智能电网暂态故障检测和电流过载预防控制研究

发布时间：2018-12-14 13:11

【摘要】：随着我国经济的高速发展，电力能源需求不断增大，电力系统变得越来越庞大和复杂。由于新的电力设备不断被接入电网中，外部干扰下电网的安全隐患增大。电力系统网络是一非线性、规模大、强耦合、动态的复杂系统，传统电力网络监控系统的测量、计算、控制，通信缺乏广泛的协作，其灵活性和效率还有待提高。智能电网的出现为上述问题的解决提供了新的机遇。面向智能电网，研究新的理论和方法提高电力系统的可靠性和安全性具有重要的意义。本文围绕提高电力系统暂态安全性，对其故障检测方法，预防控制策略展开研究。根据电力系统的传输和动态特性，构建集成统一潮流控制器(Unified PowerFlow Controller, UPFC)的电力系统暂态数学模型。提出了基于极点配置局部递归全局前馈((Locally Recurrent Global Forward, LRGF)动态神经网络建模方法，并分别讨论了基于小波提升和基于在线自适应主元分解的电网暂态故障检测。最后，针对电力系统传输线路电流过载和暂态非稳定情况，提出了采用UPFC作为控制手段，基于障碍函数和能量函数的一种预防控制策略。仿真结果验证了提出方法的有效性。 ①针对电网暂态过程基于数据的建模，提出了一种基于极点配置LRGF神经网络。由于对于动态神经元的极点存在于实轴上和一对共轭复数极点两种情况，为了避免参数到稳定区域投影的复杂性，提出的神经网络将隐层神经元内动态滤波器的极点被划分为依据极点的情况将神经元分成实极点和复极点两部分，通过函数权值的方法将这两种情况极点的动态部分加权输出，同时针对这种新的神经网络特别的采用了求导梯度下降的学习算法，通过极点投影和权值调节学习计算实现对电网暂态特性建模。 ②针对电网暂态故障检测中残差信号分析，提出了一种基于小波提升和自适应阈值的检测方法。根据残差信号和小波函数最优设计原理自适应地设计小波预测算子和更新算子。通过小波提升方法，将极点配置LRGF动态神经网络输出与电力系统输出作差得到的残差信号分解为细节信号和逼近信号提取故障特征。通过自适应阈值检测细节信号和逼近信号，以及容忍时间方法检测缓变和突变故障。仿真结果验证了此方法在电网暂态故障检测中的有效性。 ③针对电网在线暂态故障检测中残差信号分析中数据处理问题，，提出了一种基于在线自适应主元分解算法。提出的在线自适应主成份分解算法通过以残差信号为输入的主元向量迭代，快速计算主元特征向量，建立主元模型。通过主元变换降低被检测信号维度，得到残差信号的主元得分。根据主元得分计算T2统计变量和Q统计变量。通过T2统计量反应系统PCA模型内部变化，Q统计量反应PCA模型与信号偏差的原理，检测系统故障。仿真实例验证了算法的有效性。 ④为应对电网暂态过程中电网传输线路电流过载，在统一潮流控制器(UPFC)下，提出了基于障碍函数和能量函数的暂态电流过载预防控制方法。通过能量函数方法分析电力系统在故障后的暂态稳定性。根据稳定分析结果，实施预防控制策略。与基于仿真法和人工智能法的预防控制不同，文中通过构建了一种由电网能量函数和障碍函数组成的控制李亚普诺夫函数，得到控制率。控制器通过障碍函数约束边界数值无穷大的特性，以及统一潮流控制器的作用，阻止电网传输线路暂态电流过载。采用最近非稳定平衡点UEP的方法分析控制系统的稳定性，并通过优化算法调节障碍函数重塑系统稳定区域。通过对3节点电力系统和162节点电力系统的仿真结果证明了本文提出的预防控制方法的有效性。
[Abstract]:With the high-speed development of our country's economy, the power demand of electric power is increasing, and the power system becomes more and more bulky and complex. the potential safety hazard of the power grid is increased due to the fact that the new power equipment is constantly being accessed into the power grid. The network of power system is a non-linear, large-scale, strong-coupled and dynamic complex system. The measurement, calculation, control and communication of the traditional power network monitoring system lack extensive cooperation, and its flexibility and efficiency are still to be improved. The emergence of the smart grid provides a new opportunity for solving the above problems. It is of great significance to study the new theory and method to improve the reliability and safety of the power system. In this paper, the transient security of power system is improved, the fault detection method and the prevention control strategy are developed. The power system transient mathematical model of the integrated power flow controller (UPFC) is built according to the transmission and dynamic characteristics of the power system. In this paper, a local recursive global forward (LRGF) dynamic neural network modeling method based on pole assignment is proposed, and the power grid transient fault detection based on wavelet lifting and on-line self-adaptive main element decomposition is discussed. Finally, aiming at the current overload and transient non-stability of the transmission line of the power system, a control method using UPFC as a control means, an obstacle function and an energy function is proposed. The simulation results verify the effectiveness of the proposed method. Based on the data-based modeling of the transient process of the power grid, an LRGF God based on pole configuration is proposed. through the network, since the poles of the dynamic neuron are present on the real axis and the pair of common complex poles, in order to avoid the projection of the parameters to the stable region, The complexity of the neural network is that the pole of the dynamic filter in the hidden layer neuron is divided into two parts of the real pole and the complex pole according to the case of the pole, and the dynamic part of the two cases is added by the method of the function weight value. The power output, in addition to the new neural network, adopts the learning algorithm of the gradient descent of the derivation gradient, and the power grid transient is realized through the pole projection and weight adjustment learning calculation. Based on the analysis of residual signal in power grid transient fault detection, a wavelet-based lifting and adaptive threshold is presented. The method for detecting the small wave is adaptively designed according to the residual signal and the design principle of the wavelet function. and the residual signal obtained by the difference between the output of the LGF dynamic neural network and the output of the power system is decomposed into a detail signal and an approximation signal by a small wave lifting method. Taking the fault features, detecting the detail signal and the approximation signal through the adaptive threshold, and detecting the slow change by the tolerance time method. and the simulation results verify that the method is in the power grid transient fault detection In order to solve the problem of data processing in the analysis of residual signal in the on-line transient fault detection of the power grid, an on-line self-adapting is proposed. An on-line adaptive main component decomposition algorithm is proposed. The main component eigenvector is calculated by using the residual signal as the input main element vector. and reducing the dimension of the detected signal by the main element transformation to obtain a residual error, The main element score of the signal is calculated according to the main element score. Statistical variable of quantity and Q. Through the analysis of the internal change of the PCA model of the reaction system of T2 statistic, the principle of the response of the quantity of Q statistics to the deviation of the signal and detecting system failure. The simulation example verifies The effectiveness of the algorithm is presented. In order to deal with the current overload of the transmission line during the transient state of the power grid, under the unified power flow controller (UPFC), the transient electricity based on the barrier function and the energy function is put forward. The invention relates to a flow overload prevention control method, Transient stability after failure. Based on the results of the stability analysis In contrast to the control of control based on the simulation method and the artificial intelligence method, the paper constructs a control Lyapunov function composed of the power function and the obstacle function of the power grid. The controller limits the characteristic of the boundary value infinite through the obstacle function, and the function of the unified power flow controller to prevent the transmission of the power grid. The transient current of the transmission line is overloaded. The stability of the control system is analyzed by the method of the last non-stable equilibrium point UEP, and the obstacle letter is adjusted by the optimization algorithm. In this paper, the simulation results of the three-node power system and the 162-node power system prove the pre-existing problems in this paper.
【学位授予单位】：重庆大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM76;TM73

【参考文献】