Abstract:
Forecasted electrical loads are the core information required in many
processes, especially in power system expansion planning process. Several
techniques are available to forecast the loads of an area for a future period. The
pattern of load growth over the past are the basic requirements for all of these
techniques. However, in an isolated rural area historical load data may not be
available either because, the electricity might not be a source of energy in the past
in that area or the available data may not be representative ones, rather suppressed
load demands. First, research presents a technique [27, 28] offorecasting loads for
an isolated area. This technique identifies the factors responsible for the
development of electrical loads. The correlation of each of these factors with the
load growth is determined. The technique selects one or more areas with the
characteristics similar to those of an isolated area. The forecasted loads of this area
are derived from the selected area should be such that its history must be known. It
develops a relation between the load dependent factors of the selected area/areas
with those of an isolated area. The forecasted loads of the selected area/areas using
the relation developed from the load dependent factors. The methodology is
applied to an isolated area of Bangladesh. However, there is no conceptual
difficulty in applying the methodology to forecast the loads of any isolated area.
To realize a relationship between the load dependent variables and the
demand which may be highly nonlinear, a relatively new technique, using artificial
neural network, capable of solving a non-linear mapping, is investigated. The
inherent parallelism of neural network captures the past trend of the event and make projection of the best guess. The study shows encouraglllg result on
forecasting the loads of an isolated area, using the neural network method.
