Tidal turbine array optimization in Muskeget Channel, MA using a continuous genetic algorithm

Tidal in-stream energy conversion (TISEC) farms provide a highly predictable and dependable source of energy. Economic viability depends in part on optimal planning of the layout of the turbines within the farm. The layout can be treated as an optimization problem where the best design is the one which generates maximum power. This can be achieved by minimizing negative interactions between the turbines as well as exploiting potential positive interactions. This thesis applies a global optimization method based on the continuous genetic algorithm (CGA) to two distinct applications in TISEC farm layout design. In the first application, power extraction is maximized in an idealized channel of constant depth. Flow through the channel is represented by steady solutions of the shallow water equations computed with the Finite Volume Community Ocean Model (FVCOM). Momentum extraction by the turbines is modeled in FVCOM using Linear Momentum Actuator Disk Theory (LMADT). The cost function for a given design is represented by the average of the flood and ebb power generation. Two distinct optimization problems are considered for the idealized channel. In the first case, the design space is constrained to layouts on a regular grid in separate configurations : staggered and non-staggered. For this case, the design space is spanned by two variables: the lateral and longitudinal turbine spacing. In the second case, the design space is unconstrained and allows arbitrary layouts. In this approach, the dimension of the design space increases to 2N where N is the number of turbines. Numerical experiments found that at low blockage ratios where the total cross-sectional area of the array is on the order of 20% of cross-sectional area of the channel, the unconstrained design provides only 0.3% improvement over the optimal gridded design. At higher blockage ratios approaching 100%, positive interactions between the turbines were found to provide 3% benefit over both staggered and non-staggered regular arrays. This implies that, at reduced blockage ratio, turbine interactions are difficult to exploit and constrained designs are advisable due to their significantly lower computing requirements and potentially lower installation cost. In the second application of CGA, the method was applied to the preliminary design of a pilot scale (5 MW) TISEC array in Muskeget Channel, MA. For this case, interactions between the turbines were not considered. Turbines were constrained to lie on the vertices of a fixed grid overlying the entire channel. The flow field was represented by a spatial distribution of vertically-averaged annual mean tidal power density. This database was derived from the Massachusetts Tidal Model (MTM). The cost function for a given design consisted of total power density interpolated from the database at the turbine locations. For the optimal design, the annual mean energy production (AEP) of the array was estimated following guidelines established by the National Renewable Energy Lab (NREL). The approaches considered within this work are readily applicable to the optimization of tidal power at other sites and are capable of including additional constraints such as exclusions due to navigation or essential habitat.