【作者】 陈建华；

【导师】 甘德强；

【作者基本信息】 浙江大学 ， 电力系统及其自动化， 2008， 博士

【摘要】 鉴于日前市场价格的不确定性,为降低市场风险,国内多个市场引入差价合同进行风险控制,合同交易量往往占当年市场需求的80%以上。发电厂商与购电公司签订的差价合同一般为电量合同,如何将合同交易总量以一定的方式进行分解,兼顾各方利益并促进市场逐步健康发展,是一个具有重要意义且值得研究的问题。在此方面,本文做了如下主要工作:电力市场的合同分解中应用确定性合同分解算法需要制定典型负荷曲线,历史负荷中的异常数据必然影响典型负荷曲线的有效性。提出了一种新的负荷形状畸变识别方法,并将其与传统的t检验法相结合,应用于负荷异常数据的识别和修正。最后选用浙江电网的历史数据,应用上述方法进行了异常负荷识别和修正,分析结果说明该方法简单、有效。确定性合同分解算法是一种直观、简单的分解方法,但相关实证性研究较少。提出平均合同电量比例、平均合同电量比例和电价相关性两项公平性评估指标,采用期望、标准差和半标准差评估年度基数合同对电厂的收益、风险的影响。选用华东市场实际数据,分析结果表明在二次华东市场调电试运行中确定性合同分解算法公平地对待发购电双方,电厂签订年度基数合同合同可有效地降低其风险。从合同博弈角度出发,考察单位电量的净收益计算单一购买者和发电厂商的效用,分析了单一购买者和发电厂商愿意参与确定性合同分解形成联盟的理论条件。当仅有一个发电厂商时,若双方均为风险厌恶型则联盟稳定,双方均为风险喜好型则联盟不稳定;当有一个购电公司、多个发电厂商时,需满足多个不等式条件时联盟稳定。选用国内某省级市场实际数据进行仿真分析,分析结果表明模型合理有效。传统差价合同分解是以负荷预测值为基础进行,负荷的不确定性可能导致分解结果无法满足预期目标,因此在合同分解时有必要考虑负荷的不确定性。选用广义自回归条件异方差类模型建立随机负荷模型,以合同电量比例与预定比例之差的标准差最小化为目标,以实际标准差小于预期标准差的概率为机会约束,构造了最优差价合同分解的随机规划模型。采用基于蒙特卡洛模拟的遗传算法求解,算例结果表明所提出的模型和方法是有效的。传统的中长期差价合同分解都是统一分解,发电商被动接受。提出了“发电商申报法”分解差价合同电量,在考虑检修、出力约束、系统可分解合同电量和系统必需合同电量等因素后,采用基于误差的差异度评估函数建立多目标优化模型,并分别采用“线性加权求和法”和“极小极大值法”将原模型转换为单目标优化。采用国内市场的实际数据构造算例,计算结果表明模型有效,两种转换均可获得原问题的有效解,采用“极小极大值法”时各发电商调整量的差异较小。更多还原

【Abstract】 Considering the uncertainty of the day-ahead market price, in order to reduce the market risk, CfD （Contract for Difference） is introduced to hedge against market risk in some domestic markets. Usually, contract energy accounts for more than 80% of total energy. Generally, CfD that generators and buyers signed are energy contracts, how to decompose the total contract engergy while covering all participants’ benefit, is an issue which is important and worth studying. The contributions are summarized as follows:In order to apply the deterministic contract decomposition algorithm to decompose contract energy, typical load curve must be prepared. Abnormal historical load data affect the validity of typical load curve. The notion of D_ht is proposed, it is applied to identify the abnormal load data, and combined with t test to identify and correct the load outlier. The historical data of Zhejiang province is used to test the suggested approach, the result indicates that the algorithm is simple and effective.An analysis on the deterministic contract decomposition algorithm of East China electricity market is presented. Average contract quantity proportion, correlation of contract quantity proportion and price have been adopted to evaluate the equity of the algorithm. Expectation, standard deviation and half standard deviation have been adopted to evaluate the income and risks of generators. An empirical analysis is done based on the historical data of East China electricity market. The result indicates that the algorithm treats buyer and generators fairly in the trial operation, CfD contracts help to reduce the generator’s risk effectively.From the viewpoint of cooperative game, per unit net revenue is adopted to evaluate single buyer and generators’ utility, the conditions under which the generators and buyers are willing to participate in contract decomposition is derived. In a single-buyer-single-supplier market, if both parties are risk-averse, the coalition is always stable; if both parties are risk-adventurous, the coalition is always unstable. In a single-buyer-multi-supplier market, the coalition is conditionally stable. A numerical study based on the historical data of some China provincial electricity market is performed. The result indicates that the proposed model is reasonable and effective.Traditional decomposition of CfD decomposition is based on the forecasted load, the load uncertainty may lead to that decomposition result can’t satisfy the expected target, so considering load uncertain is necessary when decomposing contract. A stochastic load model based on the GARCH （Generalize Autoregressive Conditional Heteroscedasticity） model is suggested for contract decomposition. Taking the minimized standard deviation of the difference between contract ratio and appointed ratio as the objective, the probability that actual standard deviation is smaller than the expected standard deviation as a chance constraint, an optimal contract decomposition based on chance constrained programming is proposed. A genetic algorithm based on Monte Carlo simulation is used to solve the problem, the numerical result indicates that the model and method is effective.Traditional contract decomposition method is based on such idea: "contracts are decomposed in a centralized way; generation companies accept the result passively". A new contract decomposition method named "generation company declaration method" is presented. The maintenance, the power capacity of units, etc. are taken into account. A difference evaluation function based on errors is adopted to build a multiple objective optimization model, "linear weighted sum" and minimax method are respectively selected to convert the original problem into a single objective optimization. A case study of a domestic market is performed, the results indicate that the suggested model is effective, the solution of the original problem can be obtained by both conversion methods. When minimax method is selected, the difference in generation companies’ adjustment quantity is smaller.更多还原