Research on deep-learning-based photovoltaic output estimation and multi-layer collaborative load prediction in distribution networks

Abstract

In recent years, the penetration rate of distributed photovoltaics in distribution networks has been continuously increasing. The coupling of photovoltaic output with actual load with randomness forms a generalized load with greater uncertainty, posing a serious challenge to the safe and stable operation of the distribution network. Accurate distributed photovoltaic output and load forecasting is the foundation for ensuring the safety and economy of distribution network operation. However, currently most distributed photovoltaics are installed after the electricity meter and lack proprietary metering devices, making photovoltaic output invisible to distribution network operators, greatly affecting photovoltaic output prediction. In addition, the load of the distribution network has a multi-layered structure from bottom to top, and existing load forecasting methods are difficult to meet the aggregation and consistency requirements of multi-layered loads, which increases the operational decision-making burden of the distribution network. In this context, this study applies deep learning techniques such as generative adversarial networks and long short-term memory neural networks to achieve accurate estimation of photovoltaic output and collaborative prediction of multi-layer loads in distribution networks containing distributed photovoltaics.