Climate forecasting for energy: online workshop: 4 Dec 2020

Scientific meetings
, , ,
2

Presentations

Only those PDFs with open access licensing can be made available for download from this site.

Morning presentations

The presentation sessions at 10:05 and 12:00 were administered by the S2S4E project. Unlike the afternoon session, the presentations are listed without abstracts.

Afternoon presentations

Lightning talk presentations for the 15:20 time slot.

COP26 climate data hackathon brainstorm

  1. COP26 climate data hackathon. The final ten minute slot is now devoted to discussions on a climate data hackathon scheduled for 22–26 March 2021. There will be an initial “challenge brainstorm” event on 25 February 2021 for the community to generate ideas to take forward and develop during the main hackathon event.

    The hackathon is motivated by the next COP26 climate negotiations in Glasgow and is supported by the UK Met Office, University of Oxford and Reading University. This brainstorming session seeks input from participants regarding the format and goals of the data hackathon, who might potentially contribute, and how to make contact.

List of afternoon presentations

  1. Paul Westermann @pwest. Building energy surrogate models that span multiple climate zones. Machine learning surrogate models are being trained on building energy simulation in- and output data. Their key advantage is their computational efficiency, which allows modellers to explore building design performance in fractions of a second. However, these surrogate models are currently bound to the specific building energy simulation model, that was used for generating the training data set. In this study, we show how we can break that boundary by using a deep convolutional neural network which can process large annual hourly weather data. This allows the surrogate model to expand over all climates and modellers can assess the impact of climate on the building energy performance rapidly.

    To showcase the use of surrogate models they span multiple climates, we host our surrogate models on the platform www.buildingenergy.ninja. The surrogate models take building design parameters and annual hourly weather data as inputs, and produce annual hourly building loads as outputs.

  2. Adriaan Hilbers @ahilbers. Efficient quantification of the impact of climate uncertainty in energy system models. Recent studies indicate that the effects of climate uncertainty in energy system models should not be ignored. For example, running the same model with different years of demand and weather data (e.g. 2018 vs. 2019) may lead to significant spreads in outputs, and picking the “wrong year” of climate data may lead users to suboptimal energy strategy. For this reason, quantifying the impact of climate uncertainty in energy system models (creating confidence or prediction intervals) allows more robust decision-making. The standard approach involves running a model multiple times using different samples of demand and weather data. However, this is infeasible in many energy settings due to limitations in data (many different samples unavailable) or computing (many expensive model runs infeasible). In this presentation, we introduce a method that runs models across shorter samples and rescales uncertainty bounds in a statistically robust way, reducing both the data and computational burden. The paper, models, data and sample code can be found here.

  3. Ekaterina Fedotova @ekatef. Climate change impacts on the energy system under the fossil fuel curse. I’ll highlight the research works assessing the climate change impacts on the Russian energy system. A series of studies has addressed the following questions:

    • integral impacts of the warming climate on the national energy system
    • evolution of the renewable energy potential under the climate change
    • possible climate change effects on the renewables integration into power systems

    It has been shown that, while rapidly warming winters result in a significant heating demand decrease, the climate change associated shift in the load patterns is likely to create additional obstacles in escaping the fossil fuel trap. A combination of the energy systems simulation with reliable climate data seems to be crucial in resolving this issue.

  4. Adriaan Hilbers @ahilbers. Open energy system modeling for climate scientists and others. A tutorial presentation on the types of models developed and used within the open energy modeling community. It includes a simple tutorial on how climate and weather data is used in energy system models, as well as introducing a simple energy system model, designed as an introduction to the topic for climate scientists. The model, data and tutorial are available on GitHub under an MIT license:

  5. Bruno Schyska @Bruno. The sensitivity of power system expansion models on
    meteorological parameters    . Power system expansion models are a widely used tool
    for planning power systems, especially considering the integration of renewable
    resources. Studies using these models form the basis for far-reaching political
    decisions. The backbone of power system models is an optimization problem, which
    depends on a number of economic and technical parameters. Although these parameters
    contain significant uncertainties, a consistent way to quantify the sensitivity to
    these uncertainties does not yet exist. Here, we analyze and quantify the
    sensitivity of a power system expansion model to the meteorological parameter time
    series based on a novel misallocation metric. We find that the sensitivity to the
    weather data is in the same order of magnitude as the sensitivity to the definition
    of cost. By comparing different climatic periods we can, additionally, identify
    representative weather years and periods which should rather not be used for
    expansion planning. A preprint of the corresponding paper can be found
    here.

    • Schyska, Bruno U, Alex Kies, Markus Schlott, Lueder von Bremen, and Wided Medjroubi (4 December 2020). The sensitivity of power system expansion models on meteorological parameters — Presentation. Oldenburg, Germany: Institute of Networked Energy Systems, German Aerospace Center (DLR). Presented at the Climate Forecasting for Energy online workshop. CC‑BY‑4.0 license.
    • 2020-schyska-etal-sensitivity-power-system-expansion-models-meteorological-parameters.pdf (1.2 MB)