24h predictor of the ozone process for Basse-Normandie region using fuzzy approach

In this paper, Takagi-Sugeno (TS) approach is resorted to get two approaches of multi-model to describe the ozone generation process. The first approach is a model to describe the ozone phenomena over the Basse-Normandie region. The second one is a 24h predictor at a fixed time. In the first approach, the nonlinearity has been taken into consideration. The nonlinear feature is due to the structure of the Takagi-Sugeno approach. Different types of models that describe the ozone process have been discussed in this paper. Using different number of inputs and different number of membership functions. In the second approach, the focus is made on 17 hour, because in the Basse-Normandie region, the maximum value of the ozone occurs at this time. TS-based multi-model design necessitates the selection of influent inputs, membership functions (for the different input variables) and local models structure. The selection of influent input variables has been done according to the (hourly) correlation between the output and all input variables. The identification scheme, for the two approaches, involves the identification of antecedents part (membership functions parameters) and the consequence part using a least square type estimation algorithm to estimate the local model parameters. As the number of identified parameters grows exponentially with the number of inputs a compromise should be done between the models accuracy and complexity.