Networked sensors in smart grids allow techniques like sensor fusion including: sensor similarities, as well as, sensor complementarities to be integrated to obtain new information or feature that is not measured directly. On the other hand, these techniques can be extended to get trusted readings at different correlated areas based on historical observations and their corresponding probabilistic distributions of sensors at these areas. In this paper a stochastic modelling of multivariate within the platform of cyber-physical systems has been discussed. A proposed multivariate Weibull distribution (WD) modeling is adopted to predict wind speed (WS) at a certain site given data at other correlated place(s). The proposed methodology has been implemented on some cases of study to illustrate the effectiveness of the adopted technique using bivariate or trivariate models. It has been revealed that the same methodology could be extended to any multivariate WD for any stochastic modeling problem. In this paper a comparison between the proposed trivariate, and bivariate Weibull is established to show their efficiency on estimating WS at a location that has a faulty sensor, that fails to deliver its data.