经典低阶频率响应模型可快速计算各项频率指标,但由于高比例新能源系统扰动类型多样,运行方式复杂多变,难以准确获取系统参数和扰动功率大小,同时模型本身线性化会引起较大误差,导致频率预测值和实际值存在较大差异。为使频率响应模型...经典低阶频率响应模型可快速计算各项频率指标,但由于高比例新能源系统扰动类型多样,运行方式复杂多变,难以准确获取系统参数和扰动功率大小,同时模型本身线性化会引起较大误差,导致频率预测值和实际值存在较大差异。为使频率响应模型适应实际应用场景中高精度的要求,该文提出了模型-数据融合驱动的频率稳定智能增强判别方法(model-data driven intelligent enhanced method for frequency stability discrimination,MD-IEFSD),利用扰动初期频率响应数据对模型关键参数进行辨识,建立结合卷积神经网络和注意力机制的CNN-Attention频率参数预测模型,构建了融合参数预测误差和频率响应曲线预测误差的损失函数,引入了参数的敏感性和学习速率的分析,实现了频率稳定性的准确判别。最后以中国电科院万节点测试系统为算例,验证所提方法的可行性和有效性。展开更多
In order to realize information construction on settlement of pile-group foundation of Sutong Bridge, the monitoring instruments of high-precision micro-pressure sensor and hydrostatic leveling and settlement profiler...In order to realize information construction on settlement of pile-group foundation of Sutong Bridge, the monitoring instruments of high-precision micro-pressure sensor and hydrostatic leveling and settlement profiler were integrated synthetically. A set of practical multi-scale monitoring system on settlement of super-large pile-group foundation in deep water was put forward. The reliable settlement results are obtained by means of multi-sensor data fusion. Finite element model of pile-group foundation is established. By analysis of finite element simulated calculation of pile-group foundation, rules of settlement and uneven settlement obtained by monitoring and calculation results are coincident and the absolute error of settlement between them is 4.7 mm. The research shows that it is reasonable and feasible to monitor settlement of pile-group foundation with the system, and it can provide a method for the same type pile-group foundation in deep water.展开更多
文摘经典低阶频率响应模型可快速计算各项频率指标,但由于高比例新能源系统扰动类型多样,运行方式复杂多变,难以准确获取系统参数和扰动功率大小,同时模型本身线性化会引起较大误差,导致频率预测值和实际值存在较大差异。为使频率响应模型适应实际应用场景中高精度的要求,该文提出了模型-数据融合驱动的频率稳定智能增强判别方法(model-data driven intelligent enhanced method for frequency stability discrimination,MD-IEFSD),利用扰动初期频率响应数据对模型关键参数进行辨识,建立结合卷积神经网络和注意力机制的CNN-Attention频率参数预测模型,构建了融合参数预测误差和频率响应曲线预测误差的损失函数,引入了参数的敏感性和学习速率的分析,实现了频率稳定性的准确判别。最后以中国电科院万节点测试系统为算例,验证所提方法的可行性和有效性。
基金Project(2002CB412707) supported by the National Basic Research Program of ChinaProject(2006BAG04B05) supported by the National Science and Technology Pillar Program during the 11th Five-Year Plan of ChinaProject(2010B14414) supported by the Scientific Research Program of Center University in China
文摘In order to realize information construction on settlement of pile-group foundation of Sutong Bridge, the monitoring instruments of high-precision micro-pressure sensor and hydrostatic leveling and settlement profiler were integrated synthetically. A set of practical multi-scale monitoring system on settlement of super-large pile-group foundation in deep water was put forward. The reliable settlement results are obtained by means of multi-sensor data fusion. Finite element model of pile-group foundation is established. By analysis of finite element simulated calculation of pile-group foundation, rules of settlement and uneven settlement obtained by monitoring and calculation results are coincident and the absolute error of settlement between them is 4.7 mm. The research shows that it is reasonable and feasible to monitor settlement of pile-group foundation with the system, and it can provide a method for the same type pile-group foundation in deep water.
