A Reconfiguration Circuit to Boost the Output Power of a Partially Shaded PV String

Abstract

An optical isolator circuit is developed to detect and dynamically relocate the photovoltaic (PV) module under partial shading. The suggested system control structure operates in two modes. Mode 1 governs the system at global maxima (GM) by tracing the power-voltage (PV) curve. Mode 2 detects and separates all the bypassed modules from a PV string/array by means of a decentralized control and stores its power in the battery. Simulations are performed on different shading patterns to verify the efficacy of the suggested system. The results showcase that the averaged harnessed power using the proposed circuit is 25.26% more than the total cross-tied (TCT) and series-parallel (SP) array configurations. The proposed circuit does not require complex gate driver circuits and large switch counts. The circuit is scalable and can be implemented on an “N × N” array.

1. Introduction

Photovoltaic (PV) arrays are mostly used in parallel (SP) connections of PV strings to cater for the power requirements of domestic/commercial loads. A multiple number of series connected PV modules are termed the as PV string [1]. Under such circumstances, it is probable that a PV system receives different levels of irradiance due to cloudy weather, nearby structures, or even flocks of birds. This difference in irradiances causes non-uniform generation of current within a string. This non-uniformity of current is a major cause of hotspots in PV modules; hence, bypass diodes are connected to bypass the shaded modules in a PV string. Consequently, the power vs. voltage (PV) curve of a shaded PV array contains multiple peaks. The location of these peaks depends on the shading pattern and the PV module characteristics. Under such circumstances, only one peak exhibits the maximum power and is called the global maximum (GM), while the remaining peaks are termed as local maxima (LM) [2].

Consider Figure 1a which showcases a shading pattern over an arbitrary selected PV array. The resulting PV curve is shown in Figure 2. It is clear that there exists a power difference with the PV curve without shading. Further to this, it is evident that even if the array is operated at GM, the power of the bypassed PV module in a string is wasted. An earlier study found that the overall number of bypassed PV modules in a string grows when the GM comes closer to the y axis of a PV string’s current vs. voltage (I–V) curve [1]. This lost energy can be recovered if the total irradiance received on an “N × N” array is divided equally over all the PV modules as shown in Figure 1b; however, this is practically not possible. One solution is to disintegrate the bypass modules from a string and connect them to an external storage device to harness the power as shown in Figure 1c. This procedure will obviously result in harnessing more power compared to a string where the bypass modules produce no power at all. Note that the arbitrary curves shown in Figure 2 are not mathematically related to each other. The output of each case depends entirely on the shading pattern and the number of bypass modules under a given condition.

Some reconfigurable solutions to minimize the mismatch loss are available in the literature. These solutions either reconfigure the array to have an equal irradiance level, or reconnect them in TCT, HC, SP, etc., and then the one with best output power can be chosen. Occasionally, optimization methods are deployed to find a better interconnection scheme.

Methods that are based on reconfiguration are expensive, complex, and lack scalability. For example, in Reference [3], the PV array is arranged to optimize the power using PSO; however, the selection of coefficients involved in PSO, and the iteration required, makes this method expensive. In Reference [4], PSO is modified to avoid unnecessary configurations, but at the cost of a cumbersome objective function. The method in Reference [5] provides a comparison between TCT and SP, but practically both these schemes require a large number of current sensors. The requirement of a large sensory network is also a limitation for Reference [6]. In Reference [7], an estimation method for module level is designed which works on the basis of a sorting algorithm. Of course, this is a complex solution. Mathematical calculations are done for References [8,9] which require controllers with high computational power and large memory. Perhaps that is why the method presented in References [8,9,10] are based on simulation only. In Reference [11], a similar approach is presented for TCT only. Reference [12] is a good technique viz. simplicity and scalability; however, it requires irradiance sensors; the threshold of which offsets the switching operation. Moreover, matrix-based schemes such as [13] require large gate driver circuits and are therefore less reliable.

In this paper, inspired by the theme presented in a previous work by the authors in Reference [2], a reconfigurable circuit for the “n × n” PV array is presented. The proposed circuit is independent of a matrix converter; therefore, it requires reasonable circuitry. It is based on isolating the PV module with active bypass diodes, and solid-state relays can be used to realize the switches. These panels are then connected to a storage device. Any suitable GMPPT algorithm can be used with this work. The use of an optocoupler makes this circuit cost effective because of its negligible loading effect, decentralized control, and allocation of each module. It is free of complex algorithms and multiple I-V scans. The system is scalable and can be used for any size of PV array.

2. Proposed Circuit

The proposed circuit is explained in two parts: (i) the working of the optocoupler to detect bypassed PV modules and (ii) isolating the bypass PV module from the array. These two parts are explained below.

2.1. Optocoupler as a Bypass Module Detector

Inspired by our previous work Figure 3 shows an optocoupler circuit tied across a “few” PV modules in an arbitrary string [2]. The overall optocoupler circuit includes an optocoupler k817p [14], two resistors RC and RF, and a biasing supply of 5V. The bypass diode is connected across this optocoupler input via resistance RF. The internal LED of the optocoupler comes in antiparallel configuration with the bypass diode. When the PV module is active and thus, the bypass diode is OFF, then the optocoupler draws forward current IF which drives the base emitter junction of the phototransistor [15]. Consequently, collector current flows and the output point is connected to ground. In Figure 3, PV panel (b) is bypassed and the bypass diode is active. As a result, no current flows into the optocoupler and IC become zero. In such a case, all the biasing voltage appears across the output. Notice that the optocoupler works as a switch and thus, for a partial shading case, the output becomes high for the bypassed PV module only.

In Figure 3, applying KCL at node 1 relates that all the I_String flows through the bypass diode. Because panel (b) is inactive, no current flows through it. At node 2, the I_String + I_F is channelized through PV panel (a) because the bypass diode is reverse biased. At node 3, the output current is I_String − I_F.

2.2. Isolating the Bypass PV Module from the Array

Figure 4 demonstrates the practical circuit in which two switches (Q₁ and Q₂) and two diodes (D₁ and D₂) are implemented for each module in addition to the bypass diode D_b. Only one switch is active at any moment, while the other remains in cut-off mode. The output of the optocoupler is segregated into two signals (Q_g1 and Q_g2) using NOT gate and are described here for automatic and decentralized control of Q₁ and Q₂:

• If Q₁ is in the on-state it implies that the PV module operates at some V_pv value, and the bypass diode D_b is reverse biased. Thus, Q₁ offers the route for I_String to flow through the PV module, and after passing through D₁, I_String rejoins the string. At the same time, Q₂ is in the off-state and the I_String is blocked through diode D₂.
• When Q₂ is turned on, this means that the PV array does not operate on any V_pv, and it is therefore bypassed. As a result, Q₁ and D₁ quarantine the module from the PV array. This module is relocated through Q₂ and D₂, while the path for the I_String is provided by the bypass diode. The bypassed (shaded) module’s energy can be used or stored in this manner.
• The operation of Q_opt and V_B is expounded in the next section.

Figure 4. Circuit for isolating the PV module using the proposed optocoupler circuit.

Figure 4. Circuit for isolating the PV module using the proposed optocoupler circuit.

Note that either relays or semi-conductor switches such as BJT or MOSFET can realize the function of the switches Q₁ and Q₂. Furthermore, the well-known driver circuit ICs such as IR2110, etc., can be connected to the opto-IC output to quench the switches’ gating demands.

It can be inferred from the above debate that the suggested electronic circuit not only defines the status of the module without any exorbitant sensor, but also grants the decentralized control of the module’s relocation.

3. Proposed Circuit with GMPPT Tracking

A design of the system suggested is displayed in Figure 5 for one string of the entire PV system. It has two modes of operation that are explained below.

3.1. Mode 1: GMPPT Tracking

During mode 1, the main aim is to track the GMPP of the entire PV system. This is accomplished by deactivating the optocoupler’s biasing supply. At the same time, the Qsc switch is turned on for a very short time to fully discharge the scanning capacitor Csc. Thereafter, the I–V curve is scanned, and the boost converter is set at the global maximum power point. Note that during the I–V curve scanning, the Qoc switch disconnects the boost converter with the PV module. During this mode, the Q2 switch for all three modules remains off as well. During mode 1, each module forms a series connection through its corresponding Q1 and D1 switches to other modules. Once the GMPPT is set using any suitable MPPT method such as [16], the Q_opt and Q_oc switches are turned on and the energy of the entire string is transmitted to the load as shown in Figure 6.

3.2. Mode 2: Partial Shading Scenario

Since the system monitors the instantaneous power and has knowledge of the GMPPT, a threshold is imposed to detect the power difference between the two. In case the threshold is exceeded, the system takes it as an indication of partial shading. Under the partial shading condition, the system moves back to mode 1 to reset the GMPPT. Thereafter, the system enters into mode 2 and detects the PV module’s status. In Figure 5, it is assumed that the module M1b is under shading and the bypass diode D_bb is therefore active. This implies that for this particular module, the optocoupler outputs 1, and therefore, the Q_g1 becomes low and disconnects module D_bb from the string. At the same time, the Qg2 switch turns on and connects the D_bb to the battery V_B.

In mode 2, the shaded module forms a parallel connection to the battery and other modules via its corresponding Q₂ and D₂ switches. This way, the shaded modules are avoided under partial shading circumstances, and their energy is stored in the battery. It is pertinent to mention that distinct rates of irradiance on PV modules have a negligible effect in parallel connections as there is no prevalent series current condition. The battery voltage (B_V) can be set at the module V_mpp by selecting a particular battery, the value of which can be determined from the datasheet of the manufacturer. The battery can also be interfaced through an electronic power circuit.

4. Simulation Results and Comparative Study

The three settings, i.e., the suggested system, a total-cross-tied (TCT), and series-parallel (SP), are performed in PSIM. The proposed scheme is arranged in PSIM as shown in Figure 7. The comparison assessment is performed in terms of energy output between these designs. A PV array of size 10 × 3 (N_s × N_p) is used, where the specifications of each module is as follows: open circuit voltage (V_oc) = 10.78 V, short circuit current (I_sc) = 6 A, maximum power point (P_mpp) = 47.5 W, voltage corresponding to maximum power point (V_mpp) = 8.4 V, and current corresponding to maximum power point (I_mpp) = 5.66 A. Between the PV array and the 120 V battery load, a boost converter is placed, although the system works independent of the converter and converters such as [17,18] can also be used. The values of the converter parts are C_in = 100 µF, C_out = 150 µF, and L = 300 µF. Instantaneous power output of the PV array is obtained by the multiplication of voltage and current sensor output.

Figure 8 displays the power-voltage curve (PV) of the array under the impact of shading. The upper graph in Figure 8 shows the functioning of the proposed scheme, and the lower graph gives the stored power status (P_store). Initially, the PV curve is plotted, and the highlighted part in Figure 8 identifies GM. In mode 1, all opto-ICs are disabled, and the entire array PV curve organized in the SP setup is mapped. Once the PV array is settled at 5 ms on GM (=500 W), the proposed configuration control is entered in step-2, where the opto-ICs are simply activated. The bypass modules are therefore linked in parallel to the 16.5 V battery. The bypass modules can produce 178.5 W of energy, which is stored in the battery as shown in the bottom graph of Figure 8. The outcome of the simulation obviously shows the additional power storage by the suggested setup. Importantly, it requires no irradiance sensor, complicated I–V designs, search and sort algorithms, no module physical motion, and numerous I–V curve tracings.

Figure 9 displays a uniform condition of the PV array at 12 ms. The control system reaches step-1, where it disables all the opto-ICs due to the power threshold being breached. The system starts step-2 when MPP is reached at 17 ms and the opto-ICs are re-activated. The P_store is null because no module is bypassed.

Table 1. Report card of power output of the proposed scheme.

Cases	Power Extracted from PV Array (W)			%Pextra Compared to SP Configuration	%Pextra Compared to TCT
	Prop	TCT	SP
1	678.5	524.6	500	35.7	29.3
2	396.3	324.1	305.3	29.8	22.3
3	469.9	448.7	426	10.3	4.73

shows each configuration’s report card in terms of energy output. The proportion of additional power is calculated from Equation (1) below:

P_extra = (P_rec − P_conv)/P_conv × 100

(1)

where P_rec denotes the power produced by the recommended configuration and P_conv is the power generated by other conventional arrangements such as SP or TCT. In Equation (1), P_extra indicates the percentage power advantage of the proposed configuration over others.

In case-1, the suggested setup produces 35.7% and 29.3% more energy than SP and TCT settings, respectively. While TCT produces more energy than SP, it still lags by an important margin behind the energy output of the suggested system. The TCT setup PV curves against case-1 and two other instances are shown in Table 1, where 524.6 W of energy is provided in case-1. The other two examples of shading are shown in Figure 10.

In case-2, shown in Figure 11a, the effectiveness of the proposed method is proven asymmetrically (N_S ≠ N_P). The PV array of the proposed circuit outcome is shown in Figure 12. The suggested setup provides extra energy of 91 W opposed to the SP setup of 305.3 W. Note that case-2’s scenario differs from case-1, where all shaded modules at GM are segregated from unshaded modules. In case-2, when GM is tracked and due to the nature of the shading pattern as shown in Figure 12’s I-V/PV graph, the scenario occurs in a region between 2V_oc to 3V_oc where V_oc is the module’s open-circuit voltage. It means that from each string three modules must be active.

Now, S1’s shading condition is such that the 500 W/m² irradiance module determines the prevalent string current as verified by Figure 11a and Figure 12 (Diagram I-V/PV). Since the current of this module can be supported by the bottom two modules, i.e., 1000 W/m² and 700 W/m², they stay connected. At the same time, the lower two modules cannot sustain the prevalent string current because they are more shaded, with 300 W/m² and 100 W/m². The opto-ICs of just two smaller modules therefore show the bypass position and distinguish it from the string. Likewise, as shown in Figure 11a, the lower two modules of the other two strings are skipped in the system. Their harnessed 91 W energy is then deposited in the battery as shown in Figure 12’s lower graph.

In case-2, TCT derives 324.1 W of energy, which is 72.2 W less power than the system suggested. With regard to case-3, i.e., Figure 11b, the suggested setup derives again more energy than PS and TCT. Figure 13, which provides 426 + 43.9 = 469.9 W of energy, shows the operating principle of the suggested system. This is an additional power of 43.9 W and 21.2 W than the PS and TCT configurations, respectively.

5. Prototype Experimental Results

An experimental apparatus is developed for the validation of the proposed scheme. The apparatus is shown in Figure 14 along with labels. The PV array is comprised of four sub-modules, which are connected in series. The specifications of each mono-crystalline technology—model TSPM 85 sub-module is V_oc = 10.1 V, V_mpp = 8.25 V, I_sc = 6 A, I_mpp = 5.25 A, and P_mpp = 43.3 W. Solid-state relays of type DC are used as switches Q₁ and Q₂, and the energy of the bypass modules is stored in a separate battery of 16.5 V. A boost converter is installed between the PV array and the battery load of 32 V. The frequency of the converter is set at 20 kHz, and its components’ values are set as input capacitor C_in = 150 µF, output capacitor C_out = 250 µF, and inductance L = 330 µH.

Figure 15 shows the standard format of testing where the following four curves are displayed through a digital oscilloscope: I_pv (blue-line), V_pv (red-line), P_pv (green-line), and P_store (pink-line). In step-1, the PV curve is scanned for 50 ms with all the opto-ICs disabled. Notice that the PV array exhibits two LMs because of particular shading pattern on it. Once GM of 42 W is obtained through the operation of the boost converter, the proposed scheme enters into step-2. In this step, all the opto-ICs are switched on. Consequently, the power equals to 11.5 W from the bypass modules and is stored in the battery. Another test is conducted, in which the PV array is subjected to a uniform condition. Figure 16 displays the result where the PV curve exhibits a single peak. When the opto-ICs are triggered, zero power is stored as no module is bypassed during the uniform condition. The PV array is subject to another shading pattern and this time it exhibits three LMs as shown in Figure 17. Once again, the proposed scheme saves 13.1 W of more power than the SP configuration.

6. Conclusions

In this article, four switch arrangements per PV module are presented. The proposed circuit consists of two passive and two active devices which are connected across each PV module in a PV string. The proposed circuit is capable of working with relays, as demonstrated in the experiment, which makes it independent of the gate driver circuits. The proposed circuit is also capable of working under all kinds of MPPT algorithms. It only makes use of the PV modules that are bypassed through their bypass diodes. This is sensed using any commercially available optocoupler IC, which makes it a commercially viable product. The proposed two stage algorithm is easy to implement using any cost-effective microcontroller and hence adds to the advantage of the circuit. In addition to this, the presented circuitry is free from complex matrix sensors and reconfigurable PV schemes and hence presents an excellent bottom line for future work.