新能源-梯级水电混合电站运行涉及风光功率预测、电价预测与水电调度决策。传统方法通常以预测精度最优为目标开展功率与价格预测，并据此进行调度优化，但预测误差对调度收益的影响具有非线性与非对称特性，单纯追求统计精度最优难以实现收益最大化。针对上述问题，提出一种面向收益提升的可变权重价值导向组合预测优化方法。首先，构建风电、光伏及电价的上层可变权重组合预测模型；其次，建立考虑抽蓄互斥、水量平衡及机组运行约束的下层梯级水电运行优化模型；进而形成预测与决策联动的双层优化框架，并采用贝叶斯优化确定价值导向权重。算例结果表明，该方法在多源预测与梯级调度耦合场景下具有更优性能，相比精度导向方法，能够强化高收益时段的预测-调度一致性，降低偏差惩罚与调度失配，提高综合运行效益。

The operation of a renewable-cascade hydropower hybrid plant involves wind and solar power forecasting, electricity price prediction, and hydropower scheduling. Conventional approaches optimize forecasting accuracy and then perform scheduling based on predicted values; however, forecasting errors affect scheduling profits in a nonlinear and asymmetric manner, so accuracy-oriented forecasts do not necessarily maximize operational benefits. To address this issue, a value-oriented forecasting-decision integrated optimization method with variable-weight ensemble prediction is proposed. An upper-level model is developed for wind power, photovoltaic power, and electricity price forecasting, with Bayesian optimization used to determine value-oriented weights. A lower-level cascade hydropower scheduling model is constructed considering pumping-generation exclusivity, water balance, and unit constraints, forming a bi-level optimization framework. Case studies show that the proposed method outperforms accuracy-oriented approaches in coupled multi-source forecasting and cascade scheduling scenarios, enhancing prediction-decision consistency in high-value periods, reducing deviation penalties and scheduling mismatches, and improving overall operational benefits.