基于活动区光球磁特征参数图像的太阳耀斑预报建模研究

太阳耀斑及其伴随或引发的太阳质子事件、日冕物质抛射事件，可能引发剧烈的空间环境扰动，严重威胁到航天器和卫星的安全。当前耀斑预报模型仍面临预报能力有限、预报时间提前量短的问题，主要受到三个方面的影响：活动区图像信息利用的不充分、缺乏新颖有效的先兆因子、预报建模方法有待进一步优化。本项目将基于与耀斑具有强相关性的磁场梯度、垂直电流、螺度、自由磁能、剪切角、中性线梯度等活动区光球磁特征参数的图像，利用深度学习方法提取可以表征活动区磁场特性、反映太阳耀斑爆发物理机制的时序特征，进一步研究活动区的演化特征和太阳耀斑爆发的物理机制；针对具有较强空间环境效应的强太阳耀斑，分析和筛选与耀斑爆发具有强相关性的先兆因子；利用机器学习方法开展太阳耀斑预报建模研究，有针对性地对子模型进行训练、优化和融合，建立满足空间天气业务预报需求的太阳耀斑预报模型，进一步提升耀斑预报的能力和预报时间提前量。

Solar flare is one of the most important sources to cause hazardous space weather. During flare eruptions, magnetic free energy is released vigorously from the solar surface and gets converted into plasma thermal energy, kinetic energy as well as particle radiation though magnetic reconnections and other physical mechanisms. Large amount of plasma, sometimes accompanied by heavy radiation dosage, arrives at the Earth within a few hours after the flare onset, threatening our navigation, communications, control systems, and the instruments of the spacecraft. Studying the active regions (ARs) of the Sun could potentially help to predict solar flare onset and contribute to space weather prediction. In this project, we propose to implement a deep-learning algorithm to extract the features that can describe the photospheric magnetic properties of the ARs. This method paves its way out of the fundamental and deduced physical parameters of the ARs, including magnetic gradient, vertical current, helicity, free energy, shear angle, and polarity inversion lines. The evolution of ARs and the physical mechanisms of solar flares will be explored extensively based on a time series analysis of the above extracted features. Considering that strong flares have a much more significant impact on the space weather, the precursor factors that are strongly related to flare eruptions will be specifically chosen to develop sub-models for forecasting purpose based on the machine-learning algorithm. The forecasting model to be developed by targeted training, optimization, and staking methods of machine-learning could thus support a variety of space weather products and be used to predict solar flares specifically.