Handling embedded generation

Goals

Our goal is to determine the half-hourly demand of a supplier’s customers from the P114 feed. How we do this depends on whether a supplier has remote generators, embedded generators, or a mix of both.

Terminology

• Lead Party: an organisation that is registered to manage energy transactions with the grid
• \(V_t\): half-hourly metered volume of a Lead Party at time \(t\) as reported in the P114 feed
• \(D_t\): half-hourly demand of a supplier’s customers at time \(t\)
• \(S_t\): half-hourly renewable supply at time \(t\) as inferred from a supplier’s certificates

Remote generation

The easiest situation is when a supplier exclusively buys power from ‘remote’ generators which are metered by a different Lead Party. This is generally the case for big generators that are directly connected to the high-voltage transmission network.

When supplier’s generation is exclusively remote, their metered volume is very simply their customers’ demand (with a minus sign since negative \(V_t\) connotes load):

\[\begin{equation*} D_{t} = -V_{t} \end{equation*}\]

Embedded generation

Embedded generators are metered under the same Lead Party as the supplier’s consumers. This is generally the case for smaller generators that are connected to low- or medium-voltage distribution networks.

When a supplier’s generation is exclusively embedded then the renewable supply \(\left( S_t \right)\) is included in the metered volume so, in this case:

\[\begin{equation*} D_{t} = -V_{t} - S_{t} \end{equation*}\]

Mixed generation

When a supplier has a mix of remote and embedded generation we cannot trivially determine the demand since the metered volume includes embedded generation, which itself is unknown:

\[\begin{equation*} \mathrm{V_{t} = -D_t - S^{embedded}_{t}} \end{equation*}\]

However, if we assume that the embedded generation is a constant fraction of the total:

\[\mathrm{S^{embedded}_t = \alpha \cdot S_t}\]

… and, for suppliers that are 100% volumetrically matched, we assume that they are precisely matched:

\[\mathrm{\sum_t^year D_{t} = \sum_t^year S_{t}}\]

… then we the problem that is sufficiently determined for us to solve for \(D_t\) and \(\alpha\):

\[\begin{equation*} \mathrm{D_{t} = -V_t - \alpha \cdot S_{t}} \;\;\; \text{subject to } \mathrm{\sum_t^year D_{t} = \sum_t^year S_{t}} \end{equation*}\]

Validation

We welcome input into merits of this approach and ways in which it can be improved. Several enhancements will be made in future work¹ and our full results will be submitted to peer review.

Footnotes

1. The handling of mixed generation can be improved by:

   • Using dedicated metering data for generators larger than 50 MW
   • Estimating which generation is most likely to be embedded based on the size, class, and location of the asset
   • Validating the embedded generation from a decomposition of the metered volume (e.g. looking for diurnal pattens that relate to solar output, variations that relate to wind speed, and drops in generation that relate to negative pricing

   ↩︎