基于分布鲁棒优化方法的发电自调度研究

发布时间：2018-05-16 04:15

本文选题：发电自调度 + 分布鲁棒优化　；参考：《浙江大学》2017年硕士论文

【摘要】：在解除管制的电力市场中,发电公司为了在竞争中使自身获得最大收益,需要依据某种优化理论得到的竞标曲线进行投标。在联营交易模式中,发电公司和消费者分别提出供给和需求,然后由系统调度员(ISO)根据电网传输容量和网络约束等条件建立合适的市场出清规。发电公司在联营交易模式的电力市场中参与竞争时,只能通过对LMPs(节点边际电价)的预测,向ISO提供相应的发电计划。而在充分竞争的电力市场中,随着系统投标的市场化运作,必然会产生更多的不确定因素,因此发电公司面临的挑战是如何根据电力市场预测的电价,建立有效的发电自调度计划。发电自调度模型是一类含有不确定性随机变量的规划问题,模型中的负荷、节点边际电价等参数具有一定的不确定性,求解难度会随着系统规模的增大、约束条件和不确定参数的增多而增大。对于含有不确定性随机变量的规划问题,其建模方法主要是随机规划方法和鲁棒优化方法。前者通过引入随机变量来描述不确定性,但这种方法需要得知随机变量的完整分布统计特性,并且会产生大量的离散样本点,导致计算规模过于庞大;后者不需要参数的概率分布,仅需要假设不确定参数属于一个不确定集,但这种方法可能过于保守,且没有利用一些可以取得的概率统计信息。本文考虑节点边际电价的大小既有一定的范围,又具有一定的概率分布。因此可以使用一种将随机规划与鲁棒优化相结合地方法,即分布鲁棒优化方法,对含节点边际电价不确定的发电自调度问题进行求解。虽然电价的概率分布是不确定的,但通过长期统计可以得到电价的期望及协方差等确定的矩参数。电价短期的概率分布与长期统计的概率分布相似,但其矩一般是不同的,即矩具有不确定性。因此,利用矩不确定分布鲁棒优化方法来解决此类问题可以很好地处理节点边际电价这一特性。该方法首先定义一个以节点边际电价地期望、协方差为中心地不确定集合,构建一个min-max问题,然后通过Lagrange对偶原理将原问题转换为一个半正定规划进行求解。通过IEEE30节点系统进行计算分析,可以验证本文所提方法的可行性有有效性。
[Abstract]:In the deregulated electricity market, in order to obtain the maximum profit in the competition, the power generation company needs to bid according to the bidding curve obtained by some optimization theory. In the joint trading mode, the power generation company and the consumer put forward the supply and demand, and then the system dispatcher ISOO (system dispatcher ISOO) established the appropriate market rules according to the transmission capacity and network constraints of the power network. When the generation company participates in the competition in the electricity market of the joint trading mode, it can only provide the corresponding generation plan to the ISO by forecasting the LMPs (nodal marginal electricity price). In the fully competitive electricity market, with the market-oriented operation of the system bidding, there will inevitably be more uncertain factors. Therefore, the challenge faced by the power generation companies is how to forecast the electricity price according to the electricity market. Establish effective self-dispatch plan. Power generation self-dispatch model is a kind of programming problem with uncertain random variables. The load and marginal electricity price of the model are uncertain, and the difficulty of solving the problem will increase with the scale of the system. The increase of constraint conditions and uncertain parameters increases. For programming problems with uncertain random variables, the modeling methods are mainly stochastic programming method and robust optimization method. The former describes uncertainty by introducing random variables, but this method needs to know the complete distribution statistical characteristics of random variables, and it will produce a large number of discrete sample points, resulting in a large scale of calculation. The latter does not need the probability distribution of the parameter, but only needs to assume that the uncertain parameter belongs to an uncertain set, but this method may be too conservative and does not make use of some available probability and statistics information. In this paper, the size of the marginal price of the node is considered in both a certain range and a certain probability distribution. Therefore, a local method combining stochastic programming with robust optimization, that is, distributed robust optimization method, can be used to solve the generation self-dispatch problem with uncertain marginal price of electricity nodes. Although the probability distribution of electricity price is uncertain, the moment parameters such as the expectation of electricity price and covariance can be obtained by long-term statistics. The short-term probability distribution of electricity price is similar to the probability distribution of long-term statistics, but the moment is generally different, that is, the moment is uncertain. Therefore, using the moment uncertain distribution robust optimization method to solve this kind of problem can well deal with the characteristic of node marginal price. In this method, we first define a set of uncertainties with marginal electricity price expectation and covariance as the center to construct a min-max problem, and then transform the original problem into a semi-positive definite programming by means of the Lagrange duality principle. The feasibility and validity of the proposed method can be verified by the calculation and analysis of the IEEE30 node system.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM73

【参考文献】