yinrachel / pinns_renewable_wind Goto Github PK

This is a model for predicitng the Bus voltage angles under the uncertainty of renewable wind energy. I employ the power system data from MATPOWER Case 9. I combine swing equation in the loss function of neural network.

I use matlab for simulating the wind speed and generate wind power data in a time period of 60 seconds, and add the gust amplification factor from the 20s to 30s to simulate the gust. There are 300 wind scenarios

Although there are 300 scenarios which I aim to use monte carlo simulation to present the uncertainty brought by wind. But there are only 20 scenarios for training the PINNs, becasue 1) I don't have so much for computation resources, 2) the results show the model show a good performance when there are ony 20 scenario.

I employ an unseen wind speed scenario for testing the result, the model show a good performance. But it can have a lot of room for improvement.

The predicted value is smoothed

The script outputs:

A plot comparing the predicted values to the actual data. An L2 error metric printed to the console. The MSE, MSE_data and MSE_physics are printed seprarately.