Effects of system-blind prosumers in energy models

TL;DR

This paper introduces a method to incorporate prosumer behaviors into energy models, revealing significant impacts on optimal storage capacity and system efficiency, especially under various tariff schemes.

Contribution

It develops a calibrated self-generation constraint for energy models to better represent prosumer incentives and their effects on system planning.

Findings

Optimal battery capacity increases by up to 200% with prosumer constraints.

Ignoring prosumer behavior can lead to underestimating storage needs.

The method effectively approximates prosumer decisions across tariff schemes.

Abstract

Prosumer households that generate and store electricity from rooftop PV installations play an increasing role in electricity markets around the world. As retail tariffs usually do not convey time-varying wholesale price signals to households and the rollout of smart meters is low in many countries, prosumers do not necessarily self-consume and feed-in solar electricity in a system-friendly way. The effects of such system-blind behaviours are typically neglected in energy system models, which rarely account for prosumers. In this paper, we embed a calibrated self-generation constraint into a linear capacity expansion model to approximate the incentives of prosumers to minimise their electricity bills. We apply our method to a German case study for 2030 featuring sector coupling with battery electric vehicles. We show that parametrising the self-generation constraint such that the prosumer electricity bill is as low as possible approximates prosumer decisions well for a broad range of tariff schemes. Based on this, we quantify distortions that might arise in energy models that do not account for prosumers. For our case study, we find that the optimal battery storage capacity increases by up to 200% if prosumer constraints are included. The main driver is the imperfect substitutability between home and utility-scale batteries. We conclude that energy system models could benefit from implementing this straightforward method.