First order multivariate Markov chain model for generating annual weather data for Hong Kong

It becomes popular to use computer simulation technique to evaluate the performance of energy utilizations in buildings. The hourly weather data as simulation input is a crucial factor for the successful energy system simulation, and obtaining an accurate set of weather data is necessary to represent the long-term typical weather conditions throughout a year. This paper introduces a stochastic approach, which is called the first order multivariate Markov chain model, to generate the annual weather data for better evaluating and sizing different energy systems. The process for generating the weather data time series from the multivariate Markov transition probability matrices is described using 15-years actual hourly weather data of Hong Kong. The ability of this new model for retaining the statistical properties of the generated weather data series is examined and compared with the current existing TMY and TRY data. The main statistical properties used for this purpose are mean value, standard deviation, maximum value, minimum value, frequency distribution probability and persistency probability of the weather data sequence. The comparison between the observed weather data and the synthetically generated ones shows that the statistical characteristics of the developed set of weather data are satisfactorily preserved and the developed set of weather data can predict and evaluate different energy systems more accurately.

► We develop a method to generate weather data by first order multivariate Markov chain.
► Unlike TMYs and TRYs, the method is based on 15-years actual data of Hong Kong.
► The method does not need to pre-assume weighting factors of weather variables.
► Data generated is more suitable for evaluating or sizing mixed systems.
► Accuracy, statistical characteristics and persistent structure of data are better.