Probabilistic operational envelopes for demand response of new low carbon loads on low voltage networks

Low Carbon Technology (LCT) such as renewable energy generation, electric vehicles (EV) and heat pumps (HP) are expected to play a major part in decarbonising their respective sectors. The anticipated impact of LCT load suggests that continuing to plan and operate LV networks with a passive approach and common load assumptions is no longer appropriate. Increasing penetrations of LCT load on LV networks have been shown to introduce not only transformer loading issues but also voltage and thermal constraints across the network. Aggregation and active control of LV connected LCT are areas of significant development and proposed architectures specify the need for system operators to provide technical validation of demand response at various stages in the market process.
This paper presents a framework for probabilistic LV network analysis that has been developed as a planning tool to address the challenges of increased LCT load penetration on LV networks. An Active Demand (AD) optimisation algorithm is implemented within the framework to calculate a probabilistic profile of the maximum LCT load. A Maximum Sensitivity Selection (MSS) algorithm is combined with numerical calculation of a network sensitivity matrix to conduct the optimisation. The MSS approach is used to order network nodes based on the impact of load addition to the conditions on the rest of the network and to succesively add LCT load to the network based on the ordered list. Applied within the probabilistic analysis framework this approach produces a probabilistic profile of the maximum LCT load that can be supported without causing network limit violations.
A case study of a real UK network with 50% EV penetration is presented to demonstrate the application of the framework. The network comprises a single 750kVA transformer, five radial feeders and serves 226 households. Synthetic household load profiles are constructed from known transformer load profiles and a Monte Carlo simulation model of domestic EV use and availability is used to model time-specific probabilities of charging load for EV owning households. For a winter Friday 24 hour period, probabilistic maximum EV load profiles are calculated and forecast EV load probabilities are assessed in terms of their likelihood to cause network limit violations. Results indicate that potential for network limit violations is concentrated during traditional peak hours.
The paper concludes that forecast probabilities of LCT load can be assessed in terms of their risk of causing network limit violations. By specifying an acceptable level