A new validation tool of weather forecast for engineering applications

摘要

Weather conditions are related to the power output from power stations and, at the same time, to energy consumption. Therefore, these days it becomes very interesting to be able to forecast weather conditions and to relate it to the energy generation, consumption and maintenance of each system. Normally, the probability density function of other parameters such as global solar irradiation (GSI), rain, temperature and relative humidity is defined in terms of an adequate correlation factor. For example, normal distribution should be used for temperature while Gamma distribution or Pearson Type III is preferable for precipitation. Different approaches to deterministic and stochastic weather conditions models have been applied in recent years. For example, curve fit and time series were employed for different variables, including temperature, relative humidity and wind velocity, respectively. However, most of these variables are interrelated, this being a very significant factor in developing a forecast validation method. The result showed that the partial vapour pressure under saturation conditions, defined as a single function of dry bulb temperature, showed better results than the dry bulb temperature. An hourly study of correlation and determination factors showed that during the night, morning and evening the highest correlation is obtained with relative humidity versus GSI, and the acceptability index showed the highest inverse correlation during the same time period. From 14:00 to 18:00 the better index was observed in temperature versus GSI. Finally, the partial vapour pressure in saturation conditions showed similar behavior as did temperature versus GSI. A practical case study of weather forecasting developed using neuronal networks could be validated by each respective correlation and determination factor, revealing this to be an adequate methodology to be employed in future research works.