MPPT methods under partial shading conditions

Most of conventional MPPT methods assume that there is only one single maximum power point on the Power-Voltage characteristic of a PV source, However, when a PV array is under partial shading conditions, its characteristic may exhibit multiple maximum power points (see Fig. 1-9). These MPPT methods have limited reliability in dealing with the partial shading conditions [46].

Thirteen commercial inverters with MPP trackers were tested with real partially shaded PV arrays in [47]. All of these trackers have very high MPP tracking accuracy under stationary conditions. But when the P (V) curve exhibits two local maximum power points, only 7 out of 13 could track the global MPP. When P (V) curve has three local

second maximum power point. It causes up to 70% losses because the MPPT algorithms are unable to track the global MPP of the PV arrays.

Quite a few researchers [48]–[54] have worked on global MPP tracking schemes for PV systems operating under partial shading conditions.

Kobayashi et al. [48] proposed a two-stage MPP tracking scheme for a PV system under partial shading conditions. In the first stage, the operating point moves to the intersection of the I (V) curve and the load line Rpm=Vpm/Ipm, where Vpm is 80% of the open circuit

voltage and Ipm is 90% of the short circuit current. In the second stage, the operating point

moves to the MPP using conventional MPPT scheme. However, in the first stage, the operating point is not always close to the global MPP [49] [50], and it may miss the global MPP in second stage. Also, online open circuit voltage and short circuit current measurements require extra components in the PV system and increase the complexity of the control scheme. Although the measurements can be done within a short period (1ms), it will cause some power loss during the measurements.

Full range search based MPPT schemes have the advantage of easy implementation and high reliability under partial shading conditions [50]-[53].

There are two full range MPP scan methods: Power-Current curve scan [51] and Power– Voltage curve [50] scan. P (I) or P (V) curves are scanned periodically, in this way it can always find the global MPP. However, the scan interval is not based on any identified

A search based MPP tracker, designed to improve the PV system efficiency for a racing solar car, is described in [52]. The racing car is powered by a PV module and batteries. The MPP tracker includes a full range MPP searching and a conventional P&O MPPT scheme. The full range MPP searching is triggered when the operation conditions are changed, such as starting the car, system restart, the car entering shadow, etc. Also the full range MPP searching can be manually triggered by the driver. There is not much information on whether/how a judgment is made for the necessity of a full scan. And every full range MPP searching takes 3.8 seconds, which is quite long, and it may not be very suitable if the condition changes rapidly when driving the racing car.

An improved Fibonacci linear search algorithm has been proposed to track the global MPP under rapidly changing irradiance or partial shading conditions in [53]. For partial shading conditions, an equation of output power change rate is introduced for the detection of partial shading. If the output power change rate is greater than certain constant, the algorithm will be initialized to do a wide-range scan. However, this method only applies to abrupt changes of partial shading conditions. When the shading pattern changes slowly, e.g. the shade of a nearby building moves slowly during a day due to the slowly changing incident angle of the sunlight, this method will not be able to identify the change in power hence fail to track the global MPP.

A global MPP tracking scheme which only needs to scan part of the P (V) curve has been proposed [49] [54]. After some observations of the P (V) curves using the MATLAB model presented in [54], H. Patel et al. proposed an MPPT algorithm for PV systems operating under partial shading conditions in [49]. The authors observed that on either

monotonically. Based on this conclusion, a "global peak tracking subroutine" was proposed, which is called periodically in the main MPP tracking program. The subroutine searches the neighboring peaks of the current peak every 25 seconds, and if the power at one of the neighboring peaks is more than the current peak, the tracking algorithm will continue to search for the next peak in that direction until it comes to a peak that generates less power than the last one. It is claimed that the global MPP can be found without scanning the entire P (V) curve.

There are two drawbacks of this method. First of all, the conclusion obtained from the observations is not always true and this will be discussed in detail in section 4.1. Secondly, the search subroutine runs every 25 seconds even under uniform irradiance conditions, in which case, it has to scan the entire P (V) curve because there are no other local peaks.

In conclusion, because of the complexity of the PV electrical characteristics under partial shading conditions, a full range scan is necessary to find the global maximum power point. To reduce the number of unnecessary scans, judgment is required to detect partial shading conditions.