Provision of regulation service reserves by flexible distributed loads
Provision of regulation service reserves by flexible distributed loads
复制标题
通过灵活的分布式负荷提供调节服务储备
DOI：
10.1109/cdc.2012.6426025
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
Elli Ntakou
中科院分区：
文献类型：
--
作者：
M. Caramanis;I. Paschalidis;C. Cassandras;Enes Bilgin;Elli Ntakou
继我们以前的工作控制的多个设备，以响应独立系统运营商（ISO）的监管服务信号（RSS），我们模拟ISO的RSS动态-在几秒钟的时间尺度演变-作为一个两级马尔可夫过程的转移概率校准的实际数据。家电响应被建模为马尔可夫调制过程符合指数分布的时间关闭和预期的效用，是凹的价格收取时appliance turnon. Prices是动态广播的智能建筑运营商（SBO）的目标是最大限度地提高时间平均效用时获得的设备打开减去不完善的RSS跟踪的成本。我们证明了随机动态规划（DP）的政策，使我们能够制定一个近似的离散状态和控制空间DP的问题，并提出了一个合理的近似，使问题可扩展到多个设备类别的某些属性。离散化状态DP解可以作为线性规划（LP）的解来获得。LP提供了最优的动态价格控制策略，并且还产生了SBO所需的必要信息，以在提前一小时的平衡市场中最优地竞标能量和调节服务储备（RSR）容量。本文的主要贡献是求解离散化的真实的时间市场最优价格策略，用它们来标定一个连续的解析策略函数，并从真实的时间最优策略中提取出对提前一小时远期平衡市场的最优报价。
Following our previous work on the control of multiple appliances in response to Independent System Operator (ISO) Regulation Service Signals (RSS), we model the ISO's RSS dynamics - evolving in a time scale of seconds - as a two level Markov process whose transition probabilities are calibrated on actual data. Appliance response is modeled as a Markov modulated process consistent with an exponentially distributed time to switch off and an expected utility that is concave in the price charged when an appliance turns on. Prices are broadcasted dynamically by a Smart Building Operator (SBO) with the objective of maximizing the time average of utility gained when appliances turn on minus the cost of imperfect RSS tracking. We prove certain properties of the stochastic Dynamic Programming (DP) policies that allow us to formulate the problem as an approximate Discrete State and Control Space DP and propose a reasonable approximation that renders the problem scalable to multiple appliance categories. The discretized state DP solution can be obtained as a solution to a Linear Program (LP). The LP provides the optimal dynamic price control policies and in addition yields the requisite information needed by the SBO to bid optimally for energy and Regulation Service Reserve (RSR) Capacity in the hour ahead balancing market. Solving for the discretized real time market optimal price policies, using them to calibrate a continuous analytic policy function, and extracting from the real time optimal policies the optimal bid to the hour ahead forward balancing market is the main contribution of this paper.