粤公网安备44040202001840号
中文
+86-756-2137770
ToHi AI Agents — An AI Colleague for Every Role in Your Enterprise
Built on ToHi’s two decades of GRP business systems and domain expertise, our agent matrix spans fuel, procurement, equipment, inspection, safety and operations — data drives business, intelligence empowers the future.
5 Layers + 4 Cross-Cutting Systems | Data → Business → Integration → Knowledge → Semantics → Agents → Applications — business value in a closed loop
112
Business Agents
12
AI+ Business Domains
20+
Business Roles
5
Layer Enterprise Architecture
Solution Overview
WHY ToHi AI AGENTS
As AI accelerates the reinvention of enterprise operations, general-purpose LLMs struggle with complex business semantics and rule constraints. The ToHi AI Agent Solution builds on the enterprise’s own data and domain knowledge. Through a "5-layer + 4 cross-cutting" enterprise-grade agent architecture, it turns the domain models, metric definitions and business rules accumulated in GRP business systems into capabilities that AI can understand, invoke and reason over — ultimately embedding agents into every business role.
The solution has accumulated 112 business agents across 12 AI+ business domains, spanning fuel management, procurement & tendering, contract performance, materials & warehousing, equipment maintenance, operations & inspection, safety & environment, production & dispatch, and business & finance — delivering a complete closed loop from data to business value.
1Data Drives Business
Unified management of multi-source heterogeneous data — real-time, relational and document data connected — gives agents a trusted data foundation.
2Business Drives Intelligence
GRP domain services, APIs and business events are encapsulated as tools, so agents truly invoke business capabilities rather than merely answering questions.
3Intelligence Empowers the Future
Ontology-based semantic reasoning as the semantic core and orchestration as the execution core support everything from single-task automation to multi-agent collaboration.
4Security & Cost Under Control
Four cross-cutting systems — security & compliance, observability, evaluation operations and cost governance — run through the architecture, securing large-scale agent deployment.
Overall Architecture
FIVE LAYERS + FOUR CROSS-CUTTING SYSTEMS
Figure: Overall Architecture of ToHi Intelligent Agents (5 Layers + 4 Cross-Cutting Systems)
Five Vertical LayersData Source Layer → Business System Layer → AI Integration & Governance Platform → Agent Construction Layer → Agent Front-End Layer — bottom-up transformation from data to intelligent applications.
Four Cross-Cutting SystemsSecurity & compliance, observability & O&M, evaluation operations and cost governance run across the whole architecture, keeping agents stable, controllable and sustainably operated.
Dual-Core DriveThe ontology semantic-reasoning sublayer, as the semantic core, "understands the business"; the agent construction & orchestration sublayer, as the execution core, "gets tasks done".
Closed-Loop DeliveryData → Business → Integration → Knowledge → Semantics → Agents → Applications — and back to business, closing the loop of business value.
Agent Matrix
112 AGENTS ACROSS 12 AI+ BUSINESS DOMAINS
Coal Blending & Co-firing Agent
Fuel Department HeadConnects to unit load plans, coal quality and inventory of each stockpile, and emission limits; automatically performs boiler combustion simulation, linear coal-quality blending optimization, emission compliance checks and dynamic correction, and outputs blending schemes and stacker/reclaimer tasks ready for dispatch
Decision Optimization Decision Recommendation
Capability
When as-fired coal quality fluctuates, combustion destabilizes and emission indicators approach their limits, while blending still relies on experience-based trial calculations — safety and economy are hard to reconcile. This agent connects to unit load plans, coal quality and inventory of each stockpile, and emission limits; it automatically performs boiler combustion simulation, linear blending optimization, emission compliance checks and dynamic correction, and outputs blending schemes and stacking/reclaiming tasks ready for dispatch to stacker/reclaimers. When coal quality fluctuates or blending exceeds limits, it recalculates automatically and warns of combustion instability risk. It does not modify existing DCS control logic — it only issues schemes, so deployment is quick.
Intelligent Fuel Procurement Agent
Fuel Department HeadReads generation plans, inventory, coal price trends, transport capacity and supplier performance along long-term-contract and spot-market lines; automatically performs five analyses — price trend, transport capacity, inventory safety, supplier performance and cost estimation — and directly recommends purchase volumes and staggered delivery schedules
Review & Compliance Analysis & Insight
Capability
Mismatched purchase volumes versus generation demand, coal prices that spike just before procurement, and under-fulfilled long-term contracts leave fuel managers patching holes every month. This agent reads generation plans, inventory, coal price trends, transport capacity and supplier performance data along both long-term-contract and spot-market lines; it automatically performs five analyses — coal price trends, transport capacity assessment, inventory safety, supplier performance and cost estimation — and directly recommends purchase volumes and staggered delivery schedules. It flags volume mismatches, price anomalies, transport gaps, qualification downgrades, contract under-fulfillment and delivery delays, each traceable to the specific batch and supplier.
Fuel Inventory Strategy Agent
Fuel Department HeadDistinguishes winter supply-guarantee from routine operation scenarios; reads daily consumption, incoming coal plans, stockpile temperature and yard capacity data; automatically performs consumption forecasting, inventory turnover analysis, spontaneous-combustion risk analysis and capacity utilization analysis, and outputs recommended upper/lower inventory limits and stockpile relocation plans
Prediction & Early Warning Analysis & Insight
Capability
Running out of coal during winter supply guarantee is an incident, while overstocking ties up capital and aged piles risk spontaneous combustion — inventory levels are often set by gut feel. This agent distinguishes winter supply-guarantee from routine operation scenarios; it reads daily consumption, incoming coal plans, stockpile temperature and yard capacity data; automatically performs consumption forecasting, inventory turnover analysis, spontaneous-combustion risk analysis and capacity utilization analysis; and outputs recommended inventory upper/lower limits and stockpile relocation plans, stating how much more coal the yard can hold and how many additional generation hours that supports.
Fuel Risk Control Agent
Fuel Department HeadCross-checks fuel business documents, sampling/preparation/testing data and weighing records; automatically detects risk scenarios such as abnormal weighing, abnormal assay results, sampling bias, unauthorized intervention, duplicate truck plates, card-shortage beyond threshold and tampering traces, then applies semantic re-review to remove false positives
Review & Compliance Analysis & Insight
Capability
Fuel handling is a high-incidence area for integrity risk in power plants — once weighing, assaying or sampling is manipulated, after-the-fact investigation is nearly impossible. This agent cross-checks three data sources — fuel business documents, sampling/preparation/testing data and weighing records — and automatically detects risk scenarios such as abnormal weighing, abnormal assays, sampling bias, unauthorized intervention, duplicate truck plates, card shortage beyond threshold and tampering traces. Semantic re-review then removes false positives and recognizes paraphrased expressions. Finally it reconciles document errors and ledger timeline anomalies, automatically assembling a complete evidence chain ready for inspection, with one-click export of audit reports.
Inbound Coal Acceptance Agent
Fuel Department HeadFor truck, rail and barge deliveries, automatically verifies vehicle information at weighing, sampling and unloading, analyzes weighing anomalies, checks sampling compliance, optimizes receiving/unloading schedules and pre-assesses weight shortage — covering duplicate plates, tare-weight anomalies and non-compliant sampling points
Recognition & Perception Single-Task Automation
Capability
Weight shortages, tare-weight anomalies and non-compliant sampling points are discovered too late if found only at settlement, while truck queues slow down receiving. For truck, rail and barge deliveries, this agent automatically verifies vehicle information at weighing, sampling and unloading, analyzes weighing anomalies, checks sampling compliance, optimizes receiving/unloading schedules and pre-assesses weight shortage — covering duplicate plates, tare-weight anomalies, non-compliant sampling points, coal inconsistent with contract, unloading sequence conflicts and video-snapshot anomalies in one pass.
Sampling & Assay Quality Control Agent
Fuel Department HeadReads coal-sample transfer documents, sample-preparation equipment logs and assay data; automatically detects sample tampering, sample swapping, out-of-tolerance results, unauthorized retesting, coding leaks, over-time sample-jar transfers, expired retained samples, overdue equipment calibration and duplicate-sample deviations, and reconciles document errors
Review & Compliance Analysis & Insight
Capability
Sample swapping, tampering and unauthorized retesting — once sampling/preparation/assay integrity is compromised, coal quality data becomes untrustworthy and accountability is hard to establish afterward. This agent reads coal-sample transfer documents, preparation equipment logs and assay data; it automatically detects sample tampering, swapping, out-of-tolerance anomalies, unauthorized retesting, coding leaks, over-time sample-jar transfers, expired retained samples, overdue equipment calibration and duplicate-sample deviations, and reconciles document errors, logical contradictions in assay data and operational timeline anomalies, automatically tracing back to form an evidence chain. It adds only a data-verification layer without changing existing sampling workflows.
Fuel Settlement Agent
Fuel Department HeadProcesses long-term-contract and spot settlement batches; automatically performs quantity-quality matching checks, contract clause parsing, automatic deduction calculation, invoice verification and cost review — covering coal-quality deduction errors, metering-acceptance inconsistencies, wrong clause application and invoice mismatches
Review & Compliance Analysis & Insight
Capability
Miscalculated coal-quality deductions, wrongly applied contract clauses and mismatches between metering and acceptance — every such case is money given away. This agent processes long-term-contract and spot settlement batches; it automatically performs quantity-quality matching checks, contract clause parsing, automatic deduction calculation, invoice verification and cost review, fully covering coal-quality deduction errors, metering-versus-acceptance inconsistencies, wrong clause application, invoice mismatches, provisional-estimate deviations and multiple settlements on one invoice, and directly outputs a settlement variance table.
Fuel Supplier Performance Agent
Fuel Department HeadEvaluates suppliers over annual and per-batch cycles; automatically aggregates five indicator families — coal quality stability, delivery fulfillment rate, performance timeliness, incidents and cost contribution — outputs quantified scores and tiering, links directly to next-year purchase share allocation and recommends share adjustments
Analysis & Insight Analysis & Insight
Capability
Supplier quality judged by impression and next-year shares allocated by gut feel let poor suppliers win bids year after year. This agent evaluates suppliers over annual and per-batch cycles, automatically aggregating five indicator families — coal quality stability, delivery fulfillment, performance timeliness, abnormal incidents and cost contribution — and outputs quantified scores and tiering results, linked directly to next-year purchase share allocation with adjustment recommendations. It also flags persistent quality non-compliance, insufficient fulfillment, repeated delays, abnormal card shortages and frequent complaints.
AI-Based Fuel Risk Control Cockpit
Company LeadershipConnects to full-chain fuel data across planning, procurement, inbound acceptance, sampling/assay and settlement; automatically detects risks such as calorific-value jumps, sampling violations and assay outliers with graded alerts based on dynamic adaptive thresholds, auto-generates risk analysis reports, and supports multi-dimensional drill-down by plant, batch and supplier on a panoramic cockpit
Prediction & Early Warning Multi-Agent Collaboration
Capability
For groups with multiple power plants, fuel risks are scattered across links — leadership cannot see the full picture and monthly reports are assembled manually. This agent connects to full-chain data across fuel planning, procurement, inbound acceptance, sampling/assay and settlement; it automatically detects risks such as calorific-value jumps, sampling violations and assay outliers, issues graded alerts based on dynamic adaptive thresholds, and auto-generates risk analysis reports. On the panoramic cockpit, users drill down by plant, batch and supplier, forming a closed loop of alert, dispatch, handling and re-check. Reports can be submitted directly as inspection-rectification material.
Customer Cases
TYPICAL CUSTOMERS IN THE ENERGY INDUSTRY
1
Shenzhen Energy Group Co., Ltd.
Intelligent Procurement
AI + Procurement & Tendering: procurement-plan review, procurement-document compliance review, collusive-bidding detection, AI-assisted evaluation, supplier risk analysis, procurement risk early-warning
Large-model technology is embedded deep into procurement scenarios, building an intelligent analysis and decision system covering the full procurement lifecycle — upgrading procurement decisions from experience-driven to data + AI dual-driven, with end-to-end intelligence from demand forecasting and supplier evaluation to contract risk control.
2
Beijing District Heating Group
AI Equipment · Inspection · Defects
AI + Equipment Maintenance / Operations & Inspection: equipment health assessment, inspection anomaly early-warning, visual defect recognition, defect & elimination agents, hazard snap-reporting, multi-source alarm noise reduction with closed loop
Equipment condition shifted from manual reading to health scores and degradation warnings; inspection records are dictated on site and written up automatically; massive alarms are denoised and auto-converted into hazard work orders dispatched to owners — a closed loop of alarm, hazard, rectification and close-out.
3
Beijing Energy Group
AI Maintenance · Safety Control
AI + Equipment & Maintenance / Safety & Environment: maintenance knowledge graph, intelligent maintenance scheduling, work-permit safety-measure recommendation, JSA job safety analysis, AI video-based safety-behavior monitoring
Veteran experience is distilled into a queryable maintenance knowledge base, so newcomers no longer depend on personal mentoring to draft plans; maintenance scheduling balances outage duration against generation revenue; work-safety risks are intelligently controlled around the clock.
4
Ningdong Power Plant
Smart Coal Yard · Fuel Digitalization
AI + Fuel Management: inbound acceptance verification, sampling & assay quality control, fuel settlement re-check, fuel risk detection, unattended weighbill OCR recognition
Overcame the digitalization challenges of core links such as coal quality inspection and weighbridge management: OCR linked with IoT devices captures key data in seconds, infrared detection and video surveillance provide anti-fraud measures, and a traceable electronic weighbill evidence system is built.
珠海市同海科技股份有限公司
0756-2137770
tohi@tohi.cn
珠海总部：珠海市香洲区柠溪路338号太和商务中心6楼
北京办：北京市朝阳区建外SOHO西区17号楼1206
粤澳合作区：横琴天羽道48号509、510文化创意
越南办：越南海防市海阳坊生态河都市区H2双子塔楼