matlab仿真利用Matlab构造BP神经网络完成变压器故障诊断

《matlab仿真利用Matlab构造BP神经网络完成变压器故障诊断》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、利用MatIab构造BP神经网络完成变压器故障诊断a）样木数据：在汕色谱分析领域屮，基于汕屮溶气体类型与内部故障性质的对应关系,以汕屮5种特征气体为依据的判断变压器故障的方法。其特征量为H2,CH4,C2H4,C2H2,C2H6,如图3所示。在对变压器的故障识別吋，变压器的故障类型选择为:无故障、中低温过热、高温过热、低能放电和高能放电。序号故障性质-输入学习样本（相对值〉神经网络的期望输出H,CH.GH.GH.C,H>00010203041无故障46.2111.2733.04&520.614100002无故障41.1232.5114.45&360.52100003无故障33.3629.323

2、2.8927.282.25100004无故障45.8634.97&257.430.31100005低温过热14.9221.9817.1546.120.0010006低温过热0.87243.1827.1426.880.0010007低温过热3&1350.96&155.670.0010008低温过热37.9830.857.5723.010.0010009高温过热10.9921.2911.3052.982.380010010高温过热0.95816.012.895&011.160010011高温过热11.0322.513.3157.961.130010012髙温过热15.6821.8710.9853.

3、222.530010013低能放电5&0318.564.58&629.780001014低能放电86.896.485.281.130.00001015低能放电85.866.984.522.560.00001016低能放电83.687.965.153.020.560001017高能放电20.2316.961.6824.5235.740000118髙能放电26.7616.562.9838.7613.610000119高能放电43.9224.416,6223.910.5310000120高能放电4&0210.274.5222.3623.6200001b）仿真数据如图:样本H,CH

4、,实ea»17.25.63.S2.73.1iEt2120.0120.033.083.00.56低也过热320.619.67.560.91.52442971565980铁心多点按地515569334460有AO电62004616109128isastt电7981223129215展间魏檢不良8925642350中iao931.65.31.312.213.1囤鑽敖电107251213812005.6过热c）在Matlab小完成人工神经网络的创建、训练及最终的仿真，并给出仿真结果及训练误差曲线。完成如下:一、源代码如2%训练数据P=[46.2111.2733.048.520.614;41.1232.

5、5114.458.360.52;33.3629.3232.8927.282.25;45.8634.978.257.430.31;14.9221.9817.1546.120.0;0.87243.1827.1426.880.0;36.1350.968.155.670.0;37.9830.857.5723.010.0;10.9921.2911.3052.982.38;0.95816.012.8958.011.16;11.0322.513.3157.961.13;15.6821.8710.9853.222.53;58.0318.564.588.629.78;86.896.485.281.130.0;8

6、5.866.984.522.560.0;83.687.965.153.020.56;20.2316.961.6824.5235.74;26.7616.562.9838.7613.61;43.9224.416.6223.910.531;48.0210.274.5222.3623.62]';T=[10000;10000;10000;10000;01000;01000;01000;01000;00100;00100;00100;00100;00010;00010;00010;00010;00001;00001;00001;00001]';%仿真数据R二[7.25.63.52.73.1;120.012

7、0.033.083.00.56;20.619.67.560.91.52;42971565980;15569334460;2004616109128;981223129215;925642350;31.65.31.312.213.1;7251213812005.6]';S=[10000;01000;00100;00100;00010;00001;00100;01000;00001;00100]';%