Evaluating the market splitting determinants: evidence from the Iberian spot electricity prices

Abstract

Assess determinants on market splitting behaviour of Iberian electricity markets. Logit and non-parametric models to express market splitting probability response. Explanatory variables: wind, hydro, thermal and nuclear power; ATC and demand. Results: increase of market splitting probability with higher availability of low marginal cost electricity. Coordination policies governing both interconnections and renewables deployment. a b s t r a c tThis paper aims to assess the main determinants on the market splitting behaviour of the Iberian electricity spot markets. Iberia stands as an ideal case-study, where the high level deployment of wind power is observed, together with the implementation of the market splitting arrangement between the Portuguese and the Spanish spot electricity markets.Logit and non-parametric models are used to express the probability response for market splitting of day-ahead spot electricity prices as a function of the explanatory variables representing the main technologies in the generation mix: wind, hydro, thermal and nuclear power, together with the available transfer capacity and electricity demand. Logit models give preliminary indications about market splitting behaviour, and then, notwithstanding the demanding computational challenge, a non-parametric model is applied in order to overcome the limitations of the former models.Results show an increase of market splitting probability with higher wind power generation or, more generally, with higher availability of low marginal cost electricity such as nuclear power generation.The European interconnection capacity target of 10% of the peak demand of the smallest interconnected market might be insufficient to maintain electricity market integration. Therefore, pro-active coordination policies, governing both interconnections and renewables deployment, should be further developed.

It is windy in Denmark: Does market integration suffer?

“…Following the methodology described in Ref. [36], expanded to a new multi-interconnected electricity market, market splitting behaviour was modelled through logit and non-parametric models estimating the probabilities of its occurrence. In the estimated models the introduction of electricity flows and market splitting binary variables of surrounding interconnected bidding areas introduce an additional complexity in relation to the models used in Ref.…”

“…In the estimated models the introduction of electricity flows and market splitting binary variables of surrounding interconnected bidding areas introduce an additional complexity in relation to the models used in Ref. [36], where the interconnection between Spain and France was not considered. In the estimated models the probability response for market splitting of day-ahead spot electricity prices is expressed as a function of wind power generation shares, electricity demands, five interconnection electricity flow shares and five market splitting binary variables.…”

“…[35] through the use of multinomial logit and three stage least square models, reporting that the probability of interconnection congestion increases with high wind power generation exiting a bidding area and decreases with high wind power generation in the destination bidding area. For Iberia, [36] through a non-parametric approach, found that increasing wind power generation, or furthermore, increasing low marginal cost generation has a clear influence on market splitting, increasing its probability.…”

“…Following [36], expanded to a new multi-interconnected electricity market, logit and non-parametric models are herein used to express the probability response for market splitting of day-ahead spot electricity prices as a function of wind power generation share, electricity demand interconnection cross-border flows and market splitting of adjacent bidding areas. Logit models contribute with preliminary indications on market splitting behaviour, in spite of the known specification limitations.…”

“…These limitations are subsequently overcome with the use of non-parametric models as demonstrated in Ref. [36].…”