电网运行方式辅助决策的研究
|
摘要
随着计算机和网络技术的发展，电网调度自动化系统已由最初的第一代简单SCADA发展到新的第四代，新一代系统采用了先进的开放分布式应用环境的网络管理技术，以及最新的软硬件、通信技术并采用了国际标准。新系统的整体性能和功能比前几代都发生了质的飞越，已与第一代不可同日而语，为科学地电网调度提供了有力的支撑。
网络分析(NA)系统是调度自动化系统的重要组成部份。在国外，有不少电力系统已实现了EMS的部分或全部功能。在我国，大部分电网是处于SCADA阶段，只有少数大网和省网进入了SCADA／AGC的阶段。电网运行方式的实际操作调整主要是通过联络线和变压器的投退和分接头的调节以及电容器组的投切等来进行，属于多目标决策、多运行参数调整配合问题。
有关专家学者对于区域电网的最优经济运行问题已做了大量的研究并取得了很多实用的成果，尤其是无功优化问题。但大多是根据负荷的变化适时调整，这样就造成电气设备的频繁投退，从而影响设备运行寿命和对电网运行带来不安全因素。另一方面，调度日趋复杂的电网结构，必须借助现代化高科技手段，扩展EMS的高级应用软件功能。在地区电网调度自动化系统中，目前的电网运行方式调整的普遍方法是针对某一时刻电网的运行状态，采用各种不同的优化算法优化调整各控制变量实现网损最小。若在电网运行中允许时刻跟随负荷的变化不断地调整，则固然能实现最优经济运行的目的，但限于设备运行寿命以及调整和控制次数，从应用的情况来考虑是不切实际的。
本文研究将负荷曲线按时间分段进行，可控变量在该时段内保持恒定然后在每一段时间内进行优化调整，达到电网经济运行的最优。
首先进行负荷预测，负荷预测传统的方法有多种，如：多元线性回归法、指数平滑法等。它们都需要建立未来负荷与影响负荷的各个因素之间的解析表达式，这往往非常困难，也是影响负荷预报精度的原因之一。人工神经网络方法不需建立这样的解析表达式，只是通过样本对网络进行训练，建立网络参数，就可以对负荷进行辨识，对未来做出预测。本文用BP神经网络，将负荷数据和影响负荷的因素进行量化，通过样本训练建立网络参数对未来负荷进行预测，算例预测的总误差为4.63％，满足工程需要。
通过预测得到的负荷曲线，按时间进行分段，分段时依照总方差最小的原则进行，采用步长加速法寻优计算，这样确保了负荷在一段时间内的相对平稳性，从
郑州人学工学硕士学位论文
而使电网运行的经济性得到保证。
电网运行通常采用的是开式网的方式，本文运用网络拓扑理论，求出电网所有
可能的运行方式集，即求出网络图的全部树。并可根据具体的实际情况对其进行
筛选，确定运行方式集的可行子集。
对于网络运行方式的可行子集，应用优化理论中的网络最大流法进行优化和
潮流计算得到网络损耗，通过比较得出了最经济的运行方式。在此基础上，进行
变压器分接头调整和无功优化计算并保证电压的质量。
本文按时间对负荷曲线进行分段，进行电网运行的最优经济方式计算。区别
于以往的根据负荷的变化进行实时调整，避免了电气设备的频繁操作，兼顾设备
的运行寿命，提高电网运行的综合效益。算例表明本文方法可行，效果明显，具
有实用性和普遍意义。
With the developing of computer and network technology, the power dispatching automation system has evolved from the initial generation that is simple SCADA to the present fourth, the new system is the advanced system that the opening and distributing network environments are used, not only the newest software and hardware technology are applied, but the international criterions are adopted. Compared with the foregoing generations, its integrated capability and functions become flying in essence, the system has been providing the strong assistances.
The network analysis system is an important part of power dispatching automation system. Overseas part or all of EMS functions have realized in many electric systems. Most interior systems are in stage of SCADA, only several large power networks and provincial networks have been in SCADA and AGC applications. The electric grid running modes are changed through linking lines and transformer launching or dropping out, adjusting the taps of the transformer, the capacitor groups launching or dropping out. It is the problem that many objects and operating parameters are regulated and harmonized.
Lots of specialists have studied deeply on the optimization of the area electric network, and gained many applied fruits, especially on optimization of the inactive power. But most means are that the operating methods are adjusted in real time according to the loads change, thus it causes the electric equipments launching and dropping out continually, this will affect the operating life-span and bring the unsafe factors to power network. On the other hand, only the modernization high-tech means must be applied and expanded the EMS advanced functions of applying software, the increasing complex network structure is dispatched. In the dispatching automation of the area electric network, the present mean method is, according to a certain stage, that vary optimization arithmetic is used to adjust every control variable in order to realize the object that the loss of grid is least. No doubt the optimization economical running can be realized, but in practice it is not feasible because of the limits of equipment o
perating life and adjusting times.
In this paper it is studied that the loads curve is divided into several parts according to the time, the control variables keep invariable, and then the optimization adjustment
is done in the period so that the most economical running scheme is gained.
In first, the loads forecasting is done and there are many method such as the many factors linear regression method, the index flatness method and so on. In these methods the analytic expressions are required to be created among the loads and the influencing factors, it is very difficult in general and is one of the reasons for affecting forecasting precision. The analytic expressions needn't be created in ANN(artificial neural network ) method, and the network parameters will come out through the samples training on it. The loads for future may be forecasted by the trained network. In this paper the BP ANN is applied, with the transacted loads and influencing factors data the network is trained, the loads are forecasted with the trained network, the gross error of forecasted data reaches to 4.63 percent, and the results can meet the demands of the project.
The forecasted loads curves are divided into several parts by the time. The principle that the total deviations are least is applied, and the results calculated by the step accelerating method, thus assures that the loads are relatively stable in some a period so that the electric grid economical running is guaranteed.
The running mode of grid is commonly open, so the topology theory is used to compute all the running modes set, namely all the trees come out. According to the actual things the trees may be filtered and the practical subset of running mode is made certain.
With regard to the practical subset, the optimizing arithmetic such as the network maximum flow method is used, and through the flows computing the gross loss of the
网络分析(NA)系统是调度自动化系统的重要组成部份。在国外，有不少电力系统已实现了EMS的部分或全部功能。在我国，大部分电网是处于SCADA阶段，只有少数大网和省网进入了SCADA／AGC的阶段。电网运行方式的实际操作调整主要是通过联络线和变压器的投退和分接头的调节以及电容器组的投切等来进行，属于多目标决策、多运行参数调整配合问题。
有关专家学者对于区域电网的最优经济运行问题已做了大量的研究并取得了很多实用的成果，尤其是无功优化问题。但大多是根据负荷的变化适时调整，这样就造成电气设备的频繁投退，从而影响设备运行寿命和对电网运行带来不安全因素。另一方面，调度日趋复杂的电网结构，必须借助现代化高科技手段，扩展EMS的高级应用软件功能。在地区电网调度自动化系统中，目前的电网运行方式调整的普遍方法是针对某一时刻电网的运行状态，采用各种不同的优化算法优化调整各控制变量实现网损最小。若在电网运行中允许时刻跟随负荷的变化不断地调整，则固然能实现最优经济运行的目的，但限于设备运行寿命以及调整和控制次数，从应用的情况来考虑是不切实际的。
本文研究将负荷曲线按时间分段进行，可控变量在该时段内保持恒定然后在每一段时间内进行优化调整，达到电网经济运行的最优。
首先进行负荷预测，负荷预测传统的方法有多种，如：多元线性回归法、指数平滑法等。它们都需要建立未来负荷与影响负荷的各个因素之间的解析表达式，这往往非常困难，也是影响负荷预报精度的原因之一。人工神经网络方法不需建立这样的解析表达式，只是通过样本对网络进行训练，建立网络参数，就可以对负荷进行辨识，对未来做出预测。本文用BP神经网络，将负荷数据和影响负荷的因素进行量化，通过样本训练建立网络参数对未来负荷进行预测，算例预测的总误差为4.63％，满足工程需要。
通过预测得到的负荷曲线，按时间进行分段，分段时依照总方差最小的原则进行，采用步长加速法寻优计算，这样确保了负荷在一段时间内的相对平稳性，从
郑州人学工学硕士学位论文
而使电网运行的经济性得到保证。
电网运行通常采用的是开式网的方式，本文运用网络拓扑理论，求出电网所有
可能的运行方式集，即求出网络图的全部树。并可根据具体的实际情况对其进行
筛选，确定运行方式集的可行子集。
对于网络运行方式的可行子集，应用优化理论中的网络最大流法进行优化和
潮流计算得到网络损耗，通过比较得出了最经济的运行方式。在此基础上，进行
变压器分接头调整和无功优化计算并保证电压的质量。
本文按时间对负荷曲线进行分段，进行电网运行的最优经济方式计算。区别
于以往的根据负荷的变化进行实时调整，避免了电气设备的频繁操作，兼顾设备
的运行寿命，提高电网运行的综合效益。算例表明本文方法可行，效果明显，具
有实用性和普遍意义。
With the developing of computer and network technology, the power dispatching automation system has evolved from the initial generation that is simple SCADA to the present fourth, the new system is the advanced system that the opening and distributing network environments are used, not only the newest software and hardware technology are applied, but the international criterions are adopted. Compared with the foregoing generations, its integrated capability and functions become flying in essence, the system has been providing the strong assistances.
The network analysis system is an important part of power dispatching automation system. Overseas part or all of EMS functions have realized in many electric systems. Most interior systems are in stage of SCADA, only several large power networks and provincial networks have been in SCADA and AGC applications. The electric grid running modes are changed through linking lines and transformer launching or dropping out, adjusting the taps of the transformer, the capacitor groups launching or dropping out. It is the problem that many objects and operating parameters are regulated and harmonized.
Lots of specialists have studied deeply on the optimization of the area electric network, and gained many applied fruits, especially on optimization of the inactive power. But most means are that the operating methods are adjusted in real time according to the loads change, thus it causes the electric equipments launching and dropping out continually, this will affect the operating life-span and bring the unsafe factors to power network. On the other hand, only the modernization high-tech means must be applied and expanded the EMS advanced functions of applying software, the increasing complex network structure is dispatched. In the dispatching automation of the area electric network, the present mean method is, according to a certain stage, that vary optimization arithmetic is used to adjust every control variable in order to realize the object that the loss of grid is least. No doubt the optimization economical running can be realized, but in practice it is not feasible because of the limits of equipment o
perating life and adjusting times.
In this paper it is studied that the loads curve is divided into several parts according to the time, the control variables keep invariable, and then the optimization adjustment
is done in the period so that the most economical running scheme is gained.
In first, the loads forecasting is done and there are many method such as the many factors linear regression method, the index flatness method and so on. In these methods the analytic expressions are required to be created among the loads and the influencing factors, it is very difficult in general and is one of the reasons for affecting forecasting precision. The analytic expressions needn't be created in ANN(artificial neural network ) method, and the network parameters will come out through the samples training on it. The loads for future may be forecasted by the trained network. In this paper the BP ANN is applied, with the transacted loads and influencing factors data the network is trained, the loads are forecasted with the trained network, the gross error of forecasted data reaches to 4.63 percent, and the results can meet the demands of the project.
The forecasted loads curves are divided into several parts by the time. The principle that the total deviations are least is applied, and the results calculated by the step accelerating method, thus assures that the loads are relatively stable in some a period so that the electric grid economical running is guaranteed.
The running mode of grid is commonly open, so the topology theory is used to compute all the running modes set, namely all the trees come out. According to the actual things the trees may be filtered and the practical subset of running mode is made certain.
With regard to the practical subset, the optimizing arithmetic such as the network maximum flow method is used, and through the flows computing the gross loss of the
引文
[1] 陈允平，王旭蕊，韩宝亮．人工神经网络原理及其应用．北京：中国电力出版社，2002．50～51，53～56．
[2] 闻新，周露，王丹力等．MATLAB神经网络应用设计．北京：科学出版社，2000．
[3] 贺蓉，曾刚，姚建刚等．天气敏感型神经网络在地区电网短期负荷预测中的应用．电力系统自动化，2001，(17)：32～35．
[4] Zhang Y, Ding X, Liu Y, Griffin P J. An Artificial neural network approach to transformer fault diagnosis. IEEE Transaction on Power Delivery, 1996,11 (4): 1836～1842.
[5] 魏权龄，王日爽，徐兵．数学规划引论．北京：北京航空航天大学出版社．1991：216～242．
[6] 陈树柏．网络图论及其应用．北京：科学出版社．1982．7：131～155．
[7] Papadakis SEP, Theocharis JB,Kiaartzis SJ,Bakirtzis AG. A novel approach to short-term load forecasting using fuzzy neural networks. IEEE Trans. on Power Systems, 1998,13(2) :480～498
[8] 肖国泉，王春，张福伟．电力负荷预测．北京：中国电力出版社．2001．7：4～14．
[9] Kenji Iba. Reactive power optimization by genetic algorithm. IEEE Trans on Power System, 1994,9(2):685～692
[10] 李玉惠，韩曾晋．电力短期负荷预报模糊：冲经网络模型的研究．昆明理工大学学报，2000，(4)，36～40．
[11] Dash P K, Satpathy H P, Liew A C. A real-time short-time peak and average load forecasting system using a self-organizing fuzzy neural networks, engineering applications of artificial intelligence, 1998,(11):307～316.
[12] 张喜林，王丽娟．地区电力调度EMS应用功能需求及算法．电力系统自动化．2002，26(16)：71～74
[13] Daneshdoost M, Lotfalian M,Bumroonggit G. Ngoy J R Neural network with fuzzy set-based classification for short-term load forecasting. IEEE Trans. on Power System,1998,13(4):1386～1391.
[14] 周全仁，张清益．电网分析与发电计划．长沙，湖南科学技术出版社，1996．5：1～3．
[15] Qiu J, Shahidehpour S M. A new approach for minimizing power losses and improving voltage profile[J].IEEE Trans on Power Systems, 1987,2(2):287～302.
[16] 段俊东，陈昆薇．采用分时段模糊技术处理实现的短期负荷预报[J]．继电器，2002，30(11)：13～16．
[17] Francoise de Viron, Jean CLAUS, Francois DONGIER. Automation with neural network based techniques, of short-term load forecasting at Belgian nation control center. Japan: IEEE ANNPS'93 Yokohama,1993.55～59.
[18] 潘莹玉．我国电网调度自动化系统的发展与现状．继电器．2000，28(6)：58～62．
[19] Chen Jun,Wu Jie. A real-time short-term load forecasting system by artificial neural network. Journal of South China University of Technology. 1998,26(11):144-149.
[20] 辛耀中．新世纪电网调度自动化技术发展趋势．电网技术．2001，12(25)：1～10．
[21] V.Blamocanin. Optimal loss reduction of distributed networks. IEEE Transaction on Power Delivery,1990,5(3):774～782
[22] 杨秀台．电力网线损的理论计算和分析．北京：水利电力出版社．1985：37～53．
[23] Zhang Yang, Wang Haining, Zhang Zhigang,Zhang Rui. A practical method for Solving economic dispatch problem. 2002 International Conference on Power System Technology proceedings,2002, 2:241～244
[24] 陈珩．电力系统稳态分析(第二版)．北京：中国电力出版社．1995．
[25] 杨溢刚，杨其余，杨维汉．电力系统优化的理论基础．第一版．北京：水利电力出版社，1990．5：1～133
[26] Pilgrim, J.D., Li,F.,Aggarwal,R.K..Genetic algorithms for optimal reactive power compensation on the national grid system. Power Engineering Society Summer Meeting,2002 IEEE.2002,1:524～529
[27] Wu Xingping,Zhang Yang and Wang Xiwei. A new generation EMS.2002 International Conference on Power System Technology proceedings,2002,Vol.2:190～194
[28] E．米捏卡．李家莹译．网络和图的最优化方法．北京：中国铁道出版社．1984．
[29] 赵希正．中国电力负荷特性分析与预测．北京：中国电力出版社．2002．1：14～17．
[30] 陈开周．最优化计算方法．西安：西北电讯工程学院出版社,1983．8：159～183
[31] Parker C J,Morrison I F, Sutanto D. Application of an optimization method for determining the reactive margin from voltage collapse in reactive power planning. IEEE Trans. on Power System. 1996,Vol. 11,No.3.
[32] 鲁晓春，詹荷生．关于配送中心重心法选址的研究．北方交通大学学报．2000，24(6)：108～110
[33] 《运筹学》教材编写组．运筹学(第2版)．北京：清华大学出版社．1990：271～277．
[34] 张粒子，舒隽，林宪枢等．基于遗传算法的无功规划优化．中国电机工程学报．2000，6
[35] 倪伟，单渊达．具有优化路经的遗传算法应用于电力系统无功优化．电力系统自动化．2000，24(21)
[36] N.I.Deeb, S.M.Shahidehpour. Decomposition approach for minimizing real power losses in power system. IEE proceedings-c. 1991,138(1):27～37
[37] S.Granvill. Optimal reactive dispatch through interior point methods. IEEE Trans. on power system. 1998, 13(3):870～877
[38] 韩帧祥，文福栓，张琦．人工智能在电力系统中的应用．电力系统自动化．2000，24(2)
[39] 李林川，王建勇，陈礼义等．电力系统无功补偿优化规划．中国电机工程学报．1999，19(2)
[40] 王祖佑．电力系统稳态运行计算机分析．北京：水利电力出版社．1987．11
[41] 靳龙章，丁毓山．电网无功补偿实用技术．北京：中国水利水电出版社．1997．8：207～216
[2] 闻新，周露，王丹力等．MATLAB神经网络应用设计．北京：科学出版社，2000．
[3] 贺蓉，曾刚，姚建刚等．天气敏感型神经网络在地区电网短期负荷预测中的应用．电力系统自动化，2001，(17)：32～35．
[4] Zhang Y, Ding X, Liu Y, Griffin P J. An Artificial neural network approach to transformer fault diagnosis. IEEE Transaction on Power Delivery, 1996,11 (4): 1836～1842.
[5] 魏权龄，王日爽，徐兵．数学规划引论．北京：北京航空航天大学出版社．1991：216～242．
[6] 陈树柏．网络图论及其应用．北京：科学出版社．1982．7：131～155．
[7] Papadakis SEP, Theocharis JB,Kiaartzis SJ,Bakirtzis AG. A novel approach to short-term load forecasting using fuzzy neural networks. IEEE Trans. on Power Systems, 1998,13(2) :480～498
[8] 肖国泉，王春，张福伟．电力负荷预测．北京：中国电力出版社．2001．7：4～14．
[9] Kenji Iba. Reactive power optimization by genetic algorithm. IEEE Trans on Power System, 1994,9(2):685～692
[10] 李玉惠，韩曾晋．电力短期负荷预报模糊：冲经网络模型的研究．昆明理工大学学报，2000，(4)，36～40．
[11] Dash P K, Satpathy H P, Liew A C. A real-time short-time peak and average load forecasting system using a self-organizing fuzzy neural networks, engineering applications of artificial intelligence, 1998,(11):307～316.
[12] 张喜林，王丽娟．地区电力调度EMS应用功能需求及算法．电力系统自动化．2002，26(16)：71～74
[13] Daneshdoost M, Lotfalian M,Bumroonggit G. Ngoy J R Neural network with fuzzy set-based classification for short-term load forecasting. IEEE Trans. on Power System,1998,13(4):1386～1391.
[14] 周全仁，张清益．电网分析与发电计划．长沙，湖南科学技术出版社，1996．5：1～3．
[15] Qiu J, Shahidehpour S M. A new approach for minimizing power losses and improving voltage profile[J].IEEE Trans on Power Systems, 1987,2(2):287～302.
[16] 段俊东，陈昆薇．采用分时段模糊技术处理实现的短期负荷预报[J]．继电器，2002，30(11)：13～16．
[17] Francoise de Viron, Jean CLAUS, Francois DONGIER. Automation with neural network based techniques, of short-term load forecasting at Belgian nation control center. Japan: IEEE ANNPS'93 Yokohama,1993.55～59.
[18] 潘莹玉．我国电网调度自动化系统的发展与现状．继电器．2000，28(6)：58～62．
[19] Chen Jun,Wu Jie. A real-time short-term load forecasting system by artificial neural network. Journal of South China University of Technology. 1998,26(11):144-149.
[20] 辛耀中．新世纪电网调度自动化技术发展趋势．电网技术．2001，12(25)：1～10．
[21] V.Blamocanin. Optimal loss reduction of distributed networks. IEEE Transaction on Power Delivery,1990,5(3):774～782
[22] 杨秀台．电力网线损的理论计算和分析．北京：水利电力出版社．1985：37～53．
[23] Zhang Yang, Wang Haining, Zhang Zhigang,Zhang Rui. A practical method for Solving economic dispatch problem. 2002 International Conference on Power System Technology proceedings,2002, 2:241～244
[24] 陈珩．电力系统稳态分析(第二版)．北京：中国电力出版社．1995．
[25] 杨溢刚，杨其余，杨维汉．电力系统优化的理论基础．第一版．北京：水利电力出版社，1990．5：1～133
[26] Pilgrim, J.D., Li,F.,Aggarwal,R.K..Genetic algorithms for optimal reactive power compensation on the national grid system. Power Engineering Society Summer Meeting,2002 IEEE.2002,1:524～529
[27] Wu Xingping,Zhang Yang and Wang Xiwei. A new generation EMS.2002 International Conference on Power System Technology proceedings,2002,Vol.2:190～194
[28] E．米捏卡．李家莹译．网络和图的最优化方法．北京：中国铁道出版社．1984．
[29] 赵希正．中国电力负荷特性分析与预测．北京：中国电力出版社．2002．1：14～17．
[30] 陈开周．最优化计算方法．西安：西北电讯工程学院出版社,1983．8：159～183
[31] Parker C J,Morrison I F, Sutanto D. Application of an optimization method for determining the reactive margin from voltage collapse in reactive power planning. IEEE Trans. on Power System. 1996,Vol. 11,No.3.
[32] 鲁晓春，詹荷生．关于配送中心重心法选址的研究．北方交通大学学报．2000，24(6)：108～110
[33] 《运筹学》教材编写组．运筹学(第2版)．北京：清华大学出版社．1990：271～277．
[34] 张粒子，舒隽，林宪枢等．基于遗传算法的无功规划优化．中国电机工程学报．2000，6
[35] 倪伟，单渊达．具有优化路经的遗传算法应用于电力系统无功优化．电力系统自动化．2000，24(21)
[36] N.I.Deeb, S.M.Shahidehpour. Decomposition approach for minimizing real power losses in power system. IEE proceedings-c. 1991,138(1):27～37
[37] S.Granvill. Optimal reactive dispatch through interior point methods. IEEE Trans. on power system. 1998, 13(3):870～877
[38] 韩帧祥，文福栓，张琦．人工智能在电力系统中的应用．电力系统自动化．2000，24(2)
[39] 李林川，王建勇，陈礼义等．电力系统无功补偿优化规划．中国电机工程学报．1999，19(2)
[40] 王祖佑．电力系统稳态运行计算机分析．北京：水利电力出版社．1987．11
[41] 靳龙章，丁毓山．电网无功补偿实用技术．北京：中国水利水电出版社．1997．8：207～216