ARTIFICIAL INTELLIGENCE FOR THE NEXT-GENERATION SUSTAINABLE AND CARBON-NEUTRAL WASTEWATER TREATMENT SYSTEMS

THE INCREASINGLY STRINGENT WATER QUALITY STANDARDS, COUPLED WITH THE SCARCITY OF FRESHWATER RESOURCES, HAVE ACCELERATED THE NEED FOR ADVANCED WASTEWATER TREATMENT TECHNOLOGIES TO ENSURE COMPLIANCE WITH DISCHARGE LIMITS AND PROMOTE SAFE WATER REUSE. THE DIRECTIVE (EU) 2024/3019 – DIRECTLY APPLICABLE TO MEMBER STATES AND ATTRACTING GLOBAL ATTENTION FOR ITS AMBITIOUS SUSTAINABILITY AND CARBON NEUTRALITY GOALS – INTRODUCES NEW REQUIREMENTS FOR IMPROVED REMOVAL OF BIODEGRADABLE ORGANIC MATTER AND NUTRIENTS (NITROGEN AND PHOSPHORUS), EFFECTIVE REMOVAL OF MICROPOLLUTANTS AND CONTAMINANTS OF EMERGING CONCERN (CECS) THROUGH “QUATERNARY TREATMENT”, ENHANCED ENERGY EFFICIENCY, CARBON AND CLIMATE NEUTRALITY, IMPROVED PUBLIC HEALTH SURVEILLANCE, AND THE PROMOTION OF CIRCULAR ECONOMY PRACTICES. ACHIEVING THESE MULTIFACETED OBJECTIVES REQUIRES HIGH-EFFICIENCY TREATMENT PROCESSES, OPTIMIZED DESIGN AND OPERATION, AND ROBUST MONITORING AND CONTROL SYSTEMS. EMERGING DIGITAL TECHNOLOGIES – INCLUDING THE INTERNET OF THINGS (IOT), BIG DATA ANALYTICS, CLOUD COMPUTING, ARTIFICIAL INTELLIGENCE (AI), BLOCKCHAIN, ROBOTICS, VIRTUAL/AUGMENTED REALITY, AND DIGITAL TWINS – OFFER TRANSFORMATIVE OPPORTUNITIES TO DEVELOP SMART WASTEWATER TREATMENT PLANTS (WWTPS). AMONG THESE, AI STANDS OUT FOR ITS ABILITY TO MODEL NONLINEAR AND DYNAMIC RELATIONSHIPS GOVERNING TREATMENT PERFORMANCE, OVERCOMING THE LIMITATIONS OF CONVENTIONAL MATHEMATICAL AND MECHANISTIC MODELS. BY LEARNING DIRECTLY FROM HISTORICAL AND REAL-TIME DATA, AI MODELS CAN PROVIDE ACCURATE PREDICTIONS WITHOUT RELYING ON COMPLEX EQUATIONS DESCRIBING PHYSICOCHEMICAL PROCESSES, ENABLING NEXT-GENERATION WWTPS WITH SMART MONITORING, OPTIMIZATION, AND ADAPTIVE CONTROL CAPABILITIES. HOWEVER, KEY BARRIERS STILL HINDER LARGE-SCALE IMPLEMENTATION, INCLUDING INSUFFICIENT AND LOW-QUALITY DATA, POOR INTERPRETABILITY OF “BLACK-BOX” MODELS, AND LIMITED VALIDATION OF AI-DRIVEN TOOLS FOR PROCESS OPTIMIZATION. THIS PH.D. THESIS CONTRIBUTES TO ADDRESSING THESE CHALLENGES BY: (I) IMPROVING DATA QUALITY THROUGH ROBUST PREPROCESSING AND CLEANING; (II) MITIGATING DATA SCARCITY VIA CONTROLLED SYNTHETIC DATA AUGMENTATION; (III) ENHANCING MODEL INTERPRETABILITY THROUGH EXPLAINABLE AI (XAI) TECHNIQUES, SPECIFICALLY SHAPLEY ADDITIVE EXPLANATIONS (SHAP); AND (IV) DEVELOPING AND VALIDATING AI MODELS FOR PROCESS MONITORING AND OPTIMIZATION OF INNOVATIVE WASTEWATER TREATMENT TECHNOLOGIES. THE PROPOSED METHODOLOGIES WERE APPLIED TO MODEL THE PERFORMANCE OF FOUR ADVANCED SYSTEMS: A) THE LIVING MEMBRANE BIOREACTOR (LMBR) FOR ADVANCED WASTEWATER TREATMENT; B) THE TEMPERATURE SWING SOLVENT EXTRACTION (TSSE) PROCESS FOR DESALINATION OF HYPERSALINE BRINES; C) AN INNOVATIVE CARBON CAPTURE AND UTILIZATION (CCU) BIOTECHNOLOGY INTEGRATING A MOVING BED BIOFILM REACTOR AND AN ALGAL PHOTOBIOREACTOR (MBBR+APBR) WITHIN WASTEWATER TREATMENT; AND D) NANOFILTRATION FOR VOLATILE FATTY ACID (VFA) RECOVERY FROM WASTEWATER. RESULTS HIGHLIGHT THE STRONG POTENTIAL AND VERSATILITY OF AI-DRIVEN APPROACHES FOR MODELING, ANALYZING, AND INTERPRETING COMPLEX TREATMENT PROCESSES. OVERALL, THIS RESEARCH DEMONSTRATES THAT AI CAN EFFECTIVELY CAPTURE COMPLEX NONLINEAR DYNAMICS IN ENVIRONMENTAL PROCESSES, SUPPORTING ACCURATE MONITORING, PREDICTION, AND OPTIMIZATION. THE INTEGRATION OF XAI, HYPERPARAMETER OPTIMIZATION, ROBUST DATA PREPROCESSING, AND CONTROLLED SYNTHETIC DATA AUGMENTATION PROVIDES A METHODOLOGICAL FRAMEWORK TO OVERCOME MAJOR BARRIERS TO AI ADOPTION IN WASTEWATER TREATMENT, PAVING THE WAY FOR SMART, EFFICIENT, AND ADAPTIVE CONTROL STRATEGIES. THIS PH.D. THESIS PROVIDES A COMPREHENSIVE SCIENTIFIC AND MULTIDISCIPLINARY CONTRIBUTION TO THE DEVELOPMENT OF NEXT-GENERATION SUSTAINABLE AND CARBON-NEUTRAL WWTPS THROUGH THE INTEGRATION OF AI-DRIVEN MONITORING, OPTIMIZATION, AND CONTROL SYSTEMS.