基于改进深度受限玻尔兹曼机算法的光伏发电短期功率概率预测

光伏发电功率受自然环境影响具有明显的波动性、间歇性与随机性,对光伏发电进行短期功率的概率预测可以有效缓解给电网调度、能量管理等方面带来的诸多不利影响。提出一种基于改进深度受限玻尔兹曼机(RBM)算法的光伏发电短期功率概率预测模型,通过灰色关联系数法寻找待预测日的相似日,并利用遗传算法对RBM算法进行参数优化,避免模型参数寻优陷入局部最优,以提高预测模型的预测精度。仿真算例表明,所提模型可以更好地反映光伏发电功率的概率分布。

Probability forecast of short-term photovoltaic power generation based on improved depth restricted Boltzmann machine algorithm

Since the photovoltaic power is of obvious fluctuation, intermittence and random caused by the natural environment, the probability forecast of short-term photovoltaic power generation can effectively reduce the adverse effects on the aspects of power grid dispatching, energy management, etc. A probability forecast model based on the improved depth RBM(Restricted Boltzmann Machine) algorithm is proposed, which adopts the grey correlation method to find the day similar to the forecasted day, and uses the genetic algorithm to optimize the parameters of RBM algorithm, avoiding falling into local optimum and improving the forecast accuracy. A simulation example shows that the proposed model can better reflect the probability distribution of photovoltaic power generation.