Ultra-Short-Term Photovoltaic Power Prediction Based on Self-Attention Mechanism and Multi-Task Learning

Abstract

Due to the volatility and randomness of the photovoltaic power generation, it is difficult for traditional models to predict it accurately. To solve the problem, we established a model based on the selfattention mechanism and multi-task learning to predict the ultra-short-term photovoltaic power generation. First, we selected the data with the optimal timing length and input the data into the Encoder-Decoder network based on the self-attention. The validity of features extracted by the encoder was checked by the decoder. Then, we added a restriction to the middle layer of the Encoder-Decoder network to prevent the autoencoder from copying the input to the output mechanically. This condition is used to predict the photovoltaic power generation, so a multi-task learning model was established. Finally, to take full advantage of the features that are efficiently expressed and allow our main task, the prediction task, to learn some unique features autonomously, we proposed a step-by-step training method and have validated the effectiveness of this view in experiments. Through experimental contrast, it is found that compared with the Encoder-Decoder network based on CNN and LSTM, the performance of the proposed method has been increased by 14.82% and 8.09% respectively. The RMSE and MAE of the Encoder-Decoder model based on the self-attention mechanism using step-by-step training are 0.071 and 0.040 respectively. INDEX TERMS Energy, PV power prediction, self-attention mechanism, multi-task learning, autoencoder.

Short-Term Cooperative Operational Scheme of Distribution System with High Hosting Capacity of Renewable-Energy-Based Distributed Generations

“…Consequently, it is reasonable to assume that the load is almost constant during the control period. The accuracy in the prediction of PV output has been improved in recent studies; in particular, shortterm output prediction when the time unit is in minutes for the immediate future has a considerably high accuracy [31,32]. Even if the actual PV output is different from the predicted value, the short-term variability can be controlled by the proposed curtailment method to resolve any problems caused by line overcurrent and voltage violations.…”

Short-Term Forecasting Photovoltaic Solar Power for Home Energy Management Systems

“…To conclude the discussion, the authors would like to highlight that this predictive model is only one of the models necessary for our HEMS. Models of electric consumption were developed by the authors using the same methodology [30,33,44] and will be integrated into a MBPC HEMS scheme as the one described in [54]. Furthermore, in future studies the effect of dust on the PV panels will also to be considered, following the guidelines of [55], where an experimental analysis was developed for different dust types to evaluate their impact on the power output of the modules.…”

Deep Learning in Smart Grid Technology: A Review of Recent Advancements and Future Prospects

“…Compared to HA, SA is easier to implement [57]. Two variant attention mechanisms are introduced: Self-Attention mechanism (SAM) and Multi-Head Attention Mechanism (MHM) [58]. The self-attention is employed to capture spatial dimensions in the nonlocal feature dependence, which is hard to be extracted by convolution kernels.…”

“…For instance, paper [58] employed SAM and Multi-Task Learning for Photovoltaic Power Forecasting (PVPF). Despite the fundamental role of SAM in enhancing the RNN-based models' performance, a massive number of additional learn-able hyperparameters have been added into the prediction system, which requires extensive tuning, making it unsatisfactory for real-world scenarios [58]. In [59], the authors used a multi-head CNN-RNN architecture for anomaly detection.…”