For the high randomness and fluctuation of wind power, as well as the low precision of the power prediction, the traditional prediction of wind power point is not able to describe the uncertainty of wind power. A norm...
详细信息
ISBN:
(纸本)9781467371063
For the high randomness and fluctuation of wind power, as well as the low precision of the power prediction, the traditional prediction of wind power point is not able to describe the uncertainty of wind power. A normal distribution is usually used to model wind power forecast error, but it is not valid in some special cases. In this paper, non-parametric kernel density estimation is adopted to calculate the probability density errors of wind power prediction at different levels. According to system reserve capacity requirements, safety and economy of power generation dispatching, a wind power prediction interval with three spline interpolation is acquired which satisfies the certain confidence interval. The three spline interpolation is the wind power error's distribution function. An equality constrained optimization problem was simplified into an unconstrained optimization problem and Newton with the characteristics of non-parametric kernel was presented. Given a probability value at a certain precision, it's useful to use Newton to search for arguments. Then, the upper and lower range are obtained. The calculation results show that the used wind power interval forecasting method can provide wind power prediction curve and its variation range, and is more suitable for wind power uncertainly.
Subnetworks can reveal the complex patterns of the whole-genome network by extracting the interactions that depend on temporal, spatial, or condition specific context. In this paper we present an optimization framewor...
详细信息
Subnetworks can reveal the complex patterns of the whole-genome network by extracting the interactions that depend on temporal, spatial, or condition specific context. In this paper we present an optimization framework to identify condition specific subnetworks. This framework allows us to identify the most coherent subnetwork by integrating the information from both nodes and edges in the graph. Importantly we design an algorithm to solve the optimization problem efficiently. It is very fast and can extract subnetworks from large-scale network with about 10000 nodes. As a pilot study we apply our method to identify type 2 diabetes related subnetworks in the human protein-protein interaction network.
暂无评论