# app/api_utils.py import tensorflow as tf from tensorflow.keras import losses, metrics import sys import os from src.data_preprocessing import SolarDataProcessor from src.visualization import SolarCellVisualizer import tempfile import csv import matplotlib.pyplot as plt import io import base64 import numpy as np def comp_iv_curve(filename): path_data = tempfile.gettempdir() path_model = os.path.join(os.getcwd(), "models", "final_model_mlp.h5") path_input = os.path.join(os.getcwd(), "models") visualizer = SolarCellVisualizer() # Chargement modèle model = tf.keras.models.load_model(path_model, custom_objects={'mse': losses.MeanSquaredError(), 'mae': metrics.MeanAbsoluteError()}) processor = SolarDataProcessor(path_input, path_data) X_scaled, I_list, V_list = processor.preprocess_input([filename], path_data) params, I_model = visualizer.plot_comparaison_courbes(processor, model, X_scaled, I_list, V_list, Plot = False) return params, I_model def comp_iv_curve_cnn1d(filename): path_data = tempfile.gettempdir() path_model = os.path.join(os.getcwd(), "models", "final_model_cnn1d.h5") path_input = os.path.join(os.getcwd(), "models") visualizer = SolarCellVisualizer() # Chargement modèle CNN1D model = tf.keras.models.load_model( path_model, custom_objects={'mse': losses.MeanSquaredError(), 'mae': metrics.MeanAbsoluteError()} ) processor = SolarDataProcessor(path_input, path_data) X_scaled, I_list, V_list = processor.preprocess_input([filename], path_data) # Reshape pour CNN1D X_scaled = X_scaled[..., np.newaxis] params, I_model = visualizer.plot_comparaison_courbes(processor, model, X_scaled, I_list, V_list, Plot=False) return params, I_model def read_iv_csv(filepath): points = [] with open(filepath, newline='') as csvfile: reader = csv.DictReader(csvfile) for row in reader: try: V = float(row['V']) I = float(row['I']) points.append((V, I)) except ValueError: continue # Ignorer les lignes invalides return points def generate_curve_plot(points, model_points): fig, ax = plt.subplots() # Extraction des points V_input, I_input = zip(*points) V_model, I_model = zip(*model_points) ax.scatter(V_input, I_input, color='blue', label='Données expérimentales', alpha=0.6) ax.plot(V_model, I_model, color='red', label='Courbe modélisée', linewidth=2) ax.set_xlabel('Tension (V)') ax.set_ylabel('Courant (I)') ax.set_title('Courbe I-V : Expérimental vs Modèle') ax.legend() ax.grid(True) # Sauvegarde de la figure dans un buffer mémoire buf = io.BytesIO() plt.savefig(buf, format='png', bbox_inches='tight') buf.seek(0) # Conversion en base64 img_base64 = base64.b64encode(buf.read()).decode('utf-8') buf.close() plt.close(fig) return img_base64 def generate_multi_curve_plot(exp_points, model_dict): """ Génère un graphique avec les points expérimentaux et plusieurs courbes modèles. Parameters: exp_points (list of tuple): Liste de tuples (V, I) expérimentaux. model_dict (dict): Dictionnaire de courbes à tracer. Returns: str: Image encodée en base64 (format PNG). """ fig, ax = plt.subplots(figsize=(8, 5)) # Points expérimentaux V_exp, I_exp = zip(*exp_points) ax.scatter(V_exp, I_exp, color='royalblue', label='Données expérimentales', alpha=0.5, s=25) # Styles personnalisés color_map = { 'MLP': 'crimson', 'GENETIQUE': 'darkgreen', 'RANDOM': 'darkorange' } line_styles = { 'MLP': '-', 'GENETIQUE': '--', 'RANDOM': ':' } for label, model_points in model_dict.items(): V_model, I_model = zip(*model_points) ax.plot(V_model, I_model, label=label, color=color_map.get(label, 'gray'), linestyle=line_styles.get(label, '-.'), linewidth=2) ax.set_xlabel('Tension (V)') ax.set_ylabel('Courant (I)') ax.set_title('Courbe I-V : Comparaison des méthodes') ax.grid(True, linestyle='--', alpha=0.3) ax.legend(loc='upper left', fontsize='small') # Export en base64 buf = io.BytesIO() plt.tight_layout() plt.savefig(buf, format='png', bbox_inches='tight') buf.seek(0) img_base64 = base64.b64encode(buf.read()).decode('utf-8') buf.close() plt.close(fig) return img_base64