Semi-asynchronous personalized federated learning for short-term photovoltaic power forecasting

Accurate forecasting for photovoltaic power generation is one of the key enablers for the integration of solar photovoltaic systems into power grids.Existing deep-learning-based methods can perform well if there are sufficient training data and enough computational resources.However,there are challenges in building models through centralized shared data due to data privacy concerns and industry competition.Federated learning is a new distributed machine learning approach which enables training models across edge devices while data reside locally.In this paper,we propose an efficient semi-asynchronous federated learning framework for short-term solar power forecasting and evaluate the framework performance using a CNN-LSTM model.We design a personalization technique and a semi-asynchronous aggregation strategy to improve the efficiency of the proposed federated forecasting approach.Thorough evaluations using a real-world dataset demonstrate that the federated models can achieve significantly higher forecasting performance than fully local models while protecting data privacy,and the proposed semi-asynchronous aggregation and the personalization technique can make the forecasting framework more robust in real-world scenarios.

Hybrid SDS and WPT-IBBO-DNM Based Model for Ultra-short Term Photovoltaic Prediction

Accurate photovoltaic(PV)power prediction has been a subject of ongoing study in order to address grid stability concerns caused by PV output unpredictability and intermittency.This paper proposes an ultra-short-term hybrid photovoltaic power forecasting method based on a dendritic neural model(DNM)in this paper.This model is trained using improved biogeography-based optimization(IBBO),a technique that incorporates a domestication operation to increase the performance of classical biogeography-based optimization(BBO).To be more precise,a similar day selection(SDS)technique is presented for selecting the training set,and wavelet packet transform(WPT)is used to divide the input data into many components.IBBO is then used to train DNM weights and thresholds for each component prediction.Finally,each component’s prediction results are stacked and reassembled.The suggested hybrid model is used to forecast PV power under various weather conditions using data from the Desert Knowledge Australia Solar Centre(DKASC)in Alice Springs.Simulation results indicate the proposed hybrid SDS and WPT-IBBO-DNM model has the lowest error of any of the benchmark models and hence has the potential to considerably enhance the accuracy of solar power forecasting(PVPF).