import streamlit as st import pandas as pd import numpy as np import plotly.graph_objs as go import requests from io import StringIO import base64 #@st.cache_data(ttl=86400) # TTL is set for 86400 seconds (24 hours) def load_data_predictions(github_token): url = 'https://api.github.com/repos/mmmapms/Forecast_DAM_V2/contents/Predictions.csv' headers = {'Authorization': f'token {github_token}'} response = requests.get(url, headers=headers) if response.status_code == 200: file_content = response.json()['content'] decoded_content = base64.b64decode(file_content).decode('utf-8') csv_content = StringIO(decoded_content) df = pd.read_csv(csv_content, encoding='utf-8') df = df.rename(columns={ 'Price': 'Real Price', 'DNN1': 'Neural Network 1', 'DNN2': 'Neural Network 2', 'DNN3': 'Neural Network 3', 'DNN4': 'Neural Network 4', 'LEAR56': 'Regularized Linear Model 1', 'LEAR84': 'Regularized Linear Model 2', 'LEAR112': 'Regularized Linear Model 3', 'LEAR730': 'Regularized Linear Model 4', 'Persis': 'Persistence Model', 'Hybrid_Ensemble': 'Hybrid Ensemble', 'Weighted_Ensemble': 'Weighted Ensemble' }) df['Date'] = pd.to_datetime(df['Date'], dayfirst=True) df_filtered = df.dropna(subset=['Real Price']) return df, df_filtered else: st.error("Failed to download data. Please check your GitHub token and repository details.") return pd.DataFrame(), pd.DataFrame() github_token = st.secrets["GitHub_Token_Margarida"] if github_token: df, df_filtered = load_data_predictions(github_token) # Your existing logic to use df and df_filtered else: st.warning("Please enter your GitHub Personal Access Token to proceed.") #@st.cache_data #def load_data_predictions(): # df = pd.read_csv('Predictions.csv') # df = df.rename(columns={ # 'Price': 'Real Price', # 'DNN1': 'Neural Network 1', # 'DNN2': 'Neural Network 2', # 'DNN3': 'Neural Network 3', # 'DNN4': 'Neural Network 4', # 'DNN_Ensemble': 'Neural Network Ensemble', # 'LEAR56': 'Regularized Linear Model 1', # 'LEAR84': 'Regularized Linear Model 2', # 'LEAR112': 'Regularized Linear Model 3', # 'LEAR730': 'Regularized Linear Model 4', # 'LEAR_Ensemble': 'Regularized Linear Model Ensemble', # 'Persis': 'Persistence Model', # 'Hybrid_Ensemble': 'Hybrid Ensemble' #}) # df['Date'] = pd.to_datetime(df['Date'], dayfirst=True) # df_filtered = df.dropna(subset=['Real Price']) # return df, df_filtered #df, df_filtered = load_data_predictions() min_date_allowed_pred = df_filtered['Date'].min().date() max_date_allowed_pred = df_filtered['Date'].max().date() end_date = df['Date'].max().date() start_date = end_date - pd.Timedelta(days=7) models_corr_matrix = ['Real Price', 'Persistence Model', 'Neural Network 1', 'Neural Network 2', 'Neural Network 3', 'Neural Network 4', 'Regularized Linear Model 1', 'Regularized Linear Model 2', 'Regularized Linear Model 3', 'Regularized Linear Model 4', 'Weighted Ensemble'] def conformal_predictions(data): data['Residuals'] = data['Weighted Ensemble'] - data['Real Price'] data.set_index('Date', inplace=True) data['Hour'] = data.index.hour min_date = data.index.min() for date in data.index.normalize().unique(): if date >= min_date + pd.DateOffset(days=30): start_date = date - pd.DateOffset(days=30) end_date = date calculation_window = data[start_date:end_date-pd.DateOffset(hours=1)] quantiles = calculation_window.groupby('Hour')['Residuals'].quantile(0.9) # Use .loc to safely access and modify data if date in data.index: current_day_data = data.loc[date.strftime('%Y-%m-%d')] for hour in current_day_data['Hour'].unique(): if hour in quantiles.index: hour_quantile = quantiles[hour] idx = (data.index.normalize() == date) & (data.Hour == hour) data.loc[idx, 'Quantile_90'] = hour_quantile data.loc[idx, 'Lower_Interval'] = data.loc[idx, 'Weighted Ensemble'] - hour_quantile data.loc[idx, 'Upper_Interval'] = data.loc[idx, 'Weighted Ensemble'] + hour_quantile data.reset_index(inplace=True) return data # Main layout of the app col1, col2 = st.columns([5, 2]) # Adjust the ratio to better fit your layout needs with col1: st.title("Belgium: Electricity Price Forecasting") with col2: upper_space = col2.empty() upper_space = col2.empty() col2_1, col2_2 = st.columns(2) # Create two columns within the right column for side-by-side images with col2_1: st.image("KU_Leuven_logo.png", width=100) # Adjust the path and width as needed with col2_2: st.image("energyville_logo.png", width=100) upper_space.markdown("""     """, unsafe_allow_html=True) # Sidebar for inputs with st.sidebar: st.write("### Variables Selection for Graph") st.write("Select which variables you'd like to include in the graph. This will affect the displayed charts and available data for download.") selected_variables = st.multiselect("Select variables to display:", options=['Real Price', 'Neural Network 1', 'Neural Network 2', 'Neural Network 3', 'Neural Network 4', 'Regularized Linear Model 1', 'Regularized Linear Model 2','Regularized Linear Model 3', 'Regularized Linear Model 4', 'Weighted Ensemble', 'Persistence Model'], default=['Real Price', 'Weighted Ensemble']) st.write("### Model Selection for Scatter Plot") model_selection = st.selectbox("Select which model's predictions to display:", options=['Neural Network 1', 'Neural Network 2', 'Neural Network 3', 'Neural Network 4', 'Regularized Linear Model 1', 'Regularized Linear Model 2','Regularized Linear Model 3', 'Regularized Linear Model 4', 'Weighted Ensemble', 'Persistence Model'], index=8) # Adjust the index as needed to default to your desired option st.write("### Date Range for Metrics Calculation") st.write("Select the date range to calculate the metrics for the predictions. This will influence the accuracy metrics displayed below. The complete dataset ranges from 10/03/2024 until today.") start_date_pred, end_date_pred = st.date_input("Select Date Range for Metrics Calculation:", [min_date_allowed_pred, max_date_allowed_pred]) # Main content if not selected_variables: st.warning("Please select at least one variable to display.") else: st.write("## Belgian Day-Ahead Electricity Prices") # Call conformal_predictions if 'Hybrid Ensemble' is selected if 'Weighted Ensemble' in selected_variables: df = conformal_predictions(df) # Make sure this function modifies df correctly temp_df = df[(df['Date'] >= pd.Timestamp(start_date))] # Ensure correct date filtering # Initialize Plotly figure fig = go.Figure() for variable in selected_variables: fig.add_trace(go.Scatter(x=temp_df['Date'], y=temp_df[variable], mode='lines', name=variable)) # Check if conformal predictions should be added for Hybrid Ensemble if variable == 'Weighted Ensemble' and 'Quantile_90' in df.columns: # Add the lower interval trace fig.add_trace(go.Scatter( x=temp_df['Date'], y=temp_df['Lower_Interval'], mode='lines', line=dict(width=0), showlegend=False )) # Add the upper interval trace and fill to the lower interval fig.add_trace(go.Scatter( x=temp_df['Date'], y=temp_df['Upper_Interval'], mode='lines', line=dict(width=0), fill='tonexty', # Fill between this trace and the previous one fillcolor='rgba(68, 68, 68, 0.3)', name='P10/P90 prediction intervals' )) fig.update_layout(xaxis_title="Date", yaxis_title="Price [EUR/MWh]") st.plotly_chart(fig, use_container_width=True) st.write("The graph presented here illustrates the day-ahead electricity price forecasts for Belgium, covering the period from one week ago up to tomorrow. The forecasts are made every morning on day D at 08.00 for day D+1.") if not selected_variables: st.warning("Please select at least one variable to display.") else: # Plotting st.write("## Scatter Plot: Real Price vs Model Predictions") # Filter based on the selected date range for plotting plot_df = df[(df['Date'] >= pd.Timestamp(min_date_allowed_pred)) & (df['Date'] <= pd.Timestamp(max_date_allowed_pred))] model_column = model_selection # Create the scatter plot fig = go.Figure() fig.add_trace(go.Scatter(x=plot_df['Real Price'], y=plot_df[model_column], mode='markers', name=f"Real Price vs {model_selection} Predictions")) m, b = np.polyfit(plot_df['Real Price'], plot_df[model_column], 1) regression_line = m * plot_df['Real Price'] + b # Add the line of best fit to the figure with the equation as the legend name fig.add_trace(go.Scatter(x=plot_df['Real Price'], y=regression_line, mode='lines', line=dict(color='black'), showlegend=False)) # Update layout with appropriate titles fig.update_layout( title=f"Scatter Plot of Real Price vs {model_selection} Predictions from {min_date_allowed_pred} to {max_date_allowed_pred}", xaxis_title="Real Price [EUR/MWh]", yaxis_title=f"{model_selection} Predictions [EUR/MWh]", xaxis=dict(range=[-160, 160]), # Setting the x-axis range yaxis=dict(range=[-150, 150]), # Setting the y-axis range showlegend=False ) st.plotly_chart(fig, use_container_width=True) # Calculating and displaying metrics if start_date_pred and end_date_pred: st.header("Accuracy Metrics") #st.write(f"The accuracy metrics are calculated from {start_date_pred} to {end_date_pred}, this intervale can be changed in the sidebar.") st.write(f"The accuracy metrics are calculated from **{start_date_pred}** to **{end_date_pred}**. This interval can be changed in the sidebar.") filtered_df = df_filtered[(df_filtered['Date'] >= pd.Timestamp(start_date_pred)) & (df_filtered['Date'] < (pd.Timestamp(end_date_pred)+ pd.Timedelta(days=1)))] # List of models for convenience models = [ 'Neural Network 1', 'Neural Network 2', 'Neural Network 3', 'Neural Network 4', 'Regularized Linear Model 1', 'Regularized Linear Model 2', 'Regularized Linear Model 3', 'Regularized Linear Model 4', 'Persistence Model', 'Weighted Ensemble' ] # Placeholder for results results = {'Metric': ['MAE', 'rMAE']} #'sMAPE', 'RMSE', p_real = filtered_df['Real Price'] # Iterate through each model to calculate and store metrics for model in models: # Assuming column names in filtered_df match the model names directly for simplicity p_pred = filtered_df[model] mae = np.mean(np.abs(p_real - p_pred)) #smape = 100 * np.mean(np.abs(p_real - p_pred) / ((np.abs(p_real) + np.abs(p_pred)) / 2)) #rmse = np.sqrt(np.mean((p_real - p_pred) ** 2)) rmae = mae/np.mean(np.abs(p_real - filtered_df['Persistence Model'])) # Store the results results[model] = [f"{mae:.2f}", f"{rmae:.2f}"] #f"{smape:.2f}%", f"{rmse:.2f}", # Convert the results to a DataFrame for display metrics_df = pd.DataFrame(results) transposed_metrics_df = metrics_df.set_index('Metric').T col1, col2 = st.columns([3, 2]) # Display the transposed DataFrame with col1: # Assuming 'transposed_metrics_df' is your final DataFrame with metrics st.dataframe(transposed_metrics_df, hide_index=False) with col2: st.markdown("""
Equations
""", unsafe_allow_html=True) # Rendering LaTeX equations st.markdown(r""" $\text{MAE} = \frac{1}{n}\sum_{i=1}^{n}|y_i - \hat{y}_i|$ $\text{rMAE} = \frac{\text{MAE}}{MAE_{\text{Persistence Model}}}$ """) st.write("## Correlation Matrix") models_df = df_filtered[models_corr_matrix] corr_matrix = models_df.corr() fig = go.Figure(data=go.Heatmap( z=corr_matrix.values, x=corr_matrix.columns, y=corr_matrix.index)) fig.update_layout( yaxis_autorange='reversed' # Ensure the y-axis starts from the top ) st.plotly_chart(fig, use_container_width=True) st.write("## Access Predictions") st.write("If you are interested in accessing the predictions made by the models, the models themselves or how these can be incorporated into your workflows, please contact Margarida Mascarenhas (PhD Student, KU Leuven) at margarida.mascarenhas@kuleuven.be or Hussain Kazmi (assistant professor, KU Leuven) at hussain.kazmi@kuleuven.be.")