VINCULACION DEL PROGRAMA CON EL GRAFICADOR

Table 4.1 compares the explored MPPT techniques against the criteria estab- lished in Section 4.2. For each technique, one or two references are provided. Note that the cost is not included in the table due to the difficulty in estimating the likely cost of each PV system implementation. However, the systems described in the PE based approaches and those that require an increase in circuit complexity, would generally incur additional costs due to the extra power electronics components required.

In Table 4.1, the ability of the technique to reliably identify the GMPP is identified by no if it will rarely track to the GMPP, likely if the technique may converge to the GMPP, and usually if the technique can be considered to be a reliable GMPPT technique. The tracking speed is indicated as fast, slow or

varies. Where varies is indicated, shows a technique where the tracking speed is directly dependent on the choice of parameters, for instance, the step size in

Table 4.1: Comparison of MPPT techniques.

Technique Reliably Identify GMPP

Speed Speed under irradi- ance change

Steady-state oscilla- tions

Dependence on Sys- tem Specific Param- eters

Complexity Conventional MPPT techniques

Perturb and Observe and Hill Climbing [106, 128] No Varies Varies Common No Low

Incremental Conductance [120, 131] No Varies Varies Common No Low

Fractional Short-circuit Current [230] No Fast Fast No Yes Low

Fractional Open-circuit voltage [94, 231] No Fast Fast No Yes Low

Ripple Correlation Control [168] No Fast Fast No No Low

MPP Locus [102, 103] No Fast Fast No Yes Moderate

Extremum Seeking Control [169, 170] No Fast Fast No No Moderate

Sliding Mode Control [161, 162] No Fast Fast No Yes Moderate

Load current/voltage maximisation [152] No Fast Fast No No Low

Fuzzy based [61, 142] No Fast Fast No Yes High

Neural Network based [142, 150] No Fast Fast No Yes High

Bisection Search [172] No Varies Varies No No Low

β-method [179] No Fast Fast No No Moderate

Global MPPT techniques

Periodic reset [124] No Varies Varies Sometimes No Moderate

Periodic Curve Scanning [185, 186] No Varies Varies No No Moderate

Two stage methods [190, 191] No Varies Varies Sometimes Sometimes Moderate

Two stage methods (equivalent load line) [191, 192] No Varies Varies Common Yes Moderate

Observations of P-V characteristics [90] Likely Varies Varies No Yes Moderate

Refined P-V curve sweeping [22] Likely Varies Varies Sometimes Yes Moderate

Line Search (DIRECT) [20] Usually Varies Varies No Yes Moderate

Line Search (Fibonacci) [21, 198] Usually Varies Varies No Yes Moderate

Particle Swarm Optimisation [201, 202] Usually Varies Varies No Yes Moderate/High

Chaos Search [205] Usually Fast Fast No No Moderate

Power Electronics Based Approaches

Bypass diodes method [213] Usually Slow Slow No No High

Differential power processing [214] Usually Varies Varies Sometimes No Moderate

PE equaliser [215] Usually Fast Fast No Yes High

Distributed MPPT [211, 212] Usually Varies Varies Sometimes No Moderate

TEODI [220] Usually Fast Fast No Yes Moderate

the P&O method. The presence of steady-state oscillations is characterised by

no, sometimes and common. Where sometimes is indicated shows a technique which depending on the implementation may exhibit steady-state oscillations. An example of this is applying a periodic reset with the P&O technique. The dependence on system specific parameters is shown by yes, no or sometimes. The case where sometimes is indicated relates to the actual implementation for two-stage methods. Finally, the implementation is classified based on its complexity as low, moderate or high. This complexity is based on the characteristics of the technique in terms of the number of components required, potential software complexity and the ease with which the technique could be applied in a new system.

Table 4.1 clearly shows that no single MPPT technique, or approach designed for promoting global maximum power extraction under PSC, is able to meet all of the criteria defined within this chapter. In general, it can be seen that conventional MPPT techniques are usually plagued by a tendency to track a local maxima and may exhibit oscillatory behaviour around the MPP. Additionally, the simplest of these conventional techniques are limited by the definition of system dependent constants which are no longer representative of the system under PSC. Other conventional techniques designed to achieve MPP, including fuzzy and neural network approaches, add to the complexity of the tracking and may still fail under non-uniform conditions. Attempts to improve conventional tracking performance for GMPPT often increases technique complexity and cannot always guarantee GMPPT.

PE based approaches are shown to increase the energy yield under PSC but the cost involved in such systems may outweigh the benefit.

While some techniques are specifically designed to reliably track to the GMPP, these methods often increase the cost and complexity of the system and fre- quently rely on system dependent parameters which makes it difficult to adapt the technique to use in other systems.

4.8

Conclusion

This chapter has demonstrated that many approaches have been proposed to provide superior operation of PV systems experiencing PSC. In general, conventional MPPT techniques cannot achieve GMPPT without significant modification and techniques designed specifically for this purpose may not always guarantee GMPPT and may involve additional system dependent con- stants or circuit elements. While an abundance of maximum power extraction strategies have been proposed for PV systems, no single technique can meet all required objectives of a universal global maximum power extraction process.

In anticipation of real environmental conditions it is essential that a GMPPT technique is able to quickly track to the GMPP with minimal implementation complexity. The technique proposed in Chapter 5 is able to perform satisfacto- rily against the criteria defined in this chapter and is an attempt at creating a universal global maximum power extraction process.

Chapter 5

Proposed

Global

Maximum

Power Point Tracking Technique

Based on Simulated Annealing

5.1

Introduction

As presented in Chapter 4, the existing MPPT methods have limitations when tracking the maximum power of a PV system under non-uniform environmental conditions. In particular, conventional methods will often track to a local maxima due to the inherent structure of the technique such as P&O and Hill Climbing (HC) approaches. Techniques designed specifically to achieve GMPPT often increase the cost and complexity of the implementation through the use of additional circuitry and complex algorithms that must be implemented on higher cost processors. Many GMPPT strategies involve optimising the performance for specific conditions such that the method may be unable to track the GMPP under all possible environmental conditions that arise, or may rely on system dependent constants leading to a technique that cannot be easily applied to another PV system without considerable optimisation. PV systems are operated in environments where non-uniform environmental conditions, or real environmental conditions become more significant than the uniform operating conditions explored in an indoor laboratory environment. This is due to shadow from objects in the environment, dust or dirt on the module surface, module mismatch due to manufacturing tolerances, module ageing and damage with time, and the rapid changes in irradiance caused by the passage of clouds over the system. As PV systems operate in these real conditions, it is essential that an implemented GMPPT strategy can perform well under these conditions and accurately and reliably track the GMPP. Additionally, it is important that this GMPPT can be achieved without exorbitant costs and complexity to help

minimise the cost of energy associated with PV systems.

This chapter describes the proposed Simulated Annealing (SA) MPPT method. The proposed technique attempts to overcome some of the key limitations of existing MPPT techniques, as described in Chapter 4, for Partial Shading Con- ditions (PSC). In particular, it has low complexity, can easily be implemented on a low-cost microcontroller and reliably tracks to the GMPP. The performance of the proposed SA method is compared with the common P&O MPPT method and the Particle Swarm Optimisation (PSO) strategy for GMPPT in simulations in this chapter. To verify the effectiveness of the proposed SA method for practical application, an experimental implementation has also been completed.

The modelling and experimental implementation presented in this chapter shows that the SA algorithm, without a considerable amount of optimisation, can effectively converge to the neighbourhood of the GMPP within a small time frame. The technique could then be combined with a local search technique such as P&O when in the vicinity of the GMPP to track it exactly. Alternatively, once SA has located the approximate MPP, the neighbourhood size could be reduced considerably to enable the method to track exactly to the GMPP. Due to the nature of the system considered, in which the passage of clouds does not lead to each module experiencing a different level of irradiance, the main factor influencing the irradiance on each module and the PSC is the shading factor of the module which is based on the number of shaded cells within the module. The shading of the individual cells is a much slower phenomenon than the transitions in irradiance caused by the clouds, so the relative locations of the MPPs should have little variance from each sample time to the next, as has been extensively explored in Chapter 3.

The main features of the proposed SA method are summarised below with respect to the criteria established in Chapter 4:

• SA can reliably locate the neighbourhood of a global maxima.

• The SA implementation has low complexity.

• The method requires no additional circuitry and can be implemented in a low cost microcontroller.

• When optimised, the SA method can achieve GMPPT in a similar time frame or faster than the PSO method, yet with a considerably simpler implementation.

• Tracking with the SA method exhibits no oscillations in steady-state, al- though may exhibit some apparent oscillations during the searching process due to sampling different voltages.

The key advantages that the SA algorithm has when compared with other methods is that it can achieve reliable GMPPT with a limited increase in complexity beyond the common P&O method. Where the P&O method may become trapped in a local maxima, the SA method incorporates searching capabilities that allow it to escape from a local maxima. Additionally, once the algorithm has converged it remains at a constant operating voltage such that no oscillations are observed in the steady state. Unlike the PSO method, which is frequently implemented in PV systems that experience PSC, the SA method requires fewer parameters to be stored between each iteration. Other methods proposed for GMPPT often require costly processors, additional circuitry, or require knowledge of the PV system parameters or the environmental conditions as has been described in Chapter 4. The SA method is not constrained by these limitations as it can be implemented in a low cost processor, only requires circuitry to read the voltage and current and write the duty cycle (equivalent to what would be required with the P&O method), and can operate successfully without knowledge of the PV system parameters or environmental conditions.

In this chapter, the SA algorithm is firstly described and the key reasons for its good performance in global optimisation are highlighted. The SA and PSO implementations considered for GMPPT in this thesis are then described indepen- dently. Simulation results are presented to demonstrate the performance of the SA method on firstly a simple two module system and secondly, the eight module PSC system introduced in Chapter 3. Finally, experimental results are presented showing the performance of the method on the two module experimental setup from Chapter 2.

5.2

The Simulated Annealing Algorithm for