How SelectDB Cuts 60% Costs and Boosts Real‑Time Performance for New Energy Batteries

Industry Challenges

The past decade has been a golden period for the new‑energy industry. Global battery market size is projected to exceed $5 trillion by 2025, with installed power‑battery capacity reaching 1 187.1 GWh in the same year—over 60 % of which is in China. The storage‑battery market is also booming, expected to surpass $3 trillion by 2030. However, the sector is shifting from a “trial‑and‑error” model to a data‑driven paradigm, where AI‑based lifecycle management is needed to curb R&D costs.

Challenge 1: Data Flood

Each battery, once sold, continuously streams temperature, voltage, charge‑discharge status, etc., generating **millions of data points per second**. Traditional T+1 analysis cannot meet the required “seconds‑level perception, minutes‑level tuning” latency; delayed detection of anomalies can cause shutdowns, waste, or safety incidents. Customers demand sub‑second response even when handling tens of billions to hundreds of billions of records.

Challenge 2: Rapid, Diverse R&D

Product development faces dual pressure: higher standards for lifespan and energy density, and tighter cycles of 1‑2 years . Over 100 SKUs (e.g., ternary‑lithium, LFP, sodium) split R&D resources, making collaboration difficult.

Challenge 3: Long, Complex Manufacturing Process

Battery production involves dozens of steps; any minor deviation can affect final yield. Despite high automation, data silos persist, preventing fast root‑cause analysis and conflicting with the need for second‑level production cadence.

Challenge 4: Multi‑Dimensional High Standards

Capacity Utilization: Long build cycles cause under‑utilized capacity.

Quality & Safety: Yield must rise from 93 % to 99 % .

Full‑Lifecycle Traceability: Queries take hours, raising compliance costs.

Real‑Time Response: Anomaly handling requires sub‑second alerts, yet cross‑system data fragmentation hampers decision speed.

SelectDB Solution Overview

SelectDB is a real‑time, high‑performance data‑warehouse built on Apache Doris . It combines lake‑warehouse integration, full‑text and vector search, and enterprise‑grade features such as transparent encryption, disaster recovery, visual development, intelligent O&M, and expert services.

Highlight 1: Real‑Time Performance

SelectiveDB supports **million‑level data ingestion**, batch or streaming writes, updates, and deletes. A single cluster can handle **tens of thousands of concurrent queries** with sub‑second latency. Benchmark tests show industry‑leading query speed compared with traditional solutions.

Highlight 2: Lake‑Warehouse Unification & Ease of Use

The unified architecture breaks the barrier between data lake and warehouse. Users can query MySQL, Hive, Paimon, etc., with a single SQL statement, achieving **millisecond‑level** response for performance‑sensitive workloads. The design has been validated on petabyte‑scale production environments.

Highlight 3: Compute‑Storage Separation

Traditional monolithic storage uses three replicas, inflating costs. SelectDB’s compute‑storage separation isolates metadata services, allowing independent elastic scaling of compute and storage. Since its 2022 release, the model has been open‑sourced and widely adopted.

Typical Use Cases

Manufacturing Operations: Real‑time monitoring of production status and equipment health; SelectDB handles million‑level data streams and provides second‑level queries for rapid decision‑making.

Battery Traceability: Forward and reverse traceability across dozens of tables and petabyte‑scale data; query latency drops from hours to seconds/minutes.

Process Analysis: Machine‑learning‑driven optimization of coating thickness, drying temperature, etc.; SelectDB’s SQL + Arrow Flight export outperforms traditional JDBC by **hundreds of times**, supporting SPC quality‑alert models.

Battery‑State Real‑Time Monitoring: Millisecond‑level monitoring of voltage, temperature, internal resistance; enables proactive fault prevention and reduces recall cycles from weeks to hours.

AI‑Driven Real‑Time Quality Inspection: Stores vectorized image features; retrieval quality matches leading vector databases, feeding AI models for defect detection and iterative improvement.

Legacy Data‑Warehouse Upgrade: Migrates from GP/Hadoop to SelectDB, achieving **30×** query speedup, **60 %** storage cost reduction, **67 %** O&M cost cut, and millisecond‑level analytics.

Key Customer Cases & Value

SelectDB serves leading battery manufacturers such as CATL, BYD, Guoxuan, and Zhongchuang. Most have adopted the commercial edition.

Case 1: Unified Lake‑Warehouse for a Top Battery Maker

The customer faced minute‑level anomaly detection, hour‑level post‑sale fault lookup, and high storage costs. After adopting SelectDB’s lake‑warehouse integration (with Hive), data freshness improved to seconds, query speed increased **10‑15×**, storage cost dropped **60 %**, and the system has run stably for over two years.

Case 2: Battery Product Traceability

Legacy architecture indexed only the first field, causing full‑table scans and multi‑minute query latency. SelectDB’s XtoDoris migration introduced inverted and n‑gram indexes, delivering **second‑level** responses for non‑first‑field queries and **2‑3×** faster fuzzy searches on petabyte‑scale data.

Case 3: Haichen Energy Smart Manufacturing

Integrating ERP, MES, QMS data, the solution processed over 3 600 million rows, with P99 latency of **23 seconds** across **6 000+** tables, dramatically simplifying operations.

Joint Hande × SelectDB End‑to‑End Data‑Intelligence Solution

Hande contributes 30 years of IT service experience, 6 000+ customers, and 20 000+ successful projects. SelectDB, built on Apache Doris, offers extreme performance—hundreds of billions of rows with sub‑second response—and a cloud‑native elastic architecture that reduces total cost of ownership by **>50 %**. Together they provide a one‑stop data‑intelligence service covering data collection, real‑time analysis, and AI‑driven applications.

Round‑Table Discussion Highlights

Real‑time data‑driven manufacturing is mature for leading firms (70 % demand sub‑second latency) but still transitioning for SMEs.

Key bottlenecks are data governance and decision‑feedback loops, not raw processing speed.

Battery data differs from internet data: frequent updates/deletes, multi‑modal sources, and complex joins across dozens of tables.

Multi‑modal search (image + process data) requires hybrid retrieval: precise SQL filtering, keyword search, and vector similarity.

Value of real‑time traceability: second‑level queries, deep provenance from finished product back to raw material, reduced O&M cost, faster experiment cycles.

Future (3‑5 years) will see unified lake‑warehouse architectures, deeper AI integration, and data volumes expanding from hundred‑billion to trillion‑level.

Conclusion & Outlook

SelectDB provides a real‑time, high‑performance, unified data foundation for the new‑energy sector, enabling R&D efficiency, production quality, compliance traceability, and intelligent‑manufacturing upgrades. The roadmap points to tighter AI‑data fusion, architecture simplification, and continued cost reduction.