模型預測最佳化應用於具儲能之智慧電網

Model Predictive Optimization for Energy Storage based Smart Grids

楊采真(TSAI-CHEN YANG)

指導教授 ： 熊博安

國立中正大學/工學院/資訊工程學系/碩士(2016年)

摘要

智慧電網是由數個微電為電網組成，每個微電網包含數個可再生發電能源、電池儲能系統與電力用戶。在大多數的智慧電網中，電池儲能系統通常扮演著不間斷電源的角色，即備用電源。近年來，電池儲能系統不再只是充當備用電源，而是成為以減少智慧電網用電成本為目標的電力供應來源之一。然而，無論是充放電次數與充放電深度，都成為影響電池儲能系統壽命的主要因素。
因此，為了降低智慧電網的總體成本，其中包含電力買賣成本與電池儲能系統使用成本，因此應該視使用電池儲能系統為成本的支出，並降低電池儲能系統的生命損耗。為了解決總成本的問題，本論文提出一套配電管理架構，並利用模型預測優化應用於具儲能之智慧電網配電管理策略，透過自相關移動平均模型預測未來的電力狀況，再經由基因演算法找到最佳的電力買賣與電池儲能系統的充放電情形。
基於本文所提出之配電管理策略，實驗結果呈現本文所提出之方法具有兩大優勢，分別為降低總體成本與減少電池儲能系統壽命損耗。在預測用電量與發電量部分，本文採用自相關移動平均預測模型，其誤差率小於10%，而此預測模型的平均執行時間為3.13秒。本文所提出之優化方法與Model Predictive Control Look-ahead Dispatch 方法相比，不僅能降低0.85%的總體成本支出，並同時能減少12.18%的電池儲能系統壽命損耗。

關鍵字

電池儲能系統 ； 電池壽命損耗 ； 基因演算法 ； 自相關移動平均模型 ； 模行預測優化控制 ； 配電管理

並列摘要

A smart grid is usually composed of multiple micro-grids, each of which includes renewable power generators, energy storage systems (ESS) and power consumers. In most smart grids, ESS is used as an uninterruptible power supply (UPS), for power backup purposes. In recent years, ESS has also started to act as an active electricity supplier so as to minimize overall electricity costs in a smart grid. However, the ESS lifetime decrease with each charging/discharging.

For a fair evaluation of the overall cost of electricity in smart grids, besides electricity trading costs, the cost incurred due to Loss of Life (LoL) via ESS usage should also be considered. The target problem solved in this Thesis is to find a near-optimal schedule which includes the electricity usage of the smart grid such as electricity trading and ESS usage. As a solution to the target problem, the Thesis proposes a Model Predictive Optimization (MPO) method for distribution management in smart grids. Future energy states are predicted using an Autoregressive Integrated Moving Average (ARIMA) model. According to the prediction results, a near-optimal ESS usage is found through the Genetic Algorithm (GA). A distribution management system architecture is also designed to support multiple micro-grids for a global scheduling with tradeoff between electricity trading and ESS LoL within a smart grid.

The experimental results demonstrate two benefits of our proposed methods, namely reducing overall cost and ESS LoL. For predicting demand loads and renewable energy generation, the error rate of prediction model is less than 10%, and the average execution time of prediction method is 3.13 seconds. By applying the proposed MPO method, the overall cost in a smart grid is reduced by reducing both electricity cost and the ESS LoL cost. Compared to the MPC look-ahead dispatch method, MPO achieve an overall cost saving and ESS LoL reduction by 0.85% and 12.18%, respectively.

並列關鍵字

Energy Storage System ； Loss of Life ； Genetic Algorithm ； Autoregressive Integrated Moving Average model ； Model Predictive Optimization ； Distribution Management

參考文獻

[2] X. Fang, S. Misra, G. Xue, and D. Yang. Smart grid-the new and improved power grid: A survey. IEEE Communications Surveys Tutorials, 14(4):944–980, December 2012.

[3] H. S. V. S. Kumar Nunna and S. Doolla. Energy management in microgrids using demand response and distributed storage - a multiagent approach. IEEE Transactions on Power Delivery, 28(2):939–947, April 2013.

[5] Department of Energy. The Smart Grid: An Introduction. http://energy.gov/oe/downloads/smart-grid-introduction-0, 2016.

[7] F. Birol. Energy and Climate Change. International Energy Agency, June 2015.

[9] R.E. Mackiewicz. Overview of IEC 61850 and benefits. In Proceedings of the IEEE Power Engineering Society GeneralMeeting, pages 623–630, October 2006.