Subtopic 6: Laboratory and Field demonstration of MGs with PV Modules and Smart Storage

粒化計算為本的概率預測技術
Granular Probabilistic Interval Forecasting

許昭教授，香港理工大學，電機工程學系
Prof. Zhao Xu, Department of Electrical Engineering, The Hong Kong Polytechnic University

為達到更佳的能源效益，用電量的預測不可或缺。其中，粒化計算為本的概率預測技術被公認為一種相當有展望的方法。團隊與香港天文台合作，開發了性能優秀的粒化概率預測技術，能應用於太陽能及其他可再生能源的微電網。圖二示例為粒化極限學習機以15分鐘為間隔的光照輻射預測結果，其預測可信度為90%。

Electricity consumption forecasting is essential to achieve energy efficiency. Here, granular probabilistic forecasting technology has been shown to be a promising approach. In collaboration with the Hong Kong Observatory, the team developed high-performing granular probabilistic forecasting technology, which is applicable in solar and other renewable energies. Fig. 2 shows an example of 15-minute interval prediction results of solar irradiance by using granule-based Extreme Learning Machine (G-ELM) model, of which the confidence level retains 90%. 

圖一. 團隊以 Matlab 軟件，開發出再生能源預測系統 FSRE。現時，它已被應用於團隊的微電網系統中，估計太陽能光伏的概率預測上。

Fig. 1. Using the Matlab software, the team developed the Forecasting System for Renewable Energy (FSRE). Currently, it is applied to estimate the probabilistic intervals for the solar PV power in their microgrid.

圖二. 粒化極限學習機以15分鐘為間隔的光照輻射預測結果。其預測可信度為90%。

Fig 2. The 15-minute prediction intervals of solar irradiance via the granule-based model has a nominal confidence level as high as 90%.

相關文獻 Related Papers:

• Songjian Chai*, Zhao Xu* & Wai Kin Wong*. (2016). Optimal granule-based PIs construction for solar irradiance forecast. IEEE Transactions on Power Systems, 31(4): 3332-3. [Link]

Project Leaflet 研究計劃簡介	Available Patent 可授權專利