWM is consistently lower than that for SVPWM across the power range, as evidenced by spectral analysis. For instance, at 50kW output, the current THD with SVPWM is around 3.2%, whereas with DPWM, it drops to 2.5%. This improvement stems from the optimized switching sequences that minimize harmonic injection into the grid.
Efficiency measurements were taken from 10% to 100% of rated load. The results demonstrate that DPWM enhances conversion efficiency at all load points, with the most pronounced gains at medium and light loads. The following table compares the efficiencies at key operating points:
| Load Percentage (%) | SVPWM Efficiency (%) | DPWM Efficiency (%) | Efficiency Gain (Percentage Points) |
|---|---|---|---|
| 20 | 96.2 | 97.0 | 0.8 |
| 50 | 98.0 | 98.5 | 0.5 |
| 80 | 98.5 | 98.7 | 0.2 |
| 100 | 98.3 | 98.5 | 0.2 |
The efficiency improvement is directly attributable to the reduced switching losses in DPWM, as predicted by the loss analysis. At light loads, where conduction losses are less dominant, the switching loss reduction plays a more significant role, leading to higher relative gains. This makes DPWM particularly beneficial for solar inverters operating under partial shading or varying irradiance conditions, where the inverter often runs at non-rated power.
Harmonic performance is another critical aspect for grid-connected solar inverters. I measured the current THD across the power spectrum and found that DPWM consistently outperforms SVPWM. The lower THD values with DPWM indicate better power quality and reduced filtering requirements, which can further lower system costs and improve reliability. The harmonic spectrum up to the 50th order shows that DPWM attenuates higher-order harmonics more effectively, thanks to its discontinuous nature that spreads switching noise differently.
In conclusion, the adoption of DPWM algorithms in three-level solar inverters offers substantial benefits in terms of efficiency and harmonic performance. By reducing switching actions by one-third, DPWM cuts switching losses significantly, leading to efficiency gains of up to 0.8 percentage points under light loads and around 0.2-0.5 percentage points across the typical operating range. Moreover, the simplified neutral-point balancing control and lower THD make DPWM a robust choice for modern solar inverters aiming for high energy yield and grid compliance. Future work could explore hybrid modulation schemes that adaptively switch between SVPWM and DPWM based on operating conditions to further optimize solar inverter performance. This research underscores the importance of advanced modulation techniques in enhancing the viability and sustainability of photovoltaic energy systems.
The insights gained from this study are applicable not only to string solar inverters but also to central and modular solar inverter designs. As the demand for efficient renewable energy conversion grows, continuous innovation in modulation strategies will remain key to unlocking higher efficiencies in solar inverters. I encourage further investigation into loss modeling, thermal management, and real-time implementation challenges to fully harness the potential of DPWM in next-generation solar inverters.