Solar power generation/load forecasting servicesAI

Self-developed software platform

Station Controller

Station Control EMS

Station Controller is a locally deployed station-level software solution designed for high-performance operation. It facilitates millisecond-level control and monitoring communications with local devices and the power grid. Because Station Controller is technology-neutral, its deployment is not limited by device brand or application.

Key Features

  • Multi-device power distribution
  • Equipment power control
  • North-South Communications
  • Data backup

    • Strength

  • Fast implementation
  • Avoid overcharging/overdischarging/ overbooking/Contract Demand
  • Power control < 1s
  • Power adjustment in seconds

    • Solar power generation/load forecastingAI

    Operating stage

    Applying self-developed physical and artificial intelligence models, we provide solar power generation and load forecasting services that can be quickly implemented for pre-meter and industrial and commercial energy storage projects.

    Service Offerings

  • The day-ahead forecast can provide solar power generation/load forecast services up to 14 days in the future
  • Up to 10 minutes accuracy
  • Update every 15 minutes
  • Optional API format
  • Provides long-term solar power generation forecasting services

    • Cooperated with an automation control factory in Hsinchu to calculate the solar power generation data of the site for one year and compared it with the solar power generation time-sharing data (smart meter data) of the past year

    8% Monthly average error

    The model has learned the seasonal differences in solar power generation

    Accurately simulate field conditions

    0.75% Annual total error

    The model is quite accurate in estimating the annual power generation

    Accurately estimate the benefits associated with solar power generation

    We have cooperated with a PV power station in Chiba Prefecture, Japan, and conducted POC testing with permission.

    At 8 a.m. every day, the solar power generation model predicts the solar power generation every 30 minutes for the next two days and performs regression tests using actual power generation data.

    0.65% ~ 3.8% Error in clear weather

    6.43% ~ 9.29% Error in cloudy and rainy days

    (Using the data from 2023/04/08 – 2023/04/16 as example data)
    The prediction accuracy has reached the level of first-tier weather forecast manufacturers in the market.

    We conducted cooperative testing with multiple factories in Japan and, with their permission, used their load data from the past year to train the model, and then conducted a one-month POC test.
    At 8 a.m. every day, we use the day-ahead load forecasting model to predict the load every 30 minutes for the next two days and perform regression testing using actual load data.

    0.5% ~ 1.3% daily average error

    (Select the meter data of a certain road in a steel plant in Ibaraki from 2023/06/15 to 2023/06/18 as the sample data)
    The model can automatically and accurately distinguish the difference between working days and rest days, and use this as a basis to more accurately guide the behavior of the energy storage scheduling optimization model.

    Other Services