有源配电网有功无功协调控制研究

论文总字数：27702字

摘 要

16012507 詹惠瑜

指导教师 顾 伟

近年来，随着传统配电网向新一代更加智能化、数字化和自动化配电网的逐步转变，越来越多的利用可再生清洁能源的分布式电源以及新型电动汽车充电站接入了配电网，然而，由于风能、太阳能等可再生能源的间歇性和随机性，电动汽车用户充电需求的无序性，使得电网负荷呈现出的规律更为复杂，这给电网的稳定运行带来了大量的不确定性，电网中各节点处的功率越来越多的变化为波动的随机变化量，这使得配电网系统功率的预测难度增加，也给含大规模分布式电源及充电站的配电网的协调控制带来了困难与障碍。

为解决配电网节点功率预测以及有功无功协调控制的问题，支撑配电网优质高效运行，提高分布式电源和电动汽车充电站的有效利用，本文提出了一种基于小波算法的神经网络电网节点功率预测算法及基于蒙特卡洛模拟的电动汽车充电站预测算法，并加入马尔科夫链对模型更新滚动，以此作为配电网潮流计算的基础，并对配电网有功无功协调控制进行了研究，主要研究内容如下：

（1）提出了一种新型的通用于负荷以及分布式电源的电网节点功率预测建模方法。本文在对配电网日常负荷、风电、光伏数据进行分析的基础上，考虑温度、天气、风速和太阳光照强度等影响因素，提出一种通用且较为精确的小波神经网络预测模型，并结合马尔科夫链对进行滚动修正。

（2）提出了一种基于蒙特卡洛模拟算法的电动汽车充电站负荷预测方法。从电动汽车充电负荷的主要影响因素入手，分析了公交车、公用车和私用车等多种电动汽车充电行为的差异，采取蒙特卡洛模拟法对电动汽车的充电开始时段、充电结束时段、充电模式、充电能量需求以及荷电状态进行抽样，模拟电动汽车充电站的日常运行状况，对其功率预测模型进行了搭建及仿真。

（3）提出了配电网有功无功协调优化调度模型。本文将风电、光伏、电动汽车、储能以及SVC纳入配电网的优化调度中，以前文的功率预测为基础，建立了以配电网有功网损最小为目标的调度模型，并通过二阶锥优化的方法对模型进行求解，最后采用修改后IEEE 33节点配电网络进行算例的验证。

关键词：配电网功率预测；分布式电源；电动汽车充电站；有功无功协调控制；蒙特卡洛模拟；神经网络；马尔科夫过程

Abstract

16012507 ZHAN Huiyu

Supervisor： GU Wei

In recent years, with the gradual transformation of the traditional distribution network to a more intelligent, digital and automation generation, more and more distributed power supply used of renewable energy and new electric vehicle charging stations are connected to the distribution network. However, due to the intermittent and randomness of wind, solar and other renewable energy, the disorder of charging demand of electric vehicle users, the power load showed a more complex rules. It also brings a lot of uncertainty to the stable operation of the power grid, for example, the power of each node in the power grid is more and more variable. This makes it difficult to predict the power of distribution network system, and it also brings difficulties and obstacles to the coordination control of distribution network with large scale distributed power supply and charging stations.

In order to solve the problem of node power prediction and active reactive power coordinated control in distribution network, support the high quality and efficient operation of distribution network, and improve the effective utilization of the distributed power and electric vehicle charging station, a new algorithm of neural network node power prediction algorithm based on wavelet algorithm and the prediction algorithm of electric vehicle charging station based on Monte Carlo simulation is proposed in this paper. Meanwhile, the typical 33 node distribution network is modified, which is connected with wind power generation, photovoltaic power generation and electric vehicle charging station node, respectively. In addition, we add a Markov chain to update the rolling model as a basis for the calculation of active reactive power flow of distribution network, the control of active and reactive power of distribution network is studied. The main research contents are as follows:

(1) A new method for power grid node power prediction model is proposed, which is applied to the load and distributed generation. Based on the analysis of daily load of distribution network, the typical wind power generation and photovoltaic power generation in the distributed power system, a general and accurate prediction model of wavelet neural network is proposed, which is combined with the Markov chain to modify the network.

(2) Use the Monte Carlo simulation algorithm to realize the load forecasting method of electric vehicle charging station. From considering the main influence factors of electric vehicle charging load, analysis the differences of buses, public vehicles and private cars and other kinds of electric vehicle charging behavior. In their respective charging mode, the charging time of the electric vehicle charging, the charging period, the energy requirement and the charging state of the electric vehicle are sampled by Monte Carlo simulation method. In the end, the load power prediction model of electric vehicle charging station is built and simulated.

(3) The optimal dispatch model of active power and reactive power in distribution network is proposed. In this paper, wind power, photovoltaic, electric vehicles, energy storage and SVC are incorporated into the optimal scheduling of distribution network. A scheduling model with the minimum active power loss in distribution network is established, based on the previous power prediction, and the model is solved by the method of two order cone optimization. Finally, a modified IEEE 33 node distribution network is used to verify the case.

Key words: Power Prediction of Distribution Network; Distributed Power System; Electric Vehicle Charging Station; Active and Reactive Power Coordinated Control; Monte Carlo Simulation; Neural Network; Markov Process

目 录

摘 要 I

Abstract II

目 录 IV

第一章 绪论 1

1.1 课题背景及研究意义 1

1.2 课题研究现状 1

1.2.1功率预测研究 1

1.2.2 有源配电网运行控制研究 2

1.3 本文主要工作 3

第二章 典型负荷节点以及分布式电源接入点的功率预测 5

2.1 负荷与分布式电源预测模型概述 5

2.1.1 人工神经网络概述 5

2.1.2 小波神经网络概述 6

2.2 小波神经网络的算法 7

2.3 算例分析 9

2.3.1 网络模型参数确定 9

2.3.2 网络模型结构设计 11

2.3.3 程序运行与预测结果 11

2.4 本章小结 18

第三章 电动汽车充电站的负荷建模 20

3.1 考虑复杂影响因子的充电负荷模型 20

3.1.1 电动汽车充电负荷的影响因素 20

3.1.2 电动汽车充电模式分析 22

3.2 实例仿真 22

3.3 本章小结 23

第四章 配电网有功无功协调优化调度 25

4.1 配电网有功无功协调优化调度模型 25

4.2 模型的二阶锥转化 27

4.3 算例分析 28

4.4 本章小结 31

第五章 总结与展望 32

5.1 总结 32

5.2 展望 32

致谢 33

参考文献 34

第一章 绪论

剩余内容已隐藏，请支付后下载全文，论文总字数：27702字