REQUEST FOR PROPOSAL (RFP) AI-Powered Thermal Energy Storage (TES) Optimization & Predictive Analytics Integrated with Johnson Controls BMS Platform 1. Introduction [Client Name] invites qualified vendors to submit proposals for the design, implementation, and commissioning of an AI-powered optimization and predictive analytics platform for a Thermal Energy Storage (TES) system integrated with an existing Johnson Controls Building Management System (BMS). The objective is to enhance energy efficiency, reduce demand charges, enable dynamic charge/discharge optimization, and introduce predictive condition monitoring. 2. Project Objectives The proposed solution shall: Perform 48-hour cooling load forecasting Optimize TES charge/discharge scheduling using: Real-time chiller efficiency (kW/TR) Utility tariff structure (ToU + demand charges) Provide marginal cost of cooling analytics Monitor TES degradation (thermal losses) Integrate with Johnson Controls BMS via: BACnet/IP Modbus TCP/IP Provide: Advisory control (minimum requirement) Supervisory control (optional Phase 1 scope) 3. Existing Infrastructure BMS Platform: Johnson Controls Available Integration Protocols: BACnet/IP Modbus TCP/IP TES instrumentation: Tank temperature stratification sensors Flow meters Tank level sensors Chiller plant efficiency metrics (kW/TR) Utility meter interface 4. Scope of Work 4.1 Data Integration Layer Vendor shall: Integrate with Johnson Controls BMS via BACnet/IP and/or Modbus TCP/IP Extract real-time and historical data Implement secure gateway architecture Design tag mapping and data model Deliverables: Points list Integration architecture diagram Communication validation report 4.2 AI & Analytics Layer A. Load Forecasting 48-hour rolling cooling load forecast Weather-driven modeling Forecast accuracy reporting (MAPE) B. TES Optimization Engine Dynamic marginal cost calculation Charge/discharge schedule optimization Demand charge reduction logic Rolling horizon optimization (hourly recalculation) C. Condition Monitoring Thermal loss trend analysis Stratification health monitoring Degradation detection alerts 4.3 Control Mode Requirements Minimum Requirement – Advisory Mode AI generates recommended: Charge/discharge schedule Setpoints Mode change timing Visualization dashboard Operator approval workflow Optional Phase 1 – Supervisory Control AI writes: Mode request signals Charge/discharge setpoints Johnson BMS executes final control logic PLC/BMS safety interlocks remain authoritative Fail-safe fallback to standard ToU schedule Direct control of field devices by AI is NOT permitted. 5. Cybersecurity & Network Segmentation Requirements The solution must comply with OT cybersecurity best practices. 5.1 Network Segmentation AI platform shall reside in: Dedicated OT DMZ Or segregated application server network No direct access from internet to BMS network Strict firewall rule enforcement Role-based access control (RBAC) 5.2 Data Flow Architecture Allowed pattern: BMS → Gateway → AI Engine → BMS (supervisory writeback) Not allowed: Direct AI-to-field device communication 5.3 Compliance Vendor shall comply with: ISA/IEC 62443 principles Corporate IT cybersecurity policies Encrypted communication (TLS where applicable) 6. Deployment Model Options Vendor shall propose costed options for: Option A – On-Premises Deployment AI engine hosted on local server Located in data center or OT DMZ No cloud dependency Advantages: Strongest cybersecurity posture No internet dependency Preferred for critical infrastructure Challenges: Higher CAPEX Hardware procurement Maintenance responsibility on-site Option B – Cloud Deployment AI hosted on secure cloud platform BMS data forwarded via secure gateway Weather and tariff APIs integrated natively Advantages: Lower upfront infrastructure cost Faster deployment Scalable Easier ML model updates Challenges: Requires outbound OT connectivity Higher cybersecurity governance Ongoing subscription OPEX Option C – Hybrid Model (Recommended) Data historian + integration on-prem AI processing in cloud Secure outbound-only data push Supervisory writeback via secure tunnel Advantages: Balanced cybersecurity Lower CAPEX than full on-prem Advanced ML capability Operational flexibility 7. Cost Effectiveness Analysis (General Guidance) Model CAPEX OPEX Cyber Complexity Scalability Recommendation On-Prem High Low Low Medium Best for high-security sites Cloud Low Medium Medium/High High Best for fast deployment Hybrid Medium Medium Controlled High Most balanced Most Cost-Effective in Early Phase: Hybrid model typically delivers the best ROI for Phase 1 because: Avoids heavy hardware investment Enables rapid AI iteration Maintains OT security boundary 8. Performance Requirements Vendor shall specify: Forecast accuracy target (≤10% MAPE preferred) Optimization ROI projection Expected demand charge reduction % System latency Failover recovery time 9. Deliverables Detailed system architecture (layered model) Cybersecurity architecture diagram Integration design with Johnson Controls BMS AI model documentation Testing & validation plan Commissioning report Operator training materials ROI model 10. Vendor Qualification Requirements Vendor must demonstrate: Experience with Johnson Controls BMS integration Experience with BACnet/IP and Modbus TCP/IP Proven AI/ML optimization projects Industrial cybersecurity implementation experience References in HVAC or TES optimization projects 11. Proposal Submission Requirements Proposal shall include: Technical proposal Control mode clarification (Advisory vs Supervisory) Deployment architecture Cybersecurity design Detailed cost breakdown: Software Integration Hardware (if on-prem) Annual maintenance Implementation timeline Project governance structure Strategic Recommendation (Executive Perspective) For Phase 1: Start with Advisory Mode + Hybrid Deployment Design architecture ready for Supervisory upgrade Implement strong cybersecurity segmentation from Day 1 Validate forecast & optimization ROI before enabling supervisory control