Predicting Solutions to the Optimal Power Flow Problem
Predicting Solutions to the Optimal Power Flow Problem
复制标题
预测最优潮流问题的解决方案
DOI：
--
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Aditya Garg
中科院分区：
文献类型：
--
作者：
Thomas Navidi;Suvrat Bhooshan;Aditya Garg
本文讨论了梯度推进回归预测输电网络最优潮流问题的输出，以减少计算时间的实现。最优潮流问题是以输电网运行费用最小为目标的一个重要优化问题。输入是每个节点所需的负载和发电成本。输出是网络中每个发电机的功率和电压。我们使用IEEE 30总线网络作为我们的静态网络模型，并从PJM加载数据。使用MATPOWER求解地面实况解。梯度提升回归基于每个输出与真正的最优解之间的均方差提供了最高的准确度。超过90%的预测结果在真实解的5%以内。然而，大约60%的预测违反了网络约束之一。这是避免使用预测的解决方案作为一个起点的最优潮流问题。一个案例研究表明，由于可再生能源的渗透，高度可变的负载，运行最佳潮流的计算时间，这种方式减少了高达30%。
This paper discusses an implementation of gradient boosting regression to predict the output of the optimal power flow problem for transmission networks to reduce the computation time. The optimal power flow problem is an important optimization to minimize the cost of operating a transmission network. The inputs are the load demanded at each node and the cost of generation. The outputs are the power and voltage of each generator in the network. We used the IEEE 30 bus network as our static network model, and load data from PJM. The ground truth solutions were solved using MATPOWER. Gradient boosting regression provided the highest accuracy based on the mean squared difference between each output to the true optimal solution. Over 90% of predictions were within 5% of the true solution. However, approximately 60% of predictions violated one of the network constraints. This is avoided by using the predicted solution as a starting point to the optimal power flow problem. A case study shows that with highly variable load due to renewable penetration, the computation time of running the optimal power flow this way is reduced by up to 30%.