Probabilistic forecasting of power system imbalance using neural network-based ensembles

TL;DR

An ensemble of C-VSNs, which are the adaptation of variable selection networks (VSNs), which predicts the imbalance of the current and upcoming two quarter-hours, along with uncertainty estimations on these forecasts, and developed a fine-tuning methodology to effectively include new inputs with limited history in the model.

Abstract

Keeping the balance between electricity generation and consumption is becoming increasingly challenging and costly, mainly due to the rising share of renewables, electric vehicles and heat pumps and electrification of industrial processes. Accurate imbalance forecasts, along with reliable uncertainty estimations, enable transmission system operators (TSOs) to dispatch appropriate reserve volumes, reducing balancing costs. Further, market parties can use these probabilistic forecasts to design strategies that exploit asset flexibility to help balance the grid, generating revenue with known risks. Despite its importance, literature regarding system imbalance (SI) forecasting is limited. Further, existing methods do not focus on situations with high imbalance magnitude, which are crucial to forecast accurately for both TSOs and market parties. Hence, we propose an ensemble of C-VSNs, which are our adaptation of variable selection networks (VSNs). Each minute, our model predicts the imbalance of the current and upcoming two quarter-hours, along with uncertainty estimations on these forecasts. We evaluate our approach by forecasting the imbalance of Belgium, where high imbalance magnitude is defined as $|$SI$| > 500\,$MW (occurs 1.3% of the time in Belgium). For high imbalance magnitude situations, our model outperforms the state-of-the-art by 23.4% (in terms of continuous ranked probability score (CRPS), which evaluates probabilistic forecasts), while also attaining a 6.5% improvement in overall CRPS. Similar improvements are achieved in terms of root-mean-squared error. Additionally, we developed a fine-tuning methodology to effectively include new inputs with limited history in our model. This work was performed in collaboration with Elia (the Belgian TSO) to further improve their imbalance forecasts, demonstrating the relevance of our work.

Figures (13)

• Figure 1: Example of a forecast made by our model, predicting the imbalance of the current and upcoming two quarter-hours (Qh). The solid blue line represents the forecast for the 0.5 quantile, while the other quantile predictions determine the model's confidence by providing prediction intervals.
• Figure 2: The quarter-hourly average and standard deviation of the Belgian imbalance over the entire dataset. Note that the imbalance average and volatility strongly depends on the time of day.
• Figure 3: Visualization of the C-VSN model. Next to each arrow, the dimensions of the vector or matrix that is passed is indicated. The $\otimes$ symbol denotes matrix multiplication, $\odot$ represents element-wise multiplication and $\oplus$ is element-wise addition.
• Figure 4: Value of the loss weights of the training labels for $c=0.1$ in \ref{['loss weights']}.
• Figure 5: The results achieved by each model on the test set, in terms of the overall RMSE, the overall CRPS, and the RMSE and CRPS considering only instances where the true imbalance magnitude is larger than $500\,$MW. The percentages denote the increase in RMSE or CRPS compared to our C-VSN ensemble.