Wind Speed Forecasting Using ARMA and Neural Network Models

Uzair Khaleeq uz Zaman, Hamid Teimourzadeh, Elias Hassani Sangani, Xiaodong Liang, C. Y. Chung

202117 citationsDOI

Abstract

With the advancement of wind power generation technology, wind power plays an increasing role in modern power grids. To properly consider wind power for power systems planning and operation purpose, wind power and wind speed must be forecasted accurately. Wind is chaotic, random, irregular, and non-stationary in nature, which creates significant challenges in wind speed forecasting. This paper aims to forecast wind speed using both the statistical time series analysis method (autoregressive moving average (ARMA)) and neural network methods (feedforward neural network (FNN), recurrent neural network (RNN), long short-term memory (LSTM), and the gated recurrent unit (GRU)). The performance of the proposed five models is compared with the measured wind speed data, and the GRU model shows the best performance with the highest prediction accuracy. The four ANN models outperform the ARMA model.

Topics & Concepts

Wind speedWind powerAutoregressive–moving-average modelArtificial neural networkRecurrent neural networkComputer scienceAutoregressive modelTime seriesFeedforward neural networkWind power forecastingChaoticElectric power systemData modelingFeed forwardPower (physics)Moving averageArtificial intelligenceMachine learningEngineeringMeteorologyControl engineeringStatisticsMathematicsGeographyPhysicsQuantum mechanicsComputer visionElectrical engineeringDatabaseEnergy Load and Power ForecastingElectric Power System OptimizationComputational Physics and Python Applications