Non-invasive load monitoring (NILM) technology can obtain the electricity consumption information of various electrical devices of users without intruding into their premises, solely through the analysis of data from their electricity meters. NILM has been extensively researched and applied in residential load disaggregation, but its application in industrial loads is limited. On one hand, industrial loads differ significantly from residential loads in terms of load characteristics and data distribution, leading to a noticeable performance decline when methods designed for residential scenarios are applied to industrial settings. On the other hand, industrial users, concerned about privacy protection, are reluctant to disclose their electricity consumption data, making it highly challenging to effectively learn about industrial load equipment using limited data. To address these issues, an industrial load disaggregation method based on the factorial hidden Markov model (FHMM) is proposed. This method utilizes multiple independent hidden state chains of the FHMM to simulate the operational state transition process of industrial load equipment. By determining the state of the equipment at each moment, the electricity consumption of the equipment can be predicted in conjunction with state-specific energy consumption information. Finally, the proposed method is tested using on-site energy consumption monitoring data from a factory, and the results demonstrate its effective load disaggregation performance.