The electricity demand of households in the Netherlands has been growing rapidly for the last decades and will continue to grow in the near future. This is specifically the case during peak periods. High peak loads could exceed the available capacity, resulting in overloaded network components (assets) which lead to an excessive reduction in life expectancy of these assets. The present aging distribution network will not have the capacity to cope with these future peak loads. The increase of electricity demand by the end-users therefore seriously reduces the reliability and safety of the electricity distribution. This poses an important problem for the Distribution Network Operators, who are responsible for the transport of electricity, maintenance and management of the regional electricity distribution networks. The traditional method to cope with capacity availability during peak periods is to invest heavily in placing more electricity cables. However, Demand-Side Management programs using load-shifting techniques also show good potential for reducing the peak loads in the network demand pattern. Load-shifting focuses on scheduling smart household appliances from peak load periods to off-peak periods. The aim of Demand-Side Management is to increase the efficiency of the system by bringing both demand and supply to the best possible low value. To measure the effectiveness of Demand-Side Management, the Key Performance Indicators (KPIs) “Levelling Effect” (LE) and “Height of Peak loads” (HP) are used. LE measures for a day the deviation of loads from the average load of the network. HP measures the highest load (in W) that occurs during a day in the network. Both the traditional way of improving the network, and a Demand-Side Management approach will require high investments. To ensure that such investments are economically viable, DNOs should now the extent to which Demand-side Management of households will affect these KPIs. Assessing load-shifting potential by scheduling smart appliances: Because load-shifting takes place through the individual scheduling of household appliances, the focus lies on a the household’s appliances level. On this level, the irregularities of the demand pattern are important, which are caused by the simultaneous usage of household appliance. We therefore constructed a simulation model using an Agent-Based Modelling approach, which takes into account these aspects. This simulation model represents a low-voltage network with one hundred households connected to it. Each household owns appliances, which build-up the electricity demand of the household. Smart appliances are modelled as individual agents to allow the scheduling of these appliances. Non-smart appliances are combined and generate the “other-loads”. The scheduler uses a “lowest-point” principle for the scheduling of smart appliances. Furthermore, all appliances are always scheduled and they cannot be rescheduled. The model simulates the demand pattern on the network during one working. External influences (e.g. weather) are ignored. From the literature and the available data, we made a trade-off between the required accuracy and computation, and opted for a time step of 15 minutes. Simulation results: As expected, the introduction of a smart system to the network was found to level the demand pattern and lower the peaks by a maximum of 13%. In this simulation, 16% of the total demand could be shifted. Non-cooling appliances (dishwashers, washing machines and tumble dryers) represent about 8% of the total load, and cooling appliance (refrigerators and freezers) the other 8%. The rescheduling of appliances did however increase the number of excessive peak loads, which in real life could form a serious risk for overloading the network. The scheduler does not take into account the profile of the smart appliances when scheduling. Appliances with a low start demand could therefore be scheduled to a low load timeslot while the load on subsequent timeslots could increase to peak loads when the appliances reach their full demand. More advanced scheduling algorithms that also take into account the appliance demand profile should resolve this. Apart from the above mentioned aspect, a closer examination of the smartness variable also showed some additional interesting developments. As expected, better forecasting and longer operational horizons will give better results. Unexpectedly however, was the lack of effect of the scheduling schemes. This is most likely because appliances are always scheduled, which results in a lack of advantage of being first in the schedulers queue. More advanced scheduling algorithms that allow appliances not to be scheduled may resolve this. The sequencing of the cooling appliances created a layer of smart cooling load that absorbs all the small irregularities in the demand pattern. Because of their short operational time, cooling appliances may therefore successfully be used for smoothing of the network demand pattern. Scheduling appliances using a “lowest-point” principle proved to work very well, but only for non-cooling smart appliances. The multiple usages of the smart cooling appliances caused them to turn on less in low-peak periods but more on the slopes towards peak loads. This scheduling artefact is caused by the time-step of 15 minutes and a too simplistic scheduling algorithm. Too few timeslots were available for effective scheduling of these appliances. Using a smaller time step in combination with a more advanced scheduling algorithm should improve the scheduling of smart cooling appliances. However, a smaller time-step would not necessarily increase the quality of the result, this also applies to a better representation of the non-smart “other-loads” profiles. A smaller time-step would increase the variation in irregularities on the demand pattern, but the overall network demand pattern would stay the same. Although the scheduler does take into account these small variations on the demand pattern, the general network demand pattern determines the areas were the smart appliances are scheduled to. Conclusion of research Our study has shown that load shifting by scheduling smart appliances is likely to produce more levelled demand pattern. Peaks in the network demand pattern can be reduced by 13% and the gaps are filled resulting in a more levelled demand pattern. A time step of 15 minutes works well for non-cooling appliances, but it limits the effective scheduling of (the present) smart cooling appliances. But a shorter time step would not necessarily have produced better results. The overall network demand patterns, and thus the overall scheduling places of the smart appliances, will most likely stay the same. What potentially could make a difference is a more advanced scheduler. Allowing rescheduling and the possibility for smart appliance not to be scheduled could result in a higher effectiveness of the scheduling schemes and more optimal scheduling of the smart appliances.