Abstract
An artificial neural network (ANN) is an information-processing system
that has certain performance characteristics in common with biological
neural networks. Artificial neural networks have been developed as
generalisations of mathematical models, of human cognition or neural
biology. During the last three decades, artificial neural networks
have been extensively employed in numerous fields of science and
technology. For instance they have been used in signal processing,
medicine, pattern recognition, robotics, control, forecasting, speech
production, speech recognition, business, manufacturing, power systems
and also in the renewable energy and solar energy fields. As a computation
and learning paradigm, they are proposed as an alternative approach
for addressing complex problems. This chapter consists of two parts.
In the first part a bibliographic review of how artificial neural
networks have been used in the renewable energy field in general
and more specifically in the solar energy field is presented. This
review includes also applications of artificial neural networks in
the photovoltaic (PV) and in the solar radiation fields. Following
this review, the chapter is centred on the research results obtained
by the authors on using artificial neural networks, and more particularly
the Multilayer Perceptron (MLP), in the solar radiation and in the
photovoltaic fields. Among the many types of networks, supervised
models have consolidated as the most robust and easy to employ when
applicable. Usually, these models are implemented via feedforward
architectures such as, the Multi-Layer Perceptron. The Multi-Layer
Perceptron is the most widely used type of supervised neural network
employed for approximation tasks. One of the most appealing properties
of these neural network architectures is their potential use for
function approximation, which is due to their universal approximation
capabilities. The use of the MLP in the photovoltaic and solar radiation
fields has been carried out by the authors and is giving very satisfactory
results. The authors have used the MLP for generating the long-term
solar radiation series needed in the design of photovoltaic systems.
The MLP was also used as a very useful tool for generating Loss of
Load Probability (LOLP) curves for stand-alone photovoltaic systems.
Finally, the MLP was used to extrapolate V-I curves of PV modules
recorded at non-Standard Test Conditions to Standard Test Conditions
V-I curves. In all the above applications, the problems were solved
satisfactorily with a very simple structure for the MLP.
Links and resources
Tags