Consulting Project in the area of applied mathematics
FCE develops an innovative forecasting method for the energy demand of local network nodes. The focus is on a weekly forecasting model dealing with main periodicities in addition to short term effects.
An industry-specific point cloud decomposition paired with machine learning algorithms form the core of the forecasting model. It is possible to switch between different regressions such as auto-associative kernel methods, classical regressions ( such as logistic regression) and ﬁnally a special representative of neural networks.
Thereby it is possible to forecast power consumption for any location of a given network taking various weather inputs and public holidays into account.
The Dynamic Vehicle Routing Problem (short: DVRP) is a mathematical problem of combinatorial optimization and an extension of the Traveling Salesman problem (short: TSP). While the TSP seeks the optimal route for a fixed number of cities to travel, the DVRP considers both fixed goals and other stochastic random targets that are time dependent.
One application of the DVRP is the control of a collective taxi fleet, which has to pick up ad hoc passengers and take them to their destination. In doing so, the global optimum is always sought, considering the individual route optimization as well as a global allocation policy of all passengers on the existing shared taxis. Various restrictions can influence the optimum. The problem class of the DVRP is flexibly expandable to other routing applications, such as in department store logistics.