Abstract:This paper develops a new architecture that combines spatial and time‐series information for both LV and MV substation‐side short‐term load forecasting. The loop training structure naturally addresses the data scarcity problem at the distribution level. The new approach significantly improved the forecasting accuracy under both ordinary and unexpected extreme cases, for example, extreme weather and pandemic outbreaks. Compared to load forecasting at the national level, two challenges arise in providing accurate forecasting for LV and MV networks: (1) customers within LV and MV networks are much less, implying greater volatility within those load profiles; (2) not all customers have smart metres. Particularly, the two challenges would exacerbate forecasting performance under unexpected events, such as extreme weather and COVID‐19 outbreaks. To secure accurate short‐term load forecasting for LV and MV networks, this paper customised a Spatio‐Temporal Edge‐Cloud‐coordinated (STEC) approach on a loop training structure—LV networks to MV networks to LV networks. For each LV network, this approach utilises XGboost to learn the relationship between weather data, substation‐side loads, and a few accessible customer‐side load data to deliver rough forecasting. Then, it adopts the rough forecasting results and accessible data for all LV networks within an MV network to train the convolutional neural networks and gated recurrent unit (CNN‐GRU) network. This step provides load forecasting for MV networks and simultaneously refines load forecasting for LV networks by generating the interacting relationship between LV substations of different locations. Case studies reveal that the STEC approach successfully extrapolates the demand‐varying information from long‐term datasets and improves short‐term forecasting performance under both normal scenarios and newly occurring unexpected scenarios for LV and MV networks. The loop training structure halves the forecasting error, compared to classic methodology by utilising the local data only for MV networks.

A Collaborative Cloud-Edge Approach for Robust Edge Workload Forecasting

ELASTIC: Edge Workload Forecasting Based on Collaborative Cloud-Edge Deep Learning

Microservice Auto-Scaling Algorithm Based on Workload Prediction in Cloud-Edge Collaboration Environment

AI-oriented Workload Allocation for Cloud-Edge Computing.

Collaborative Cloud and Edge Computing for Latency Minimization

One for All: Unified Workload Prediction for Dynamic Multi-tenant Edge Cloud Platforms

Edge-cloud collaboration for low-latency, low-carbon, and cost-efficient operations

A Collaborative Edge Computing Framework for Predictive Maintenance in Nuclear Power Plants

EsDNN: Deep Neural Network based Multivariate Workload Prediction Approach in Cloud Environment

Robust Auto-Scaling with Probabilistic Workload Forecasting for Cloud Databases

Dynamic Load Combined Prediction Framework with Collaborative Cloud-Edge for Microgrid.

Adaptive Task Offloading for Mobile Edge Computing with Forecast Information

Design and Evaluation of a Prediction-Based Dynamic Edge Computing System

A lightweight method of integrated local load forecasting and control of edge computing in active distribution networks

Dynamic Parallel Multi-Server Selection and Allocation in Collaborative Edge Computing

Task allocation algorithm and optimization model on edge collaboration

Accurate workload prediction for edge data centers: Savitzky-Golay filter, CNN and BiLSTM with attention mechanism

Short‐term load forecasting facilitated by edge data centres: A coordinated edge‐cloud approach

Collaborative Resource Management and Workloads Scheduling in Cloud-Assisted Mobile Edge Computing across Timescales

Providing Worst-Case Latency Guarantees with Collaborative Edge Servers.

Robust Task Offloading in Dynamic Edge Computing.