Regression ANN getting high root mean squared error values after re scaling dataset

Question

I have a data set on predicting solar power generation, I am getting root mean squared loos of 0.3196 on training set on scaled values, but when I inverse transform them my loss rises to 298 on training and 488 on test set. but my r2scores are .883 and .69 on tests and training sets. R2 scores seems acceptable but why i am getting large value of error.

Here is my code:

    data_path = r'drive/My Drive/Proj/S.P.F./solarpowergeneration.csv'
    dts = pd.read_csv('solarpowergeneration.csv')
    dts.head()
    X = dts.iloc[:, :-1].values
    y = dts.iloc[:, -1].values
    print(X.shape, y.shape)
    y = np.reshape(y, (-1,1))
    y.shape
    from sklearn.model_selection import train_test_split
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=42)
   
   from sklearn.preprocessing import StandardScaler
   sc_X = StandardScaler()
   X_train = sc_X.fit_transform(X_train)
   X_test = sc_X.transform(X_test)

   # outcome scaling:
   sc_y = StandardScaler()
   y_train = sc_y.fit_transform(y_train)    
   y_test = sc_y.transform(y_test)


def get_spfnet():
  ann = tf.keras.models.Sequential()

  ann.add(Dense(X_train.shape[1], activation='relu'))
  ann.add(BatchNormalization())
  ann.add(Dense(32, activation='relu', kernel_regularizer=regularizers.l2(0.01)))
  ann.add(BatchNormalization())
  ann.add(Dense(64, activation='relu', kernel_regularizer=regularizers.l2(0.01)))
  ann.add(BatchNormalization())

  ann.add(Dense(1))
  ann.compile(loss='mse',
              optimizer='adam',
              metrics=[tf.keras.metrics.RootMeanSquaredError()])
  return ann

spfnet = get_spfnet()
#spfnet.summary()
hist = spfnet.fit(X_train, y_train, batch_size=32, validation_data=(X_test, y_test),epochs=250, verbose=2)

the accuracy and loss graphs are

plt.plot(hist.history['root_mean_squared_error'])
plt.plot(hist.history['val_root_mean_squared_error'])
plt.title('Model error')
plt.xlabel('Epochs')
plt.ylabel('error')
plt.show()

what changes are needed to improve my model

dataset:

colab link of code