风光互补并网发电系统中的资源评估与短期功率预测研究

Effective resource assessment and short-term power forecast of wind/solar hybrid power generation system become the key to improve capability of stable operation of grid-connected systems, however, random fluctuating wind and solar energy can be affected by many factors, which has brought a serious challenge for the sustained stability of assessment and forecast. Given that the existing research cannot effectively solve this problem, theoretically in views of data mining algorithm, this project will carry out systematic research in four stages successively: (1) Establish meteorological and geographic feature information databases, then accordingly construct data analysis and preprocessing methods; (2) Construct dynamic adaptive comprehensive assessment system to adapt to weather types identification, and design manual intervention mechanism to cope with predictable events; (3) On the basis of fuzzy decision theory, establish fast forecast models which can select appropriate factors and determine the weight coefficient, finally realize the complete adaptation for different neural network; (4) Make full use of existing artificial intelligent forecast models, enhance and generalize forward factors extraction methods, establish fast and refined forecast models which can select appropriate factors and their associated two-way interaction terms; The completion of this project has the crucial theoretical and practical significance, which can not only enrich factor extraction forecast theory, but also provide reliable references on sites selection for wind/solar hybrid farm and optimization scheduling for power system.

风光互补发电系统中有效的资源评估与短期功率预测成为提高并网电力系统稳定运行能力的关键，但随机波动的风能与太阳能受诸多因素影响，这给评估与预测性能的持续稳定带来了严峻的挑战。鉴于现有研究不能切实解决这一问题，本项目将以数据挖掘算法为理论基础，从四个阶段依次开展系统的研究：（1）建立气象地理特征信息库，构建数据分析与预处理方法；（2）构建适应天气类型识别的风能与太阳能动态自适应综合评估系统，并为其设计人工干预机制以应对可预期事件；（3）基于模糊决策理论，建立能够提取合适影响因子并确定权重系数的快速预测模型，最终实现对不同神经网络的完整适应；（4）充分利用现有人工智能预测模型的优势，完善并推广向前因子提取方法，建立能够提取合适一次项与二次项因子的快速精炼预测模型。本项目的完成不仅可以丰富因子提取预测理论，同时为风光互补发电场选址和电力系统优化调度提供可靠的参考依据，具有重要的理论与实际意义。

风能太阳能的随机性与间歇性给风光互补并网电力系统和消纳带来了严峻挑战，准确的风电预测和光电预测是解决问题的有效途径。本项目致力于构建能够挖掘重要特征因子的可靠预测模型，不仅能提高风光互补发电预测的准确性，还能降低预测模型的复杂度，为确保电力系统安全稳定运行和提高风光互补发电穿透功率极限提供参考依据。项目组从数据预处理、特征因子自适应选择、构造新型复合惩罚函数、未知参数的动态寻优和模型优选等角度提出了适应多种高维特征数据环境下的多个预测模型，同时通过智能优化算法、因子提取方法和集成学习方法预测风速和太阳辐射强度，评估了影响它们的重要因素，实验结果显示提出的模型具有鲁棒性、稀疏性和较高的预测精度。本项目是预测理论与方法、因子提取技术、能源经济与管理的交叉研究，对完善预测理论，推广能源预测领域的因子提取方法，以及评估风能太阳能发展具有一定的理论意义和实际价值。通过项目资助项目组共发表学术论文16篇，其中SCI收录的15篇，中文核心论文1篇，培养硕士研究生7名，协助培养留学博士研究生1名。