This thesis investigates how optimized electric vehicle (EV) charging and discharging can improve load balance and reduce overload risks in constrained power grids. A simulation- based model was developed using the Flower Pollination Algorithm (FPA) to schedule EV charging based on priority and system needs. The model was tested using real transformer data from a primary substation in Västerås under a future high-load scenario. By coordinating 200 EVs, the model achieved peak shaving and valley filling, reducing peak load from 9300 kW to 8133 kW and increasing valley utilization from 5500 kW to 6687 kW. The load factor improved from 72.8 % to 85.5 %, and the total deviation from the target load was minimized to 501 kW. The results show that smart EV charging can significantly support grid stability, especially when applied under conditional grid agreements such as those evaluated by Mälarenergi. In addition, analysis of summer load patterns revealed duck-curve behavior, where daytime charging using public EV station data could help reduce evening ramps. Together, the findings demonstrate that EVs can evolve from passive consumers to active grid-supporting resources in future energy systems.