Novel Control Strategy to Improve Photovoltaic System Performance under Low Irradiance Level

1Department of Electrical Engineering, National Taiwan University of Science and Technology, No. 43, Sec. 4, Keelung Rd., Da'an Dist., Taipei City 106, Taiwan (R.O.C.) 2Circular Line Operations Division, Taipei Rapid Transit Corporation, No. 7, Lane 48, Sec. 2, Zhongshan N. Rd., Zhongshan Dist., Taipei City 104, Taiwan (R.O.C.) 3Department of Electrical Engineering, National Chin-Yi University of Technology, No. 57, Sec. 2, Zhongshan Rd., Taiping Dist., Taichung City 411030, Taiwan (R.O.C.)


Introduction
Solar power greatly reduces environmental pollution as it does not emit greenhouse gases or air pollutants. Solar power has been widely used in various fields, including power banks, smartphones, cars, and wireless networks, as well as industrial power (e.g., motor and water pumps) and building power. (1)(2)(3)(4) Solar power has provided much convenience in people's lives.
However, solar power has two major shortcomings. First, its output is poor during cloudy days with an irradiance level of less than 150 W/m 2 . (5) Second, solar power output depends greatly on climatic factors such as irradiance level and temperature. (6) A maximum power point tracking (MPPT) controller is thus indispensable for enhancing solar power efficiency.
Numerous algorithms have been extensively investigated and are available for the MPPT of solar energy. (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17) The hill climbing (HC) and perturbation and observation (P&O) algorithms have been widely used as they are simple and cost-effective. (10,16,17) However, they have four main drawbacks: (1) slow convergence near the maximum power point (MPP), (7) (2) when the irradiance is steady, the tracked power point oscillates around the MPP and causes power loss in the system, (7)(8)(9)11,12) (3) on cloudy days (irradiance level < 150 W/m 2 ), both algorithms have difficulties in accomplishing MPPT, (7,9,12) (4) when the irradiance changes rapidly, both algorithms are prone to divergence. (8,10) Here, we propose a novel MPPT algorithm for solar power generation systems, which is based on the P&O algorithm integrated with irradiance change per unit area threshold control (ICPUATC). The proposed algorithm can achieve quicker and more precise detection of the MPP, and improve the efficiency of photovoltaic (PV) modules on cloudy days. Experimental results for a PV module under irradiance levels of 480, 140, and 65 W/m 2 confirm that the proposed algorithm has better performance, higher reliability, and greater suitability than the HC algorithm.

HC Algorithm
The HC algorithm is popular owing to its simplicity and low cost. By using sensors to measure the output voltage (V pv ) and output current (I pv ) of a PV module, the HC algorithm is able to calculate its output power (P pv ). It involves disrupting the duty cycle of a power converter to affect P pv for MPPT. (18) However, the disruptive character of the algorithm causes the actuating point to oscillate around the MPP, resulting in suboptimal power output. In addition, the algorithm compares the power points only with the adjacent power points. Hence, when an actuating point is trapped near a local minimum, divergence may occur (Fig. 1). (13) Figure 2 shows a flowchart of the P&O algorithm, which has a simple structure and uses voltage and current sensors to measure the output voltage V pv and output current I pv of a PV module for calculating its output power P pv . The P&O algorithm is based on the P pv − V pv characteristic curve of a PV module with the characteristics of oscillation of the tracked power point around the MPP and the subsequent system power loss. (19) To improve the P&O technique in the case of continuous disturbance, the strategies proposed in this work are as follows: (1) When the slopes of dP pv and dV pv are 0, a fixed point of the track is located at the MPP and the duty cycle D is fixed. (2) ICPUATC is added to the P&O technique.

Relationship between R pv −1 and I pv
When the proposed algorithm executed in MPPT reaches the MPP and enters ICPUATC, then Eq. (1) for the impedance R pv of the PV module is satisfied. Equation (1) is the basis of the proposed algorithm to calculate the real irradiance level. (1) Equation (2) can be used to obtain I pv .
( ) In  If the values of R pv −1 and I pv fall on line A in Fig. 4, they correspond to line A.1 in Fig. 5 Fig. 4, they correspond to line B.1 or C.1 in Fig. 5, respectively, and G can be calculated using Eq. (4).

ICPUATC
In this work, the ICPUATC value was set at 1 W/m 2 . This value was selected after testing; a very large value led to a slow response of MPPT and a very small value led to a fast response of MPPT. When MPPT could not run at the MPP, power loss occurred. Figure 6 shows a flowchart of the proposed algorithm, where V pv (n) is the present voltage, V pv (n − 1) is the previous voltage, I pv (n) is the present current, P pv (n) is the present power of the PV module, P pv (n − 1) is the previous power of the PV module, G(n) is the present irradiance, G(n − 1) is the previous irradiance, R pv −1 (n) is the present equivalent conductance, R pv −1 (n − 1) is the past equivalent conductance, and D is the PWM duty cycle during MPPT.  Figure 7 illustrates the system framework of a boost converter using the proposed algorithm, which is used to compare the proposed algorithm and the HC algorithm. (20)(21)(22)(23) The boost converter had an inductance L of 1 mH and an output capacitance C out of 220 μF. The microcontroller unit (MCU) used in this work was the 18F452 model manufactured by Microchip Technology. This actual test circuit used an optical coupler as a V pv sensor and a current transducer as an I pv sensor.

Experimental Results
The performance of the proposed algorithm under various irradiance conditions was verified by conducting experiments under irradiance levels of 480, 140, and 65 W/m 2 . It was observed that the proposed algorithm yielded better results than the HC algorithm, as shown in Figs. 8-10 and Table 1. Figure 8 shows the waveforms of the PV output when the proposed and HC algorithms were employed in the MPPT controller with the PV module under the irradiance of 480 W/m 2 and a temperature of 35 ℃. Figure 8(a) shows the test results of the proposed algorithm. The proposed MPPT algorithm was activated at time t 0 . After the V pv and I pv sensors transmitted measurement signals to the MCU, the MCU performed MPPT calculation, then sent out a PWM signal to control the power MOSFET S 1 (Fig. 7). In contrast to the HC algorithm, at time t 1 , the proposed method could maintain a fast and stable performance at the MPP.
When the conductance of the PV module R pv −1 was 0.13 S, then I pv, Curve2 was 4.8 A from Eq. (3), compared with the measured I pv of 4.7 A. Because 1.025·I pv, Curve2 > I pv > 0.975·I pv, Curve2 , I pv falls on line B (Fig. 4), which corresponds to line B.1 (Fig. 5). Equation (4) is used to calculate the irradiance G, which is 480 W/m 2 . Figure 8(b) shows the test results of the HC algorithm, for which MPPT was activated at time t 0 . After the proposed MPPT algorithm was activated at time = t 0 and the V pv and I pv sensors delivered measurement signals to the MCU, the MCU executed the MPPT calculation, then sent out a PWM signal to drive the power MOSFET S 1 (Fig. 7). It was observed that the tracked power point of the HC algorithm oscillated around the MPP, which led to power loss of the system. The experimental results in Table 1 verify that the proposed algorithm yielded higher MPPT efficiency and greater output power P o than did the HC algorithm. Figure 9 illustrates the waveforms of the PV output when the proposed and HC algorithms were used in the MPPT controller for the PV module under irradiance of 140 W/m 2 and a temperature of 25 ℃. Figure 9(a) shows the test results of the proposed algorithm. The proposed MPPT algorithm was activated at time t 0 . Following the transmission of measurement signals from the V pv and I pv sensors to the MCU, the MCU performed the MPPT calculation then sent out a PWM signal to drive the power MOSFET S 1 (Fig. 7). In contrast to the HC algorithm, at time t 1 , the proposed method maintained a fast and stable performance at the MPP.
When the conductance of the PV module was R pv −1 = 0.041 S, I pv, Curve2 was 1.47 A according to Eq. (3), whereas the measured value I pv was 1.54 A. Because I pv > 1.025·I pv, Curve2 , R pv −1 and I pv fall on line A (Fig. 4) and correspond to line A.1. From Eq. (4), the irradiance G is 140 W/m 2 . Figure 9(b) shows the test results of the HC algorithm. At time t 0 , MPPT was activated but diverged for 10.8 s, preventing the algorithm from operating at the MPP. However, at time t 1 , the MPPT continued to diverge. In Fig. 9(b), the MPP diverged because of the difficulty in completing the MPPT procedure under a low irradiance level (140 W/m 2 ), resulting in the confinement of the actuating point at a local minimum. The experimental results in Table 1 verified that the proposed algorithm yielded higher MPPT efficiency, greater output power P o , and a shorter convergence time than did the HC algorithm. Figure 10 shows the waveforms of the PV output when the proposed and HC algorithms were used in the MPPT controller for the PV module under irradiance of 65 W/m 2 and a temperature of 20 ℃. Figure 10 PWM signal to control the power MOSFET S 1 (Fig. 7). In contrast to the HC algorithm, at time t 1 , the proposed method could maintain fast and stable performance at the MPP. When the conductance of the PV module was R pv −1 = 0.0189 S, then I pv, Curve2 was 0.68 A from Eq. (3), whereas the measured I pv was 0.72 A. Because I pv > 1.025·I pv, Curve 2 , R pv −1 and I pv fall on line A (Fig. 4), corresponding to line A.1 (Fig. 5). From Eq. (4), the irradiance G is 65 W/m 2 .   Figure 10(b) illustrates the test results of the HC algorithm. At time t 0 , MPPT was activated but diverged for 16.4 s, preventing the algorithm from operating at the MPP; however, MPPT continued to diverge at time t 1 . The MPP in Fig. 10(b) diverged because of the difficulty of completing the MPPT procedure under a low irradiance level (65 W/m 2 ), resulting in the confinement of the actuating point at a local minimum. The experimental results in Table 1 confirm that the proposed algorithm yielded higher MPPT efficiency, greater output power P o , and a shorter convergence time than did the HC algorithm.

Conclusions
The proposed algorithm in this study operated consistently at the MPP, even when the PV module was under low irradiance, thus avoiding the power loss due to oscillations around the MPP, which occur in the HC and P&O algorithms. Experimental results confirmed that the proposed algorithm yielded higher MPPT efficiency for PV modules under irradiance levels of 480, 140, and 65 W/m 2 than did the HC algorithm. Whereas the HC and P&O algorithms exhibited difficulties in MPPT for PV modules under low irradiance levels (<150 W/m 2 ), the proposed algorithm quickly and precisely facilitated MPPT for the PV module under the irradiance level of 65 W/m 2 and substantially improved the module performance under poor weather conditions. Therefore, the proposed algorithm was shown to be highly reliable when used with PV modules under various irradiance levels. More specifically, the proposed algorithm is suitable for long-term low irradiance levels and areas with rapidly changing conditions. However, its performance for MPPT under partial shadow conditions should be improved in the future. Currently, when the irradiance level is insufficient and a solar power generation system cannot supply enough power to the load, users have to switch back to batteries or diesel generators for their power supply. The proposed algorithm can maximize the available power from PV modules in harsh environments to better meet the load demand.