What a calibration sheet taught me about every reading since

The alarm log at the Beiyanling Coal Company monitoring centre printed every methane reading above threshold, and in my first month I was asked to summarise it for the chief safety engineer. I produced a neat table of alarm counts by sensor. He returned it with one question pencilled at the top: which of these sensors had been calibrated that week? I could not say, because the calibration sheets were kept in another office. That question has followed me through twenty-five years in energy. A number without its collection history invites a confident answer to the wrong question.

I joined Beiyanling in August 2001, a month after completing a degree in applied statistics at Lushengtai Institute of Technology. For five years I analysed sensor readings, alarm logs and calibration records for the safety-monitoring system of two underground mines. I had no authority over ventilation or production; the safety engineers decided what an alarm required. My work was to make sure they saw each reading with the context needed to trust it, including sensors overdue for calibration and periods when a monitoring line was down.

In September 2006 I moved to Taolingshan Thermal Power, a coal-fired plant with four units. As a performance and fuel analyst I reconciled coal receipts, weighbridge records, laboratory analyses and unit heat rates. The fuel account rarely closed cleanly. Moisture, sampling practice and stockpile surveys all changed the answer, and I learnt to show which difference came from which source before anyone called it a loss. From 2012 I supervised fuel and emissions data, including continuous emissions monitoring and the plant's first carbon-emissions reports. I reported findings to the plant's safety and environmental committee. When coal-sample testing later came under official scrutiny across the industry, I was glad our chain of custody from sample to laboratory result had been written down years before anyone asked for it.

The power-sector reforms of 2015 opened electricity retail to new companies, and in May 2016 I joined Hanxiyuan Energy Services as its first load and market analytics manager. The company had a retail registration, a few industrial customers and almost no data of its own. Customer consumption arrived through the grid company's metering service a day or more after the fact. I built our first customer load forecasts for annual and monthly contracts, and later prepared analytics for the province's spot-market trial runs. Traders decided the positions. My team's task was to tell them how far each forecast deserved their trust.

Hanxiyuan bought a public charging operator in 2019, and I became head of data and analytics for both businesses. Charging brought a different kind of record: a driver's phone number, a plate number for parking-fee waivers, a session time and a station location. Retail customers are companies; drivers are people, and their movements are personal information. I spent the next three years moving both businesses onto one platform while keeping drivers' records separately governed.

I became chief data officer in March 2022. I report to the general manager and present to our board and its risk committee, though I have never held a directorship. The three cases on this site show the work I consider central to the role. Paying for customer metering gateways meant postponing a price-forecasting service our head of trading preferred. Declining a vehicle maker's offer for session-level charging records gave up RMB2.5 million a year. Locking out or limiting 232 connectors before the Spring Festival peak meant queues, complaints and lost revenue in the busiest weeks of the year.

I lead through a few written rules rather than frequent approvals. Every dataset has an owner outside my office, a stated purpose and a retention period. My teams may refuse a request that arrives without a purpose, and they are expected to explain the refusal. I spend more time with traders, field engineers and legal colleagues than with technology vendors, because the useful questions about data usually arise where someone is about to act on it.

My mistakes have mostly been about timing. Charger alarms and maintenance work orders sat side by side from 2021, and I linked them only in 2025 because I had given trading analytics priority. I now ask each quarter which safety-relevant records nobody in my office has examined, and I accept that some answers will be uncomfortable.

I would like to bring that discipline to a board where data, safety and customer trust meet commercial pressure. I would also need to learn the new company's obligations properly, rather than assume that an electricity retailer's rules carry over unchanged.

Open to Board appointments

Languages / Mandarin · English