LSTM-RNN-FNN Model for Load Forecasting Based on Deleuze’s Assemblage Perspective

Abstract

Read online

Reliable load forecasting is essential for electricity generation and even for people’s lives. However, the existing load forecasting theories cannot match the requirements of complex systems (e.g., smart grids). Deleuze’s metaphysical complexity theory is seen as the theoretical foundation for comprehending complex systems, and thus, a new perspective based on Deleuze’s assemblage concept is given. According to the assemblage perspective, the electrical load is a quantitative representation of the mutual becoming of people and electricity, and load forecasting is an attempt to control this continuous process of deterritorialization and reterritorialization. We built an LSTM-RNN-FNN model for load forecasting based on the assemblage perspective, and the assessment results demonstrated that the model has high prediction performance. Furthermore, the performance of adding the temperature parameter into the network is also tested, while the correlation between the temperature and load is not strong enough and may not be suitable for load prediction. The assemblage perspective has significant implications for future load forecasting and potentially smart grid research.

Keywords

• load forecasting
• assemblage
• neural networks
• LSTM-RNN-FNN
• Deleueze