M Michiel Nijhuis

Gaussian Mixture Based Probabilistic Load Flow For LV-Network Planning

Due the many uncertainties present in the evolution of loads and distributed generation, the use of probabilistic load flow in low voltage (LV) networks is essential for the evaluation of the robustness of these networks from a planning perspective. The main challenge with the assessment of LV-networks is the sheer number of networks which need to be analyzed. Moreover, most loads in the LV-network have a volatile nature and are hard to approximate using conventional probability distributions. This can be overcome by the use of a Gaussian mixture distribution in load modeling. Taking advantage of its radial nature and high R/X ratios, the LV-network can be analyzed more efficiently from a computation viewpoint. By the application of simplifications defined in this paper, the backward–forward load flow can be solved analytically. This allows for the direct computation of the load flow equations with a Gaussian mixture distribution as load. When using this new approach, the required calculation time for small networks can be decreased to 3% of the time it takes to generate a similar accuracy with a Monte Carlo approach. The practical application of this load flow calculation method is illustrated with a case study on PV penetration.

Scenario analysis of generic feeders to assess the adequacy of residential LV-grids in the coming decades

The energy transition will bring large changes to residential load profiles. In order to assess whether the current low voltage (LV)-grids are capable of handling these changes, future scenario analysis must be used to assess the grid adequacy. To this end, an approach is proposed, incorporating scenario-based household load modelling, LV-feeder clustering and probabilistic load flow calculations. All combinations of independent scenarios on the drivers of change in the residential load have been assessed with probabilistic load flow calculations . Generic feeders are chosen, which are representative for the LV-grid of a Dutch distribution network operator. The scenario assessment shows the chances of overloading and/or exceeding of the voltage limits for the LV-feeders over different time horizons and scenarios. The approach can also be used to assess the main scenario drivers and grid parameters responsible for changes in the future adequacy of the LV-grid.

Incorporation of on-load tap changer transformers in low-voltage network planning

The introduction of distributed generation in the low-voltage (LV) network can generate bi-directional power flows and thus voltage increases instead of decreases from consumers along the feeder towards the substation. The new generation installed at the consumer premises may induce voltage problems while the loading of the cables is still under nominal values. Conventionally possible resulting voltage violations are solved by reinforcing the network, however smart grid alternatives like voltage control in the LV network can also alleviate the network problems. LV-networks are traditionally designed with medium to low voltage transformers equipped with off-line tap changers. The addition of an on-load tap changer (OLTC) for voltage control can decrease the voltage violations in the network, however this needs to be considered within the optimisation method applied for the planning of the LV-network. In this paper the smart grid alternative of installing an OLTC in this optimisation has been performed. By assessing different OLTC control strategies under conditions with increasing distributed generation over many types of networks, the effectiveness of the OLTC becomes apparent. The OLTC is included in the optimization problem formulation by the introduction of additional voltage constraints and relaxing the constraints in the form of a penalty function. When the introduction of an OLTC is more efficient rather than the conventional strengthening of the network is demonstrated with a case study on the impacts of distributed generation.

Stochastic household load modelling from a smart grid planning perspective

Changes in the residential load warrant the investigation into more advanced planning methods for distribution grids. Smart grid alternatives require detailed information on network loading, and a risk based evaluation of different planning options, via a probabilistic approach. To limit the increased computational burden associated with this increased complexity, the required level of detail in modelling the household load is assessed. The effect of different aggregation levels of household load curves on the error in estimated voltage deviations is demonstrated, as well as the impact of varying degrees of availability for data regarding demand-side management (DSM) on the expected peak load reduction. The required level of load curve aggregation is determined depending on the feeder characteristics and the grid operators risk appetite. We show that incorporating DSM in network planning requires a high level of data availability, as the amount of expected DSM drops significantly when less measurement data is available.

Demand response: Social welfare maximisation in an unbundled energy market - case study for the low-voltage networks of a distribution network operator in the Netherlands

With the introduction of smart meters, dynamic pricing and home energy management systems, residential customers are able to react to changes in electricity prices. In an unbundled market, the energy supplier and the network operator may have conflicting interests with respect to demand response (DR) programs. As the customer participation is essential to a well-functioning DR program, it is needed to assess which DR programs offers the most benefits to customers. Two DR program options are analysed for low-voltage feeders: a program from the energy supplier based on the electricity price, and a demand response program from the network operator based on the loading of the network. Depending on the grid topology the benefits can change significantly between the two DR programs. DR from an energy supplier point of view might induce under-voltages which lead to grid reinforcements, while load shifting from a network point of view can generate higher electricity cost.

Appliance-based residential harmonic load modelling

An adequate modelling of the harmonic load is required to analyse the harmonic impact and perform propagation studies. Since more and more nonlinear loads are connected with the network, such as energy saving lamps, electric vehicles, photovoltaics systems etc., the accurate modelling of the harmonic injection of new appliances becomes more important. In this paper, a bottom-up stochastic model is proposed for the modelling of harmonic loads in residential networks. First a Markov Chain Monte Carlo approach is employed for household occupancy modelling with time use survey database. The occupancy, weather conditions, neighbourhood features and behavioural survey data are subsequently applied to obtain loading patterns of the household appliances. The harmonic spectra of various appliances are established based on measurements. Based on these, a harmonic load flow can be run to calculate the harmonic load of each household. Measurements of harmonic magnitude carried out in a residential low voltage network in the Netherlands are used to validate the proposed approach.

Integrating direct and indirect load control for congestion management in LV networks

With the energy transition, capacity challenges are expected to occur more frequently in low-voltage (LV) distribution networks. In the literature, several direct and indirect load control methods have been suggested as solutions to alleviate network congestion. Direct methods involve the network operator directly controlling appliances at the households, while indirect methods aim to motivate end-users to shift their consumption through price changes. In this research, the direct and indirect methods are combined into an integrated approach, making use of the advantages of both methods. An agent-based architecture is adopted so that distributed and computational intelligence can be combined to ensure a smooth coordination among the actors. A sensitivity- based curtailment scheme is used to incorporate the unbalanced loading condition of the LV networks. The efficiency of the proposed integrated approach is investigated through simulations in the unbalanced IEEE European LV test feeder. Simulation results reveal up to 94\% reduction in congestion by the integrated approach, while maintaining the required levels of supply in the network.

Bayesian-Inference-Based Voltage Dip State Estimation

Voltage dip state estimation (VDSE) tries to estimate the voltage dip characteristics at nonmonitored buses from measured voltage dip values at monitored buses. In this paper, the VDSE is addressed through the method based on Bayesian inference. A priori including the fault position among other grid conditions is used to estimate the residual voltage at each bus based on the measurement quantities, including their uncertainties. The dip duration is calculated with the time setting of protection system incorporating the uncertainties due to dip detection algorithm of the root mean square values. The proposed method has been applied to the IEEE 13-bus and IEEE 123-bus distribution test systems for multiple simulation scenarios, such as with or without distributed generation and different types of faults. The simulation results show good observability of the network.

Effects of power quality limits on LV-network design

For the design of the LV-network, not only the requirements with respect to the loading of the cable play an important role, but also power quality aspects such as flicker and voltage variations should be taken into account. In this paper, various requirements with respect to adequate earthing, voltage variations, flicker and overloading are investigated and compared. As the current approach with respect to the assessment of these quantities is limited, more risk based asset management approaches have been evaluated in this study. By evaluating these requirements for both rural, sub-urban and urban conditions, the differences in the driving factors for the LV-network design become apparent for geographical customer distributions. This study showed among others that the optimisation of the fuse breaking capacity based on the number of connected customers can for instance greatly increase the possible length of a LV-feeder. Moreover, for a network structure consisting of a single line the impedance based requirements, i.e. flicker and safety, are driving. If a branched network structure is used, the requirements with respect to voltage variations and overloading become more important.

Application of resilience enhancing smart grid technologies to obtain differentiated reliability

With the introduction of smart grid technologies in the distribution network the resilience of the network can be increased. The advantage of the use of these technologies is that they could be more effective than conventional technologies in creating a local increase in the resilience of the network. This approach would allow for differentiating the reliability per (groups of) customers. Different smart grid technologies are evaluated based on how cost-effectively they can increase the local reliability. The creation of microgrids in combination with local energy storage and the reconfiguration of the network by remote switching, are the two smart grid technologies which are investigated in this paper. The use of energy storage can increase the reliability of a single node without affecting the other nodes, however the costs of implementing energy storage are still similar to the cost of conventional network reinforcement methods to locally increase the reliability. The cost of using remotely-controlled switches to increase the reliability at a single node is much lower. With this method, the increase in reliability is however not only achieved at a single node, but at other nodes as well. This makes it difficult for the DNO to charge in a fair manner for this additional reliability.