ARTIFICIAL INTELLIGENCE BASED DROOP CONTROL FOR OPTIMAL POWER DISTRIBUTION IN MULTI-TERMINAL DC GRID

Abstract

In recent years, there has been an increased development and deployment of renewable energy resources to meet the ever-increasing electric power demand and to limit the use of fossil fuels. This has spurred offshore wind farms due to vast offshore wind energy potential. Large scale windfarms integration challenges such as need for long distance subsea power transmission and managing the variability of wind power variation on in the power grid. These challenges can be properly met by the use of Multi-Terminal Voltage Source Converter High Voltage DC transmission (MTDC) grid. MTDC system consists of three or more HVDC converter station connected to a common DC transmission network. Several control schemes were designed and analysed to use DC voltage as feedback signal and paid less intention by researchers. The DC voltage measured signal is used to vary DC reference voltage in AC side outer loop control and develops coordination between AC and DC sides. Thus, any uncertainty/variation in DC voltage feedback signal directly reflects on the performance of ith terminal in terms of reduction in power and voltage sharing capability. The DC terminal voltage is connected to DC bipolar transmission line and as results; it is enhanced probability of uncertainty. Hence, power sharing and voltage sharing capability can be modified using DC voltage feedback signal. As in the literature survey, it is found that most of the researchers have been taken only to improve power sharing capability without consideration of terminal DC voltage variations against AC grid load demand change, outage of terminals and integration of wind farms etc. It is not possible to get simultaneous minimum change in both terminal power and DC voltage without consideration of intelligent control strategy. Thus, this thesis is focused to reduce following research gaps as:  The fixed droop control method is not capable to maintain optimal power share among terminals in presence of abnormal conditions.  Also, it is required to power transfer from a remote generation over a large distance is regulated with enhanced power share and minimum oscillations against sudden change in load.  Further, the power sharing capability and DC voltage regulation is also distorted in presence of transmission line resistance.  During outage of a terminal in VSC-MTDC, the power share is effectively distributed.  The unpredictive natures of renewable energy are affected on MTDC grids stability in terms of over loading, oscillations in power and voltage. However, in addition of above gaps, the contribution of the research work can be put into three areas, namely control strategies used in VSC-MTDC system, analysis of ANN based droop control for optimal power distribution in VSC MTDC, impact analysis of transmission resistance, outage of terminals, wind power penetration impact and analysis of optimal power sharing in a meshed MTDC Grid via a terminal based Current Flow Controller in the following thesis Chapters. Chapter 1 deals with the introduction to VSC-MTDC system, need of MTDC terminals, types of various MTDC topologies, and discussed with their advantages and disadvantages, introduction to different control schemes used for MTDC control and VSC-MMC schemes also described. Chapter 2 discusses the detailed Literature survey in terms of scalability of controller design approach for VSC grid. The various control methods like voltage droop control, control based on network topology, voltage margin control, master slave control, PI control, optimal control, robust control, genetic control, coordinated control, adaptive droop control, Artificial Intelligence based droop control and control strategy when MTDC integrated with renewable energy sources. The advantages and drawbacks of each method have been discussed. Chapter 3 proposes a neural network based intelligent adaptive droop control strategy on a five terminals MTDC system to achieve optimal power share with minimum DC voltage deviations according to terminals rating and its topology. A three-layer Artificial Neural Network (ANN) error back propagation algorithm-based learning rule is merged with a Linear Quadratic Regulator (LQR). The ANN learning rule is used to estimate droop constants adaptively and LQR is utilized to minimize transient in power flow at minimum DC voltage fluctuation. The robustness of the proposed control strategy is illustrated with respect to transmission line admittance matrix uncertainty, step and random change of reference power at constant power terminals. Also, a comparative scenario is demonstrated in terms of peak with existing two controllers. The contribution in the area of ANN based intelligent adaptive droop control comprises results of step