ABSTRACT
The power grid has become the foundation and important component of the
development of the industrialization and the information society. At the same time, the
grid is also absorbing the achievement of the industrialization and information, which
makes all kinds of advanced technologies being applied in the power grid, and greatly
promotes the electricity grid system function. Smart grid is the inevitable trend for the
grid. In recent years, with combining with the conventional electricity technology,
communication, computer and automation technologies have been widely and deeply
applied in the power grid, which greatly improve the level of the intelligent power grid.
Load forecasting of power system in the environment of the smart grid is the main
foundation work for the management department of the power grid. Due to the more
difficulties made by utility grid connect across the provinces and new energy grid, the
traditional forecasting methods have failed to meet the requirement of accuracy, and
more of the artificial intelligence algorithms are used in the load forecasting, which
improves the load forecasting accuracy and has the extremely vital significance.
In this paper the difficulties of load forecasting brought by the development of the
power grid are discussed. And based on the complex system knowledge, the complex
characteristics of the power system load are particularly analysed. Therefore, it is
proposed that dividing an area by different sizes and the types of the load and then
getting the whole load of this area plus the load of all districts. The two different
intelligent algorithms are respectively used in the load forecasting of the same area.
The results show the feasibility of load division and the increase of accuracy by the
two algorithms than by neural network. Finally considering the real requirement of
power departments in load forecasting, the relative data of the weather are predicted by
neural network, which not only avoids the weather data loss because of the special
reasons such as bad weather or human error operation, but save unnecessary spending
for the power supply departments. The results represent the availability and feasibility
of the weather forecasting method by comparing with the data of weather department.
Key words: smart grid, complex system, short-term load forecasting,
neuron-fuzzy network, support vector machine, weather
information