Capacity Sizing Optimization And Control Strategy For Stand Alone New Energy Hybrid Power Generation System

With the development of global economy, the demand for energy is growing gradually. Energy crisis and environmental protection tension have also been paid attention all over the world. Therefore, renewable energy power generation technology has been studied and developed both at home and abroad. Setting up hybrid renewable energy system in remote areas not only is consistent with the national energy policy, but also is an effective way to solve the electricity supplying problem in remote areas. Considering the renewable energy application demands and referring to a lot of related books and literatures, this dissertation explores some key technical and difficulties in this field, and focuses on the optimal design and control strategy in hybrid renewable energy system. The main contents of this dissertation are as follows.(1) At the beginning of this dissertation, the structure of a hybrid renewable energy power system will be introduced. Mathematical models of the main components in the hybrid renewable power system are also described in the first part, including wind generator system’s model, photovoltaic (PV) power generation’s model, fuel cost of diesel generator’s model and the model of batteries. Then, a genetic algorithm (GA) based optimization method is proposed to optimize the capacity allocation of the hybrid renewable power system. The optimization process uses the minimum operating cost of the power system as the fitness function and constructs the constraint conditions like the power balance, subsequently, GA is applied to search the optimal sizing results of system components. Using the practical data, the simulation results in MATLAB have showed that the hybrid renewable power system optimized by GA has the advantages of high reliability, low comprehensive cost and less emissions.(2) In order to utilize the solar power effectively, this dissertation proposes a short-term power forecast method of photovoltaic using a combination model. The proposed forecast model is a combination of weighted vertical sub-model and horizontal sub-model, and each sub-model adopts support vector machine (SVM) to predict its output power. In the proposed combination model, the input variables of the horizontal sub-model is irradiation intensity, the sun incidence angle, temperature, weather conditions and so on, and the output of this sub-model is short-term power of photovoltaic. The input variables of the longitudinal sub-model is four sequence values of solar power, and the output value of this sub-model is the solar power in the next moment. To achieve the optimal short-term power forecast, a least square method is used to optimize the weighted coefficient of each sub-model. Besides, this dissertation introduces the design principle and implementation process of the proposed combination model in detail. The applied SVM method has excellent learning ability and generalization performance, which can improve the forecast accuracy after learning. At the same time, the simulation results under the different weather conditions have also verified the effectiveness of the proposed short-term power forecast model of photovoltaic.(3) Generally speaking, the wind power is greatly influenced by the climate, weather and other natural conditions, which makes the wind power unstable and uncertain. For this reason, this dissertation proposes a novel wind power short-term forecast model based on empirical mode decomposition (EMD) and relevance vector machine (RVM). This forecast approach integrates the advantages of EMD and RVM. Firstly, by using the good filter performance of the EMD, the wind power series data are classified and combined into the intrinsic mode functions which have similar time-frequency characteristics. Secondly, RVM is applied to set up the corresponding power prediction models for all kinds of signals decomposed by EMD. Last, the sum of several RVM prediction models is achieved which is just the forecast model. In this way, the wind power short-term forecast model is achieved and its accuracy is also improved. Simulation results have also shown the high accuracy of the proposed forecast method.(4) For a nonlinear plant as variable speed wind turbine (VSWT), this dissertation proposes an adaptive control strategy for VSWT using RBF neural network. In the proposed neural network based adaptive control system, a neural network based identifier is employed to estimate nonlinear behavior of VSWT. The identifier can provide complex information about relationship between VSWT’s input and output. At the same time, another neural network is used to design an inverse model controller. This dissertation presents the implementation process and the training algorithm of the neural network for the adaptive control system. The proposed adaptive control system can overcome the influence of VSWT’s strong nonlinear. Simulation results have shown the good ability and performance of the proposed control system.(5) This dissertation carries out research on energy dispatch strategy of the hybrid renewable power system in a case study. On the basis of energy balance, the proposed energy management system describes the flowchart of the implementation of the system’s energy management and dispatch strategy. Then, for the micro grid projects in Dong’ao Island, the hybrid power system is established with system allocation scheme and the MATLAB simulation model. Simulation results show that these strategies are simple to implement, but the stability of the hybrid power system is high. Meanwhile, they improve the system’s economy and have good engineering application value.In the end, the main innovations of the dissertation are summarized, and the fields for further investigation are expected.