Analysis of a utility-interactive wind-photovoltaic hybrid system with battery storage using neural network.

摘要

This dissertation investigates the application of neural network theory to the analysis of a 4-kW Utility-interactive Wind-Photovoltaic System (WPS) with battery storage.; The hybrid system comprises a 2.5-kW photovoltaic generator and a 1.5-kW wind turbine. The wind power generator produces power at variable speed and variable frequency (VSVF). The wind energy is converted into dc power by a controlled, tree-phase, full-wave, bridge rectifier. The PV power is maximized by a Maximum Power Point Tracker (MPPT), a dc-to-dc chopper, switching at a frequency of 45 kHz. The whole dc power of both subsystems is stored in the battery bank or conditioned by a single-phase self-commutated inverter to be sold to the utility at a predetermined amount.; First, the PV is modeled using Artificial Neural Network (ANN). To reduce model uncertainty, the open-circuit voltage V_OC and the short-circuit current I_SC of the PV are chosen as model input variables of the ANN. These input variables have the advantage of incorporating the effects of the quantifiable and non-quantifiable environmental variants affecting the PV power.; Then, a simplified way to predict accurately the dynamic responses of the grid-linked WPS to gusty winds using a Recurrent Neural Network (RNN) is investigated. The RNN is a single-output feedforward backpropagation network with external feedback, which allows past responses to be fed back to the network input.; In the third step, a Radial Basis Functions (RBF) Network is used to analyze the effects of clouds on the Utility-Interactive WPS. Using the irradiance as input signal, the network models the effects of random cloud movement on the output current, the output voltage, the output power of the PV system, as well as the electrical output variables of the grid-linked inverter.; Fourthly, using RNN, the combined effects of a random cloud and a wind gusts on the system are analyzed. For short period intervals, the wind speed and the solar radiation are considered as the sole sources of power, whose variations influence the system variables. Since both subsystems have different dynamics, their respective responses are expected to impact differently the whole system behavior. The dispatchability of the battery-supported system as well as its stability and reliability during gusts and/or cloud passage is also discussed.; In the fifth step, the goal is to determine to what extent the overall power quality of the grid would be affected by a proliferation of Utility-interactive hybrid system and whether recourse to bulky or individual filtering and voltage controller is necessary.; The final stage of the research includes a steady-state analysis of two-year operation (May 96–Apr 98) of the system, with a discussion on system reliability, on any loss of supply probability, and on the effects of the randomness in the wind and solar radiation upon the system design optimization.