风电功率预测对当前风电并网及智能电网的建设、规划及生产调度都至关重 要。精准且有效的风电功率预测是发展能源互联网的必然要求。现今生活中，风电 场的开机规划及供给量判断不仅要考虑季节节点及运载负荷，更要应用精确量化 的风电功率预测以合理分配清洁电力供给，保证能源网络的平稳运行。风电功率的 有效预测，关键在于处理风能的复杂影响因素与波动性、间歇性等特征。在能源互 联网环境下，能够消除不确定性的风能概率预测方法更具实际价值。 为有效降低风电功率预测中的复杂度和不确定性，持续改善风能发电功率的 预测精度，优化预测成本。本文根据风能数据序列特征，提出了基于 EMD 与自助 分位数回归的概率密度预测方法和一种结合误差修正模型的自助分位数回归概率 密度预测方法。所构建模型首先运用经验模态分解（EMD）方法进行数据预处理 以有效提取数据关键信息，其次将自助法（Bootstrap）、分位数回归（QR）方法与 误差修正（EC）相结合，提出了考虑误差修正的自助分位数回归模型，最终以核 密度估计方法来优化模型的预测性能，并完成风能源概率密度预测实验研究。结果 表明，模型既给出了高精度的未来时刻预测值及其区间范围，而且获得了未来风电 完整的概率密度曲线。为展示所提模型方法的优越性和稳健性，本文从数据统计方 向来验证，通过概率密度预测方法所提供的均值、中位数和众数结果来评估模型的 点预测效果，与此同时根据合理的区间预测评价准则对预测区间进行评价分析。 本文主要利用基于 EMD 和自助分位数回归的概率密度预测方法进行风电功 率单步预测，运用基于 EC 与自助分位数回归的概率密度预测方法进行风电功率多 步预测。在预测过程中，同时考虑原始序列特征、滞后矩阵维度和误差分布修正以 构建稳健有效的预测模型。文章使用四个历史数据集进行算例分析，包含：加拿大 安大略地区 2019 年冬夏两季节的风电功率预测，西班牙西北部加利西亚风电场 2018 年不同频率冬季数据的风电预测。通过实验分析并与其他先进方法的对比， 进一步证明本文提出的风能概率密度预测方法可以在降低不确定性和优化复杂度 上取得均衡的同时提高预测精度。在科学研究上，本文方法提供了风电功率预测不 确定性与复杂度均衡建模策略，并获取了完整的风能概率密度曲线和预测区间，较 好地解决了风电能源的波动性和不确定性问题，为争取风电预测模型的实际应用 和保障电力网络的稳定运行提供了有效的技术改进和突破方向。 关键词：经验模态分解；误差修正；自助分位数回归；概率密度预测；不确定性分 析；风电功率III ABSTRACT Wind power forecasting is critical to the construction, planning, and production scheduling of current wind power integration and smart grids. Accurate and effective wind power forecasting is an inevitable requirement for the development of the energy internet. In reality, the wind farm's start-up planning and supply judgment must not only consider seasonal nodes and carrying loads, but also apply accurate and quantified wind power forecasting to reasonably allocate clean power supply to ensure the smooth operation of the energy network. The key to effective prediction of wind power is to deal with the complex influencing factors of wind energy and the characteristics of volatility and intermittency. In the energy Internet environment, the wind power probability prediction method that can eliminate uncertainty is more practical. In order to effectively reduce the complexity and uncertainty in wind power forecasting, continuously improve the forecasting accuracy of wind power generation power and optimize the forecasting cost. Based on the characteristics of wind energy data series, this paper proposes a probability density prediction method based on EMD and bootstrap quantile regression and a bootstrap quantile regression probability density prediction method combined with error correction model. The constructed model first uses the empirical mode decomposition (EMD) method for data preprocessing to effectively extract key data information, and secondly combines the bootstrap method, quantile regression (QR) method and error correction (EC) to propose A bootstrap quantile regression model with error correction is considered, and finally the kernel density estimation method is used to optimize the prediction performance of the model, and the experimental study of wind energy probability density prediction is completed. The results show that the model not only gives high-precision forecast values and interval ranges inthefuture, but also obtains the completeprobabilitydensity curveof future wind power. In order to show the superiority and robustness of the proposed model method, this paper verifies from the statistical direction of the data, and evaluates the point prediction effect of the model through the mean, median and mode results provided by the probability density prediction method, while Reasonable interval prediction evaluation criteria evaluate and analyze the prediction interval. This paper mainly usesthe probability density predictionmethod based on EMD and bootstrap quantile regression for single-step wind power forecasting, and uses the probability density prediction method based on EC and bootstrap quantile regression for multi-step wind power forecasting. In the prediction process, the original sequencecharacteristics, the lag matrix dimensions and the error distribution correction are considered simultaneously to build a robust and effective prediction model. The article uses four historical data sets for example analysis, including: the forecast of wind power in the winter and summer of 2019 in Ontario, Canada, and the forecast of wind power in 2018 with different frequency winter data from the Galicia wind farm in northwestern Spain. Through experimental analysis and comparison with other advanced methods, it is further proved that the wind energy probability density prediction method proposed in this paper can reduce the uncertainty and optimize the complexity while improving the prediction accuracy. In scientific research, the method in this paper provides a balanced modeling strategy for uncertainty and complexity of wind power prediction, and obtains a complete wind energy probability density curve and prediction interval, which better solves the volatility and uncertainty of wind energy The problem provides effective technical improvement and theoretical direction for the practical application of the wind power forecasting model and ensuring the stable operation of the power network. Keywords: Empirical mode decomposition; Error correction; Bootstrapping quantile regression; Probabilistic density Forecasting; Uncertainty analysis; Wind power forecastingV 目 录 第一章 绪论 ..........................................................1 1.1 研究背景与意义 ................................................ 1 1.2 国内外研究现状 ................................................ 5 1.2.1 风电功率预测研究现状....................................... 6 1.2.2 概率性预测................................................. 8 1.3 研究思路和研究方法 ............................................ 9 1.3.1 研究方法................................................... 9 1.3.2 本文的主要创新点.......................................... 10 1.3.3 研究内容.................................................. 10 第二章 相关理论概述.................................................13 2.1 经验模态分解 ................................................. 13 2.2 分位数回归 ................................................... 14 2.3 误差修正 ..................................................... 15 2.4 自助分位数回归模型 ........................................... 15 2.4.1 自助法.................................................... 15 2.4.2 自助分位数回归............................................ 15 2.5 贝叶斯网络模型 ............................................... 16 第三章 基于 EMD 与自助分位数回归的风电功率概率密度预测 ...............18 3.1 基于 EMD 与自助分位数回归的风电功率预测模型 ................... 18 3.2 基于 EMD 与自助分位数回归的风电功率概率密度预测模型 ........... 18 3.3 预测结果评价原则 ............................................. 21 3.3.1 点预测结果分析............................................ 21 3.3.2 区间预测评价准则.......................................... 22 3.4 加拿大安大略地区冬季和夏季风电功率案例分析 ................... 22 第四章 考虑误差修正的自助分位数回归单步风电功率概率密度预测 .........30 4.1 考虑误差