Search

A multi-resolution approach to dynamic programming is presented, which reduces the computational effort of solving multistage optimization problems with long horizons and short decision intervals. The approach divides an optimization horizon into a series of subhorizons, discretized at different state space and temporal resolutions, enabling a reduced computational complexity compared to a single-resolution approach. The method is applied to optimizing the operation of a residential energy storage system, using real 1-min demand and rooftop PV generation data. The multi-resolution approach reduces the required computation time, allowing optimization to be rerun more frequently, increasing the robustness of the receding-horizon-control approach to forecast errors. In an empirical study, this increases the cost-saving offered by a 2 kWh behind-the-meter battery energy storage system by 120% on average, compared to an approach using a single fine-grained resolution.

Energy storage systems have the potential to deliver value in multiple ways, and these must be traded off against one another. An operational strategy that aims to maximize the returned value of such a system can often be significantly improved with the use of forecasting — of demand, generation, and pricing — but consideration of battery degradation is important too. This paper proposes a stochastic dynamic programming approach to optimally operate an energy storage system across a receding horizon. The method operates an energy storage asset to deliver maximal lifetime value, by using available forecasts and by applying a multi-factor battery degradation model that takes into account operational impacts on system degradation. Applying the method to a dataset of a residential Australian customer base demonstrates that an optimally operated system returns a lifetime value which is 160% more, on average, than that of the same system operated using a set-point-based method applied in many settings today.

The combined challenges of a high fire risk and poor vehicle access in mountainous regions have led planners and emergency management authorities to consider non-traditional alternatives to complete evacuation of a region under threat. Community fire shelters have been put forward as one such alternative; however, their benefits remain contested. In this paper a series of experiments are designed in the Dandenong Ranges in Australia and presented to elucidate the relationship between shelters and community safety in wildfire scenarios. Our approach utilises a multi-stage simulation workflow to evaluate the outcome of 64 shelter configurations subject to three different fires. When compared to a scenario without shelters, some shelter configurations result in up to 10% reduction in the median exposure count, while some other configurations increase it. It is found that the efficacy of shelters strongly depends on the relative location to the ignition point and the trend of fire progression. The most effective shelters are identified for the specific fires that we simulate. The results demonstrate that sophisticated modelling and simulation is necessary for decision makers to determine a beneficial shelter placement strategy that remains effective across a number of likely wildfire spread scenarios.

The economical operation of many distributed energy assets relies on effective on-line control, which in turn often requires forecasts to be made. To produce and evaluate forecasts, the error metric by which one measures forecast accuracy must be selected. A new method is presented which customizes a forecast error metric to a given on-line control problem instance, in order to improve the controller's performance. This method is applied to the real-time operation of a battery with the objective of minimizing the peak power drawn by an aggregation of customers over a billing period. In the empirical example considered, customizing the forecast error metric to each problem instance, improved performance by 45% on average, compared to a controller provided with a forecast of the same type, but trained to minimize mean-squared-error. Error metric customization is made possible by two newly proposed parametrized error metrics. The proposed method can be applied to any on-line optimization problem which requires a point forecast as an input, and which can be accurately simulated ahead of time. The method is likely to be most effective in applications where forecasting errors are quite high, as in these applications the choice of forecast error metric significantly affects the forecasts which are produced.

Much research is being conducted into the potential value of residential energy storage. The temporal resolution of the analyses carried out in these studies is typically driven by the available data, which is often only at 30-minute or 1-hour intervals. This study uses higher temporal resolution data to examine the effect of input time-series resolution on the value determined for residential storage. In the case study considered, an analysis carried out at a 30-minute interval underestimates the cost-saving delivered by 5kWh of residential energy storage by 17% on average, compared to the same battery analyzed at a 1-minute interval. The sensitivity of storage value to temporal resolution is found to vary significantly from one customer to the next. A method for improving estimates of the realtime value of energy storage using coarser time-series data is introduced. Finally the impact of temporal resolution on storage technology selection is also evaluated.

Distributed energy technologies, such as residential energy storage, embedded generation, and microgrids, are likely to play an increasing role in future energy systems. Getting the most value from these distributed assets is often dependent on the ability to optimize their operation in a distributed manner. This distributed optimization, in turn, calls for effective short-term forecasts of the output of small-scale generating assets, and the demand of small-scale aggregations of users. This paper introduces the integration of data-driven forecasting and operational optimization methods into a single model, avoiding the need to explicitly produce forecasts. The method is tested against two empirical energy storage operational optimization problems, the minimization of peak energy drawn by a small aggregation of customers, and the minimization of the energy costs of a collection of households which have rooftop PV systems. The integrated forecasting and operational optimization approach performs well at the peak demand minimization problem for intermediate-sized aggregations (50 residential customers or more), while an approach with separate forecasting and optimization performed better on the energy cost minimization problem. These results suggest that the integrated approach can be effective in applications where (i) forecasting difficulty is intermediate, and (ii) the exact operational optimization formulation can be well approximated by a data-driven model trained on a small fraction of the available forecast training data.

This paper presents and empirically evaluates two approaches to accounting for forecast uncertainty when attempting to optimize the operation of a residential battery energy storage system. Data-driven methods are used for forecasting, and dynamic programming, within a receding horizon controller, is used for operational optimization. The first method applies a discount factor to costs incurred at later intervals in a deterministic dynamic programming control horizon, provided with point forecasts. In the second approach probabilistic (scenario) forecasts are generated using Lloyd-Max quantization of the distribution of forecast errors, to allow the use of a stochastic dynamic programming formulation. These methods are applied to maximizing the cost-savings delivered from a residentially owned and operated battery, using a case-study of residential consumers with roof-top PV systems in New South Wales, Australia. It is found that scenario forecasts can offer an 8% increase in annual cost-savings, on average, when using a univariate multiple linear regression forecast.

In this research we outline an approach to the provision of services in the presence of power constraints and energy constraints. The primary focus is power and energy constraints in heterogeneous computing environments. These environments are descriptive of the computing environments present in many organisations. Our goal in this research is to explore smarter methods of operating in power and energy constrained environments to maximise the total benefit to an organisation from the services these resources provide.

Smart grids utilise omnidirectional data transfer to operate a network of energy resources. Associated technologies present operators with greater control over system elements and more detailed information on the system state. While these features may improve the theoretical optimal operating performance, determining the optimal operating strategy becomes more difficult. In this paper, we show how a decision tree ensemble or ‘forest’ can produce a near-optimal control strategy in real time. The approach substitutes the decision forest for the simulation–optimisation sub-routine commonly employed in receding horizon controllers. The method is demonstrated on a small and a large network, and compared to controllers employing particle swarm optimisation and evolutionary strategies. For the smaller network the proposed method performs comparably in terms of total energy usage, but delivers a greater demand deficit. On the larger network the proposed method is superior with respect to all measures. We conclude that the method is useful when the time required to evaluate possible strategies via simulation is high.

The frequency of adjustments to the supply temperature set-point in district heating networks influences the overall operating cost in two ways: adaptability to changes in network conditions; and availability of time for determining an appropriate response. In this paper we investigate this trade-off and show that operating costs can be reduced through appropriate control period selection. Eleven hypothetical scenarios spanning five different networks have been simulated under a variety of control policies. When compared to a constant supply temperature strategy, the best policy in each case achieved a loss reduction of 5–24%.

Wildfires are particularly dangerous in areas where communities colocate with regions of dense vegetation. Early detection helps minimise response time and community impact, with networks of wireless sensors widely accepted as the best available early warning solution. However, financial constraints often cause sensors to be spatially distributed in a sparse and random (or pseudouniform) manner. This paper presents a new approach to sensor placement by employing maps of wildfire impact. Such maps pinpoint ignition loci that lead to more destructive fires and hence, locations where early identification is essential. We leverage IBM evacuation planner (EVA) to generate these maps from a pipeline of simulation components including: fire progression, evacuee behaviour and traffic simulation. Accordingly, these yield insights into potential community impact, and from them, we propose and evaluate two algorithms for sensor placement. The effectiveness of our approach is demonstrated through a case study in Mount Dandenong, Victoria, Australia.

The majority of renewable energy sources are non-dispatchable, meaning that it is not possible to control when and how much power they produce. For non-dispatchable renewable energy sources to meet a greater proportion of global electricity demand, the industry must develop and implement strategies that directly address the intermittency challenge. This paper considers electrical storage and transmission assets as alternative means of matching non-dispatchable generation with non-deferrable demand. It seeks an optimal combination of storage and transmission assets for a simplified representation of Australian population centres, assuming that demand is met entirely with solar PV generation. This problem is solved using a Mixed Integer Linear Program. Under the baseline assumptions it is found that the optimal (lowest cost) solution has significant quantities of storage in all load centres, as well as transmission assets installed over large distances. The storage selected was 10–15% Li-ion batteries by energy; with the remainder being pumped hydro storage.

There are currently numerous heuristic algorithms for combinatorial optimisation problems which are commonly described as nature-inspired. Parallels can certainly be drawn between these algorithms and various natural processes, but the extent of the natural inspiration is not always clear. This chapter attempts to clarify what it means to say an algorithm is nature-inspired. Additionally, we will discuss the features of nature which make it a valuable resource in the design of successful new algorithms. Not only does nature provide processes which can be used for optimisation, but it is also a popular source of useful metaphors, which assist the designer. Finally, the history of some well-known algorithms will be discussed, with particular attention to the role nature has played in their development.

This paper proposes a model for metaheuristic research which recognises the need to match algorithms to problems. An empirical approach to producing a mapping from problems to algorithms is presented. This mapping, if successful, will encapsulate the knowledge gained from the application of metaheuristics to the spectrum of real problems. Information theoretic measures are suggested as a means of associating a dominant algorithm with a set of problems.