Gas is commercial product, and its quantity, that will be purchased by the customer, has to be assessed in advance, in other words the customer has to determine his consumption of the next gas day. In addition, the participants of the market have an interest in being able to forecast the consumption of the next gas day as exactly as possible, in order to avoid deficit. For these reasons in the last few years it became more and more important to be able to give as accurate forecast of the gas-consumption as possible.
My Thesis Project tries to find a solution for this problem, too. I work not with hourly gas data, but with gas day data, and it is supposed that the gas consumption of the previous day and the average daily temperature and the average daily wind speed of the given day are available at the modelling. This is not true in reality, but my purpose is that the models I work with should provide results that can be used in a real modelling. During my work ten emphasised gas transfer stations are analysed.
At the beginning of my Project after a theoretical summary a simple temperature sensitivity model is presented. Following that I come to the focus of the Project: the basic models. These are linear regressive and autoregressive models. Firstly models are built on the rough data, than by using these results data are manipulated, and on the partly new data received in that way the models mentioned above are built again.
After basic models with the same parameters a non-linear model is fitted on the data, and the relation between the results received in that way and the results received with the basic models is analysed.
On the models that performed better a new variable is brought in, in which way models are extended with a parameter peculiar to wind, and it is analysed how far the models received in that way become better than the basic models.
Finally the part of my Project about modelling is ended with cluster analyses. More than one gas transfer stations are brought in on clustering. In the light of the results received during clustering I try to explain the results received previously, and on the grounds of these I try to draw some conclusions.
As the finish of my Project the received results are summarised, as well as some ideas are outlined, which better and more precise models may be constructed with.