Abstract In the context of smart cities, the growing share of solar power induces uncertainty in power generation due to inherent climatic variations. Accurate forecasting will be a key point for future residential microgrids since inability to do so could dramatically impact power balance and grid stability. This paper proposes an enhanced energy management framework which aims to efficiently address uncertainty issues due to local climatic variations in a peninsula context. The proposed framework uses a ten-state Markov chain to generate stochastic solar irradiation as well as a forecast correction method based on recursive least-squares updated every hours in order to efficiently take part in hour-ahead power bidding process. Numerical results highlight benefits obtained by combining proposed forecast correction method and storage in a practical example of Saint-Nazaire, a city located in peninsula of Guerande, France. Besides, a sensitivity analysis regarding impact of storage size and aggregator penalty on operation cost and commitment indices is investigated. Obtained results demonstrates better accuracy in delivering power to the grid and will lead residential microgrids dealing with strong climatic variations to decrease their operation cost and increase power balance for all grid stakeholders.

中文翻译：

---

考虑灵活性服务机会和预测不确定性的住宅微电网能源管理

摘要 在智慧城市的背景下，由于固有的气候变化，太阳能发电份额的增加导致发电的不确定性。准确预测将是未来住宅微电网的关键点，因为无法这样做可能会显着影响电力平衡和电网稳定性。本文提出了一个增强的能源管理框架，旨在有效地解决由于半岛环境中当地气候变化引起的不确定性问题。所提出的框架使用十态马尔可夫链来生成随机太阳辐射以及基于每小时更新一次的递归最小二乘法的预测校正方法，以便有效地参与提前一小时的电力招标过程。数值结果突出了在位于法国盖朗德半岛的城市圣纳泽尔的实际示例中，通过将提议的预测校正方法和存储相结合而获得的收益。此外，还研究了有关存储大小和聚合器惩罚对运营成本和承诺指数的影响的敏感性分析。获得的结果表明，向电网输送电力的准确性更高，并将引导住宅微电网应对强烈的气候变化，以降低其运营成本并增加所有电网利益相关者的电力平衡。