Automated Residential Demand-Side Management Strategies

Abstract

There is growing evidence that climate change is becoming one of the major issues facing mankind within the next century. In this context, the electricity sector has been receiving increasing attention over the past decade due to the large amount of carbon it emits. Demand-side management - controlling overall power demand - has been heralded as a potential solution to this problem and while it is common in industry today, residential demand response remains elusively untapped. This is because residential electricity consumers are generally unwilling to sacrfice comfort for often meagre incentives, and the difficulty of controlling such a complex system increases its cost. This work specifically targets these barriers to adoption, proposing an entire demand- side management topology, control algorithm, and novel way of viewing controllable loads. To start, a uni ed load model is introduced which sees all controllable loads represented as batteries with time-varying power and energy values bounded by consumer comfort limits. Then, an algorithm called NES (Net Energy Stored) control is introduced that controls loads modelled in this fashion using a simple, scalable control system. The algorithm operates without making any predictions, greatly enhancing its ability to respond to unexpected events such as reserve provision, real-time pricing, unpredictable local generation, and the load profiles of diverse, human consumers. A practical system was built on low-cost microcontrollers and the algorithm was implemented thereon. A method of validating demand response algorithms is also proposed in this thesis, employing a Monte-Carlo simulation of 100 statistically representative 7-house New Zealand communities using a bottom-up load modelling tool based on large behavioural studies. Several scenarios were explored, covering electric vehicle adoption, varying allowable temperature deviation in hot water cylinders and different wind generation profiles. Then an optimal scenario operating with perfect foresight was implemented as a Mixed Integer Program and used as a benchmark. Both the optimal savings and performance of NES control varied significantly across the 100 communities and across the scenarios, notably performing very well in the most volatile scenarios. On average, NES control obtained 57.4% of the theoretical optimal savings assuming the use of existing residential loads. Increasing the allowable temperature range of hot water cylinders using a mixing valve on the output allowed NES control to extract significantly more savings relative to the uncontrolled case, extracting 68.2% of the optimal benefits if the water temperature is allowed to fluctuate between 55-85 C. At 30% electric vehicle adoption Net Energy Stored (NES) extracts 66.4% of the optimal value. Under different wind profiles NES extracts between 38.8% and 66.0%, showing its sensitivity to the timing and magnitude of wind power. Combining the extremes of all of these scenarios, NES control achieves 63.6% of the optimal results. NES performs marginally worse under a real-time pricing scheme, with 48.2% of the optimal savings extracted in the status quo situation and 56.3% with all technologies combined. Overall these results are promising because NES control does not make any predictions of load, price or wind while its benchmark is afforded `perfect foresight'. Furthermore NES control has been successfully implemented on an ultra-low-cost, ultra-low-bandwidth embedded communications system which performed almost identically over a networked control system as it did in simulation. NES control is therefore a low-cost, readily implementable solution that is well suited to the complex problem of residential demand-side management.