Incremental Adaptive Time Series Prediction for Power Demand Forecasting

Vrablecová, Petra; Rozinajová, Viera; Ezzeddine, Anna Bou

doi:10.1007/978-3-319-61845-6_9

Cited by 3 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Early work on prediction in this kind of scenario includes modifying ARIMA models to an online manner [3,52], predicting with kernel-based methods [63], and eforts on elastic resource scaling to reduce cloud system operating cost [9,68] More recent work leverages deep learning on streaming data. For instance, Vrablecová et al [73] proposed a stream change detection method to identify the ongoing changes or concept drifts of the power meter data. Guo et al [37] proposed an adaptive gradient learning method which aims to minimize impacts from outliers as well as leverage the local features, but this work is solely based on RNN and only targets univariate time-series prediction.…”

Section: Related Workmentioning

confidence: 99%

Graph Deep Factors for Probabilistic Time-series Forecasting

Chen

Rossi

Mahadik

et al. 2023

ACM Trans. Knowl. Discov. Data

View full text Add to dashboard Cite

Effective time-series forecasting methods are of significant importance to solve a broad spectrum of research problems. Deep probabilistic forecasting techniques have recently been proposed for modeling large collections of time-series. However, these techniques explicitly assume either complete independence (local model) or complete dependence (global model) between time-series in the collection. This corresponds to the two extreme cases where every time-series is disconnected from every other time-series in the collection or likewise, that every time-series is related to every other time-series resulting in a completely connected graph. In this work, we propose a deep hybrid probabilistic graph-based forecasting framework called Graph Deep Factors (GraphDF) that goes beyond these two extremes by allowing nodes and their time-series to be connected to others in an arbitrary fashion. GraphDF is a hybrid forecasting framework that consists of a relational global and relational local model. In particular, a relational global model learns complex non-linear time-series patterns globally using the structure of the graph to improve both forecasting accuracy and computational efficiency. Similarly, instead of modeling every time-series independently, a relational local model not only considers its individual time-series but also the time-series of nodes that are connected in the graph. The experiments demonstrate the effectiveness of the proposed deep hybrid graph-based forecasting model compared to the state-of-the-art methods in terms of its forecasting accuracy, runtime, and scalability. Our case study reveals that GraphDF can successfully generate cloud usage forecasts and opportunistically schedule workloads to increase cloud cluster utilization by 47.5% on average. Furthermore, we target addressing the common nature of many time-series forecasting applications where time-series are provided in a streaming version, however, most methods fail to leverage the newly incoming time-series values and result in worse performance over time. In this paper, we propose an online incremental learning framework for probabilistic forecasting. The framework is theoretically proven to have lower time and space complexity. The framework can be universally applied to many other machine learning-based methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Graph Deep Factors for Probabilistic Time-series Forecasting

Chen

Rossi

Mahadik

et al. 2023

ACM Trans. Knowl. Discov. Data

View full text Add to dashboard Cite

show abstract

“…In this section, we introduce our method for coping with concept drifts. For our method, we assume like [20] that concept drifts in load forecasting mainly influence the level of a time series. Based on this assumption, we propose to model the level and the remainder of a time series separately and to add them later on (see Figure 1).…”

Section: Profiles For Concept Driftsmentioning

confidence: 99%

“…With regard to ensemble methods, Jagait et al [11] combine the OARNN with a moving ARIMA to address concept drifts. Lastly, Vrablecová et al [20] propose a detection-based method where they retrain the linear regression model when the threshold-based detection recognizes a concept drift. However, all proposed methods rely on retraining or model selection methods that tend to be computationally expensive.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Adaptively coping with concept drifts in energy time series forecasting using profiles

Heidrich

Ludwig

Turowski

et al. 2022

Proceedings of the Thirteenth ACM International Conference on Future Energy Systems

View full text Add to dashboard Cite

Accurate electrical load forecasts are necessary to stabilize the electricity grid, e.g., by optimally operating energy storage systems or using demand-side management. However, an implicit assumption of most load forecasting methods is that future data looks similar to past data. Unfortunately, this assumption often does not hold; for example, recruiting new staff or a pandemic can lead to demand changes resulting in so-called concept drifts in the underlying data. Most methods for coping with such concept drifts rely on computationally expensive retraining. We propose a new method for coping with concept drifts based on profiles and a linear regression model that avoids expensive retraining. Compared to a simple baseline and five state-of-the-art benchmark models on two different data sets, our method has lower computational costs and higher forecast accuracy, making it especially interesting for smart grid applications. CCS CONCEPTS• Applied computing → Forecasting; • Computing methodologies → Machine learning.

show abstract