Scope
This work develops an AI-driven dynamic pricing model for electricity consumption. It examines various tariff structures—block rate, seasonal, time-of-use (including critical peak pricing), and real-time pricing—to incentivise consumers to adjust their usage patterns, thereby reducing peak demand and associated costs.
Summary
In (Holtschneider et al, 2013) the authors concern models for incentivising consumer behaviour in the context of electricity consumption. A major issue in this domain is the high demands in peak hours which can lead to higher costs for the generators to respond with the appropriate supply. Dynamic pricing can be particularly useful for changing consumer patterns, and it can come in various forms. Some of the prevalent ones in the literature are the following:
- Block rate tariffs. Here the price varies based on the quantity of consumption. This means that there can be multiple price tiers, characterised by the consumption.
- Seasonal tariffs. Such schemes offer different rates in different seasons so as to reflect varying demand levels during a calendar year (e.g. in countries with a high tourist season).
- Time-of-use (TOU). These are pre-declared tariffs that vary during different times of the day. A particular form of a TOU scheme is referred to as Critical peak pricing (CPP), where the prices are higher during a few peak hours and discounted during the rest of the day.
- Real time pricing (RTP). With this policy, the prices change at regular intervals (like intervals of 1 hour).
Further, it proposes a real time pricing scheme using an AI-driven approach. Namely, they train a neural network to learn the consumers’ behaviour and the way that they will respond to price changes. The behaviour of the consumers depends on both their motivation to respond to dynamic pricing as well as on their ability to reduce the consumption in some time interval and increase it in another one. The training is done via a 2-layer feed-forward neural network with sigmoid hidden neurons. The outer process then is an optimisation problem of finding the right prices that would steer the population towards the desirable pattern of electricity consumption. The authors use the heuristic of Mean Variance Mapping Optimisation (MVMO), which outperforms other similar heuristics for such problems. At the end the model outputs the mean value and the standard deviation of the changes in load that are expected because of the dynamic pricing scheme.
Relevance for EXIGENCE
(Holtschneider et al, 2013) paper provides a way to steer the population’s behavior into consuming more energy at off-peak hours. This certainly leads to a more efficient operation of the electrical grid, with less costs for the generators. At the same time, it serves as a representative example of how dynamic pricing can be expected to incentivise users to adapt to a desirable behaviour that is overall more energy-efficient.
Holtschneider, I. Erlich Ι. “Optimization of Electricity Pricing Considering Neural Network based Model of Consumers’ Demand Response”, IEEE Computational Intelligence Applications in Smart Grid (CIASG), 2013.
