In order to minimize greenhouse gas emissions and operational cost related to the ICT-infrastructure the power management method of a Radio Base Station (RBS) or Base Transceiver Station (BTS) at Ericsson is studied and modelled. The major load demand at RBS’s are due to uplink and downlink data traffic. Ericssons RBS’s contains power resources and energy storage systems that are controlled by a power controller module in order to meet load demand. Such power resources may be AC-grid, generator and photovoltaic cells etc. Using a certain power resource entails a distinct cost and emission. This thesis investigates the influence that the power controller logic has on cost and emission using two frameworks. One framework models the power resource management in the form of a hybrid dynamical system. Using this framework a new management policy is investigated and compared with the current management policy. The second framework investigates the usage of MPC for controlling the power sources during operation. Using this framework a third management policy is introduced and compared with the current implementation. The performance of the three management policies are compared by simulating a standard RBS operation scenario. The results show that during a RBS scenario where the grid will never fail the third management method using MPC outperforms the other two management policies. In the RBS scenario that the grid may fail the second management method outperforms the current implementation. The implication of these results may be that during a scenario of no grid failure it is more advantageous to decrease emission and cost by using a supervisory control method. During a scenario where grid failure may occur using the second management method, where battery is used as more than just a back-up unit, is more advantageous for decreasing emission and cost. / För att minimera utsläpp av växthusgaser och driftskostnader inom ICT-infrastrukturen studeras och modelleras effekt-hanteringen för en radiobasstation (RBS) eller Base Transceiver Stations (BTS) på Ericsson. Det stora belastningsbehovet hos en RBS beror på upplänk och nedlänksdatatrafik. Ericssons RBS innehåller kraftresurser och energi lagringssystem som styrs av en effektkontrollmodul för att möta lastbehovet. Sådana kraftresurser kan vara AC-nät, generator eller solceller etc. Användandet av en viss kraftresurs medför en distinkt driftkostnad och utsläpp. Detta examensarbete undersöker effekten som kontrollmodulens logik har på kostnad och emission genom två tillvägagångssätt. Ett tillvägagångssätt modellerar effekt-hanteringen i form av ett hybrid dynamiskt system. Med detta tillvägagångssätt utreds en ny resurs förvaltningspolicy och jämförs med den nuvarande implementerade policyn. Det andra tillvägagångssättet undersöker användningen av MPC för att styra effekt-hanteringen under drift. Detta tillvägagångssätt leder till en tredje förvaltningspolicy som jämförs med den nuvarande implementerade policyn. Prestandan av de tre förvaltningspolicyerna jämförs genom att simulera olika scenarion av RBS-drift. Resultaten visar att under ett RBS-scenario där nätet aldrig kommer att förloras överträffar den tredje förvaltningspolicyn som använder MPC, de andra två förvaltningsmetoderna. I ett RBS scenario där nätet kan förloras överträffar den andra förvaltningsmetoden den nuvarande implementeringen. Resultaten tyder på att under ett scenario utan nätfel är det mer fördelaktigt för att minska utsläpp och kostnader genom att använda den tredje förvaltningsmetoden. Under ett scenario där nätfel kan inträffa är den andra förvaltningsmetoden, där batteriet används som mer än bara en reservenhet, mer fördelaktig för att minska utsläpp och kostnader.

This thesis deals with chemical plants constituted by parallel batch-continuous production lines with shared resources. For such plants, it is highly desirable to have optimal operation schedules which determine the starting times of the various batch processes and the flow rates of the continuous processes in order to maximize the average plant productivity and to have a continuous production without interruptions. This optimization problem is constrained by the limitation of the resources that are shared by the reactors and by the capacities of the various devices that constitute the plant. Such plants are "hybrid" by nature because they combine both continuous-time dynamics and discrete-event dynamics. The formalism of "Hybrid Automata" is there fore well suited for the design of plant models. The first contribution of this thesis is the development of a hybrid automaton model of the chemical plant in the Matlab-Simulink-Stateflow environment and its use for the design of an optimal periodic schedule that maximises the plant productivity. Using a sensitivity analysis and the concept of Poincaré; map, it is shown that the optimal schedule is a stable limit cycle of the hybrid system that attracts the system trajectories starting in a wide set of initial conditions. The optimal periodic schedule is valid under the assumption that the hybrid model is an exact description of the plant. Under perturbations on the plant parameters, it is shown that two types of problems may arise. The first problem is a drift of the hybrid system trajectory which can either lead to a convergence to a new stable sub-optimal schedule or to a resource conflict. The second problem is a risk of overflow or underflow of the output buffer tank. The second contribution of the thesis is the analysis of feedback control strategies to avoid these problems. For the first problem, a control policy based on a model predictive control (MPC) approach is proposed to avoid resource conflicts. The feedback control is run on - line with the hybrid Simulink-Stateflow simulator used as an internal model. For the solution of the second problem, a classical PI control is used. The goal is not only to avoid over- or under-filling of the tank but also to reduce the amplitude of outflow rate variations as much as possible. A methodological analysis for the PI controller tuning is presented in order to achieve an acceptable trade-off between these conflicting objectives.