The VIP project aims at an automation of the processing of firing processes in coal-fired power stations on the basis of visual information.
Global target of a processing of firing processes is the emission minimization during simultaneous efficiency maximization. A problem of automation consists among other things of the fact that only time-delayed information about the firing is available. Due to these time delays in the sensor technology the process control can achieve only suboptimal emission values and efficiency. On the other hand, directly and without delay obtained visual data from the firing could be used to improve the process control. However, these data cannot be directly used due to their spatial-temporal complexity.
Target of the VIP project is therefore to develop neural procedures which analyse this data and extract process control relevant features. This approach utilizes the ability of neural networks to learn approximations of nonlinear spatial-temporal dependencies between visual input and relevant firing process data.