What exactly is 'bias' in a Perceptron model?

We know that Prediction is based on Activation function in a Perceptron model. \({\hat y=f(z)}\) \(z=\sum_{i=1}^{n}w_ix_i+b\) \(\hat y\) = Predicted output \({f(z)}\) = Activation function \({w}\) = W