Least-cost Distribution Network Tariff Design in Theory and Practice.

• Author: Schittekatte, Tim

1. INTRODUCTION

   Technological breakthroughs on the consumer side are challenging the use of volumetric distribution network charges ([euro]/kWh). Specifically, volumetric charges with net-metering, implying that a consumer's network charges are proportional to its net consumption from the grid over a certain period (e.g. a month), are deemed inadequate given the massive deployment of solar Photovoltaics (PV). Consumers with solar PV pay significantly lower network charges but still rely on the distribution grid as much as before. This means that if cost recovery is respected, consumers that have not installed solar PV would have to contribute more.

   There is no easy fix for distribution network tariff design. Regulators in many European countries are thinking of suspending net-metering and moving more towards capacity-based ([euro]/kW), fixed network tariffs ([euro]/connection) or a combination of both (CEER, 2017). However, many practitioners as well as academics, e.g. Abdelmotteleb et al. (2017), Batlle et al. (2017), Passey et al. (2017), Pollitt (2018), Perez-Arriaga et al. (2017) and Simshauser (2016), warn against possible issues constraining the implementation of improved or more efficient distribution tariffs. In this paper, we go one step further by demonstrating quantitatively how such constraints affect distribution network tariff design. We focus on two often-discussed constraints which are of a different nature: implementation issues with cost-reflective charges and fairness in the allocation of network costs among consumers.

   To capture the impact of these two constraints on network tariff design in this new reality with active consumers investing in Distributed Energy Resources (DER), it is indispensable to consider how consumer incentives change as a function of network tariff design. Therefore, we introduce a game-theoretical model which closes the loop between network tariff design, incentives for active, self-interest pursuing consumers, and the aggregate effect of consumer actions on the total network costs which need to be recovered by the network charges. Although the rise in active consumers is rightly welcomed, the model takes into account the fact that it can also be a double-edged sword. On the one hand, the more consumers have the ability to react to price signals, in this case network charges, the more welfare gains can be made from efficient consumer behaviour as an alternative to the historical practice of 'fit-and-forget' (Ruester, Schwenen, Batlle, & Perez-Arriaga, 2014). On the other hand, the more significant negative welfare impacts can result if these price signals are badly designed and are 'guiding' consumers in the wrong direction. In that case, the network charges avoided by active consumers will simply be transferred to more vulnerable passive consumers who see their electricity bill increase. The more consumers have the possibility to react to price signals, the more important it becomes to get the network tariff design right.

   The mathematical structure of the presented model is a bi-level optimisation problem which is reformulated as a Mathematical Program with Equilibrium Constraints (MPEC). At the upper-level, a regulator sets the distribution network tariff. Besides volumetric charges, the regulator has two other 'traditional' network tariff design options: capacity-based and fixed network charges, or can opt for a combination of the three. The regulator anticipates the reaction of the consumers represented at the lower-level and the network tariff is determined in a way that the total system costs (including network costs, retailer energy costs and DER investment costs by consumers) are minimised. The regulator is subject to the constraint that the total network charges collected need to equal the network costs. (1) Modelled consumers can be passive or active. Passive consumers are assumed not to react to prices; active consumers pursue their own self-interest, i.e. their objective is to minimise their cost to satisfy their electricity demand. Active consumers have the option to invest in two technologies: solar PV and batteries.

   Using a numerical example, we illustrate a trade-off between cost-efficiency, for which the proxy is the total system costs, and fairness, for which the proxy is the increase in grid charges for passive consumers compared to a baseline. We find that some cost-efficiency can be sacrificed to limit the distributional impact resulting from network tariff redesign, and we show how this trade-off is impacted by the implementation issues with cost-reflective network tariffs. However, our main finding is that if the regulatory toolbox is limited to the three 'traditional' tariff design options, it will be hard to design a distribution network tariff that is cost-reflective and future-oriented, while at the same time also fair in the allocation of costs between active and passive domestic consumers. We argue that other, more creative, regulatory tricks are needed to combine and satisfy different policy objectives.

   The paper is structured as follows. In Section 2, we discuss the two considered constraints a regulator faces when designing the distribution network tariff and include relevant literature. In Section 3, we introduce the modelling approach. In Section 4, the setup and data for the numerical example are introduced. In Section 5 and 6, the two considered tariff design constraints are introduced, their modelling implication is described, and the results for the numerical example are presented to gain insights into their impact on network tariff design. In Section 7, we discuss the results and derive policy implications. Lastly, a conclusion is formulated and future work is proposed.

2. PRACTICAL CONSTRAINTS WHEN REDESIGNING THE DISTRIBUTION NETWORK TARIFF

   Perez-Arriaga et al. (2017) discuss and Abdelmotteleb et al. (2017) show with simulations and numerical examples that in a new world with active consumers the least-cost distribution network tariff consists of a forward-looking-peak-coincident capacity charge plus a fixed charge. (2) If the capacity-based charge is computed as the incremental cost of the network divided by expected load growth, the tariff is cost-reflective; consumers will make optimal choices with regard to the tradeoff between their consumption levels and grid reinforcements. A fixed network charge complements the capacity-based charge so as to collect the remaining residual network cost in a non-distorting manner.

   However, there are many difficulties which constrain the implementation of this theoretical optimal tariff. A first constraint relates to the implementation difficulties with cost-reflective tariffs. In practice, so-called cost-reflective tariffs are only a proxy for the actual cost driver(s) in distribution grids because it would be too complex to consider all of them or because we simply lack the necessary information. Gomez (2013) describes how a distribution network is more difficult to oversee than a transmission network as it involves a much larger number and a wider variety of equipment and components. Cohen et al. (2016) use actual load and load growth data to show that grid usage in California is very heterogeneous. They also show that the costs of accommodating incremental demand/injection can be very location specific. Passey et al. (2017) analyse a dataset of 3,876 residential consumers in the Greater Sydney Area in Australia and observe that demand profiles and the timing of the network peaks vary widely across networks and at different voltage levels, depending on the mix of consumers connected. Designing a truly cost-reflective capacity-based charge is a challenging task. The coincident-peak of a distribution system, identified as the main network cost driver, is hard to target. Targeting the wrong network peak implies an efficiency loss; for example, DER adoption can be under- or over-incentivised without resulting in much change in the total grid costs.

   Perez-Arriaga et al. (2017) and Pollitt and Anaya (2016) agree that from an efficiency point of view, a network tariff with very fine temporal and locational granularity is optimal. Examples are critical peak-pricing (mainly temporal) or even user-by-user charges as an extreme case (temporal and locational). However, such dynamic charges with fine locational granularity are hard to attain in the current context. This is mainly true due to a lack of information about the network flows in real-time, requiring significant investments in IT infrastructure. Moreover, even if the distribution network became extremely 'smart', the implementation constraint could persist, as in most countries regulation requires that a uniform distribution tariff should be in place on a regional level or per area operated by a Distribution System Operator (DSO) (European Commission, 2015). This regulatory requirement is mainly based on arguments of simplicity and predictability for the consumer. Therefore, in this work, we limit ourselves to the application of the three 'traditional' tariff design options: volumetric charges ([euro]/kWh), capacity-based ([euro]/kW) and fixed network charges ([euro]/connection). Besides simplicity and predictability, fairness is an important regulatory requirement (e.g. Batlle et al., 2017; and Neuteleers et al., 2017), thereby leading us now to the second considered constraint in this paper.

   There is a fear that network tariff reforms, which aim to increase cost-efficiency, may result in an unfair allocation of the network costs, i.e. passive, often smaller or poorer, consumers would see their electricity bills increase. Pollitt (2018) notes that under certain conditions, it can be optimal from an efficiency point of view to recover a large share of the network costs through fixed network charges: when an over-dimensioned network in place, there is low load growth, there is a limited possibility to fully disconnect from...