Taipower

AMI Deployment

Taipower  has installed AMI meters for large (all extra-high-voltage and high-voltage) customers in 2013. The load structure of Taiwan's power system is mainly composed of large-scale (high-voltage) industrial and commercial users, and their power consumption is about 60%. For small (low-voltage) customers, Taipower plans to install up to two hundred thousand, one million and three million AMI meters in 2018, 2020 and 2024 respectively. The schedule was approved by Executive Yuan in September 2016.

There are more than 15 million electricity customers supplied by Taipower. Even if finishing 3 million small customers’ AMI deployment in 2024, the penetration of AMI users is only about 21%. However, the power usage of large-scale customers with installed AMI are about 60%, after finishing 3 million customers’ AMI deployment in 2024, Taipower could handle around 81% of power usage information of Taiwan’s power system.

Deployment Schedule

Taipower conducted low voltage AMI smart meter deployment via separate tendering of meter, communication, and meter data management system. The current status of deployment is as shown below:

• Meter
  The procurement of 200,000 units of low voltage AMI smart meters was completed in 2017. The accumulated installation of 230 thousand smart meters was completed in 2018 and 696 thousand smart meters in 2024. Last year Taipower has deployed about 3.4 million smart meters for commercial and residential customers, which occupies about 81.5% of all customers' power usage in Taiwan.

• Communication module
  The procurement of 200 thousand units of communication modules was completed in July 2018. The installation of 230 thousand communication modules was completed in March 2019. In 2024, we have installed about 3.4 million communication modules. The total system connection rate between MDMS and smart meter is over 97%.

• Metering data management system
  Metering data management system (MDMS) has been on-line system in 2020, and it can successfully collect and manage 2 million meters data.

MDMS Data Interfaces & Service Applications

Along with the schedule of AMI deployment, Taipower is planning to construct a MDMS (Meter Data Management System) which can accommodate with some millions of customers’ AMI data.

MDMS is extremely important in the architecture of AMI. It should be able to offer processed data, which is consistent and can be directly used by other applications, to make sure the applications can be integrated on the same basis. The repetitive data pre-processing work (ETL, Extract-Transform-Load) can also be avoided. MDMS can help earn valuable application by providing good data, related application such as power dispatch, renewable energy, equipment performance, equipment maintenance and power restoration, shown as figure below.

Achievements

The section below will highlight some key subjects/items of research and development group as examples to illustrate the application contents of the plan. Every other application subject/item within the framework have been assigned to responsible departments for the administration, which are regularly tracked and controlled. 

1. xEMS (Community, Building, and Home Energy Management System)
   xEMS is the concept of energy management in the area of community, building, and home, which also known as CEMS, BEMS, and HEMS, respectively.

   1. 1.1   CEMS (Community Energy Management System)
      The introduction of CEMS can help to manage the energy usage inside a community effectively. With measuring and monitoring the instant power load for every responsible area, the system can calculate reduced demand potential immediately within the community to support overall load management and operation. 

   2. 1.2   BEMS (Building Energy Management System)
      BEMS focuses on the power management inside each building. With checking the electrical loops and equipment inventory in the building, measurement and control elements are installed to master the real-time electricity usage. Besides, with the participation in DR programs of Taipower, users can reduce the peak load and get further rewards. 

   3. 1.3   HEMS (Home Energy Management System)
      HEMS assists home members to master the electricity use inside the home. With the communication network technology, integration of AMI and HEMS and connection with energy-saving household appliances, HEMS can provide instant power usage information, manage and control appliances by intelligent schedule, analyze the power usage status on smart home platform, perform demand response request and energy management service functions. 

   

2. Analysis of Demand-Bidding Participants Performance

   

   Demand-Bidding program is one of the most effective DR programs of Taipower. However, the load reduction or even the transfer affects the accuracy of load forecasting. This analysis provides system operators with better understanding of DR management. The graph shows the transfer effect from DR participants’ performance. The x-axis shows a 48-hours timeline. During the event (D-day 15-17), the load reduction is obvious while transferred to the early morning of both D-day and the next day.

3. Potential Load Reduction Quantification of DR programs

   

   The effectiveness of DR is decided by participants’ performance. Taipower uses ML (Machine Learning) and AI (Artificial Intelligence) technologies to estimate the amount of potential participants’ load reduction, which can help us to find the most wanted customers. After taking the experience and modeling technologies from previous result, we develop this analysis to precisely identify target customers with their reducible load.

   The targeted customers with their individual profiles such as industry, location, contracted capacity and predictive load, etc. the result has been visualized as a dashboard to facilitating DR apartment for precision marketing.

4. Industrial Parks Power Usage Profiling

   

   A descriptive statistics analysis was done for the pilot project of chosen industrial parks, the above figure illustrates an analysis result of the amount of power usage and users of different industries. More detailed cross-profiling has also been done, which takes feeder information into account. The analytics procedure is expected to duplicate to other industrial parks for investigating precise industrial features and development trend. A visualization map of all industrial parks can be done if all the data and locations(borders) were clearly defined.

5. Non-Intrusive Appliance Load Monitoring (NIALM)

   This project applies to the advanced Artificial Intelligence (AI) based Machine Learning (ML) theory to develop the Non-Intrusive Appliance Load Monitoring (NIALM) model which uses the minute and 15 minutes interval based active and reactive power provided by smart meter as learning patterns. Frequently used appliances such as air conditioner, lighting, refrigerator, electric water heater, dehumidifier, and fans are assessed in the laboratory house for evaluating the performance of NIALM model. The F-Score achieves 83%, which outperforms the published performance.

   The NIALM model was also evaluated in 30 residentials for testing air conditioner and lighting. The average F-Score achieves 85%, which also outperforms the latest domestic published performance.

   

6. Electricity Usage Routine Indicator

   As Taiwan’s aging population continues growing, challenges of aging society is increasing. Based on caring for the elderly who do not live together, we have developed the “Electricity Usage Routine Indicator” using AMI (Advanced Metering Infrastructure) data which record every 15 minutes electricity consumption, allowing real-time monitoring of household electricity usage. By analyzing AMI data to derive customers’ electricity usage patterns, we can detect unusual status according to their electricity usage routine patterns and provide information for reference in real time. This application is launched in March 2025 on the Taipower APP for free.

   

   At the beginning, we find an electricity consumption threshold of each user based on past 28 days AMI data, then compare electricity consumption with the threshold. If the load exceeds the threshold, the status is set to 1. If the load is lower than the threshold, the status is set to 0. Next, summary status to derive patterns. For example, at the time point of 00:00, average status of the past 28 days. We can derive that the customer has around 85% chance using electrical appliance during 00:00 to 01:00. On the other hand, this customer never uses electrical appliances at 06:00 during the past 28 days. With the derived patterns, we can provide electricity usage information to customers to enhance communication between families.

   

Open Data Platform of Government and Taipower

As of March 12, 2025, Taipower has 187 Datasets available on the open data platform for downloading and use by the public. In accordance with the Ministry of Economic Affairs' data openness promotion strategy and set goals, Taipower regularly takes stock of business data and evaluates the demand for external data from time to time on the basis of the three principles of non-openness, including personal data privacy, business secrets, and national social security. After structuring, de-identifying, and passing data quality checks, we will promote the diversification of data applications and their scope of application and open them to the public in open formats or APIs to create higher value in the electricity market together with the public.

Optimizing Taipower's Operational Efficiency Through Advanced Data Analytics Infrastructure.

Introduction

Taipower strategically invests in a sophisticated big data platform, emphasizing the optimization of data analysis, with a specific focus on high-frequency Smart Meter Data collected every 15 minutes. This optimization not only enhances operational efficiency but also brings tangible benefits, such as cost savings and faster decision-making. Combining this data with customer attributes, weather patterns, and economic activities requires a robust data warehouse with massively parallel processing (MPP) capabilities, ensuring seamless data collection, storage, computation, and analysis.

Taipower achieved a significant milestone with the successful implementation of the "Meter Data Management System (MDMS)" in 2020, encompassing all Advanced Metering Infrastructure (AMI) customers' data. Additionally, Taipower ‘s Enterprise Data Platform commenced operations on December 1st, 2021.

The Plan of Enterprise-level Platform

1. Origin of the plan

   With the continuous advancement of Taipower's operations, driven by the deployment of smart grids, widespread adoption of smart meters, and the flourishing influence of renewable energy sources, the Taipower's business operations is becoming increasingly complex. This requires the establishment of a comprehensive enterprise analytics platform to bridge data silos among business units and establish a forward-looking, data-driven strategy. The primary goals include enhanced grid management, more efficient demand-side operations, improved power quality, heightened operational efficiency, increased customer satisfaction, and an overall enhancement of company profitability and competitiveness.

2. Project goals

   The project delineates three specific objectives:

   1. [1]Integration of Company-wide Operational Data

      Establish an Enterprise analytics platform that seamlessly connects smart grid data, including AMI meters and ERP transactions. This initiative aims to create a smart enterprise integrated structure, enabling real-time insights for faster and more informed decisions.

   2. [2]Training Data Scientists and Promoting AI Application

      Invest in cultivating top talent by empowering data scientists with professional training platforms, which equips them with a comprehensive skillset, including data extraction, cleaning, enrichment, and model building. This approach fosters unique core skills combined with domain-specific knowledge, positioning the company advantageously in the evolving landscape of data-driven strategies.

   3. [3]Implementing Data Transparency and Enhancing Data Activation

      This involves constructing an enterprise-level analysis environment built on three key pillars: data source, data storage platform, and data visualization platform. These components work together to streamline data collection, boost storage and computation, enabling effective visual analysis.

3. Major Tasks:

   The core tasks involve constructing an enterprise-level analysis environment, which includes:

   
   Figure 1 Architecture Diagram

   1. [1]Data source

      Focus on collecting and cleaning data from multiple sources, ensuring data integrity through validation, quality assurance, and translation. Employ automatic scheduling mechanisms to manage transmission processes efficiently.

   2. [2]Data storage platform

      Implement a highly scalable platform that provides resources for high-speed computing and flexible data storage using MPP architecture.

   3. [3]Data visualization platform

      Empower users with visual analysis tools for creating interactive reports. These tools enable data-driven decision-making through diverse chart types and robust statistical functions. Additionally, the platform allow users to leverage algorithms and build advanced analytics models for predictions and optimizations.

   The implementation of this enterprise-level platform will not only streamline data management but also empower Taipower to make informed decisions, improve operational efficiency, and keep a competitive edge in the evolving energy landscape.

Use Cases

1. TPC e-Bill service
   1. [1] Service Description

      The TPC e-Bill service with simple registration and customer number setting has been provided to TPC users for quick inquiry of power consumption data and easy bill management. It is presented via easy-to-understand electronic interface and micro-animation such that users can have quick access to bill related information to facilitate the self-management of amount power consumption behavior and to achieve the energy saving effect.

      The main visualization design of the front page is based on the illustration style with the layouts of four seasons. And there are the renewable energy subjects of wind power, solar energy, and hydraulic power generation to attract users to TPC e-Bill service in order to achieve the goal of paper reduction and environmental protection.

      
      Figure 1 Front page main visual design (Spring)

      
      Figure 2 Front page main visual design (Summer)

      
      Figure 3 Front page main visual design (Fall)

      
      Figure 4 Front page main visual design (Winter)

   2. [2] Function Description
      1. a. General User

         “General User” can have access to the website related information including payment status enquiry, bill pre-estimate, website tutorial, FAQ, and news.

         
         Figure 5 Front page of TPC e-Bill service

      2. b. Members

         Users can register to become the “Member” of TPC e-Bill service. In addition to periodic bill notification and payment certificate in the form of email, users can also have access to historical bill data. A member could login the TPC e-Bill service to have access to the functions of basic information management, bill inquiry, and analysis of amount consumption chart. To make the operation easily, the bill information of all registered customer numbers will be listed on the login screen, and the provided functions are the latest monthly bill, total amount, payment status, the payment barcode to be downloaded (the unpaid bill can be directly downloaded), re-sending of electronic notification, and historical data.

   3. [3] AMI Real-Time Inquiry

      To facilitate the real-time inquiry of amount consumption data by AMI smart meter users and thus promoting the power consumption management, the “AMI Area” icon behind the customer number of AMI could be selected to enter the AMI dashboard function, which provides various functions such as comparison of various real-time amount consumption data, self-defined upper limit of amount consumption, analysis of amount consumption statistical chart, and rate service evaluation.

      Note: Users could enquire real-time amount consumption data of the previous day by noon.

      
      Figure 6 My Bill

      1. a. The percentage comparison of total AMI amount consumption and total amount trend

         The percentage comparison of total AMI amount consumption of current year and last year will be shown on the indicator dashboard. The monthly power consumption trend for the two years next to the dashboard to directly compare whether the total consumption has increased or decreased.

         
         Figure 7 The percentage comparison of total AMI amount consumption and total amount trend

      2. b. Comparison of average total consumption in the same area/industry

         The users could check their own total consumption and the average power consumption in the same area or industry. And it would provide different contents via different scenarios to guide users to change their energy consumption behaviors.

         
         Figure 8 Comparison of average total consumption in the same area/industry

      3. c. Setting Daily Power Limit

         Provide users with the ability to set the upper limit of daily electricity consumption. When the electricity consumption exceeds the self-set degree, the user will be notified by email.

         
         Figure 9 Self-defined amount consumption upper limit and abnormality notification

      4. d. Carbon Emission

         According to whether the amount consumption exceeds the " average total consumption in the same area " or " Self-defined amount consumption upper limit ", the icons of carbon emission and trees are provided to strengthen the connection between amount consumption and energy saving.

         
         Figure 10 Carbon Emission

      5. e. Monthly Consumption

         The total amount consumptions of all months are presented and compared in the form of monthly pie chart, and different colors (green, yellow, red) are used to show whether or not it is increased as compared to the base month.

         
         Figure 11 Monthly Consumption

      6. f. AMI Statistics

         The real-time and detailed AMI amount consumption charts are presented in graphics to shown the statistics based on “year”, “month, “day”,“15 minutes”,and the daily (24 hours) comparison.

         
         Figure 12 AMI real-time statistical chart

      7. g. Rate Suggestion Pre-estimate

         The trial calculation comparison amount of simple rate plan is provided according to the past AMI real-time amount consumption information of users to help users assess which rate plans are more appropriate according to their energy consumption habits.

         
         Figure 13 Rate Suggestion Pre-estimate

      8. h. Unbilled Electricity Bill Amount Estimate

         According to the past AMI real-time amount consumption information of users, we provide the estimate of the unbilled power consumption and the bill amount to help users control electricity usage.

         
         Figure 14 Unbilled Electricity Bill Estimate

      9. i. Amount consumption analysis chart

         Users are provided with the power consumption charts of recent 48 hours for comparative analysis with the same timeframe of last week and last month.

         
         Figure 15 Amount consumption analysis chart

      10. j. Maximum Demand Alert

          If the Maximum Demand(kw) exceeds 95% of the Contracted Capacity(kw), the user will receive reminder email, which remind user to avoid Customer Charge(Non-Contracted) fee.

          
          Figure 16 Contents of notification letter

   4. [4] Future planning
      1. A. Enhancement of interactive design

         We will continue enhancing the interactive service for users and simplify operation interface and enhance the friendliness of website by user friendly and dynamic guidance approach in order to enhance users’ willingness of registration and use while achieving the energy saving efficiency of all people.

      2. B. Add new application functions

         A new method is added to estimate the number of days of use based on the amount of electricity bill, allowing users who need to prepay electricity bills to estimate the number of days that the prepaid amount can be used, and then determine the prepayment amount according to needs.

      3. C. Continue to integrate member accounts with Taiwan Power APP

         Allow users to use services on various platforms of TPC through a single set of account passwords.

2. Future planning- Taipower APP
   1. [1] Electricity information and comparison:

      Provide users to select different time intervals to view the electricity consumption overview, such as every 15 minutes, hours, days, months, or choose two intervals for self-consumption comparison, or with comparing the average electricity consumption of users in the same area. Self-management of electricity consumption.

      
      Figure 17 Electricity information and comparison

   2. [2] Residential electricity analysis:

      By utilizing power usage data and survey responses from users, we provide an analysis and estimation of home appliance usage patterns. To access this service, first-time users must click "Subscribe Now" and complete the electricity analysis questionnaire. The analysis results will be available the following day under 'Residential Electricity Analysis', displaying a breakdown of power consumption for each household appliance.

      
      Figure18 Residential electricity analysis

   3. [3] My category:

      Users with stable communication low-voltage AMI power service. Electricity consumption before scheduled meter-reading day: We provide trial calculation for the amount occurred from the last scheduled meter-reading day until 0:00 yesterday. We also provide the amount of the same period last year for comparison and you can make the prepayment immediately.

      
      Figure 19 My category

   4. [4] Rate trial calculation:

      We provide rate package based on the trial calculation of the latest electricity bill for your application. Rate overview: Introduction of all the rate packages. Rate trial calculation: We provide rate package that is applicable to your electricity meter and will demonstrate the current and the optimal one (with the lowest rate) for reference.

      
      Figure 20 Rate trial calculation

   5. [5] Electricity consumption reminder(Only the account holder can use):

      A daily limit or budget for electricity consumption can be set, and the system will schedule a review of the previous day's usage or calculate the electricity bill. When the set limit is exceeded, a push notification will be sent to the person who set the limit.

      
      Figure21 Electricity consumption reminder

   6. [6] The prepayment:

      We provide different methods of prepayment for general users and AMI users(users with stable communication low-voltage AMI power service).

      1. a. General users: Display the amount that has been calculated (this amount has not been billed in the New Billing System) User can enter any amount and pay it off.

         
         Figure 22 General users

      2. b. AMI users: We provide trial calculation for the amount occurred from the last scheduled meter-reading day until yesterday for reference. You could choose to input the amount or days of used. If the input amount is greater than that of the trial calculation, the system will access and show the remaining available days. If you choose to input days of used, the system will conduct a trial calculation to estimate the future electricity bill(including the amount calculated until yesterday). Any unpaid bills must be included in the prepayment.

         
         Figure 23 AMI users

   7. [7] Relocation settlement:

      In addition to self-meter reading、taking photos of meter and making an appointment for Taipower personnel to read the meter for settling bill. AMI user that has been registered in the New Billing System can input the settled date and if the meter-reading was successful, the settlement amount will be calculated automatically. In the event of meter-reading failure, it will be the same as the non-AMI users that users could choose the method of meter-reading.

      
      Figure 24 Relocation settlement

3. Explore potential target customers through AMI data analysis

   Leveraging AMI data analytics, we have identified potential participants for the Demand Response Program. By constructing predictive models based on historical electricity consumption patterns, we have precisely targeted high-potential users, thereby significantly enhancing program participation and effectiveness. In 2024, 1,064 target users were selected for the Demand Response Program. On the night peak load day of July 22, 2024, these users achieved a demand reduction of 1,180 MW, representing 90% of the total demand reduction across all participating users (1,310 MW). This contribution bolstered the system's operating reserve by approximately 3.5%.

Hosting Capacity Visualization System for Transmission Power Supply System and Power Distribution System

1. Background Information

   In recent years, with the government expanding its efforts to promote the green energy development policy, related renewable energies such as photovoltaics, wind power, hydroelectricity and geothermal power, all of which are Distributed Energy Resources (DER), will be massively grid-connected to the distribution system, and it will significantly change the entire distribution network structure. The electric power will no longer be generated by the generator units in large power plants, but rather be generated by small and decentralized devices. In order to fully understand the grid-connection status of the renewable energies, Taipower began to develop the Distribution Planning Information System (DPIS) since years ago, attempting to regulate the possible impact to the system when those energies are grid-connected by adopting the application review mechanism for Distributed Energy Resources.

   According to records, there are nearly 10,000 feeders in the power distribution lines in Taiwan. The amount of data is huge and its number in the hosting capacity sections is as high as 3.26 million. When presented on a traditional web-based graphics platform, it would consume a considerable amount of software and hardware resources and may result in performance problems caused by a large amount of map data loading. Therefore, the Map Tile display technology will be used to generate the map tile files after the incorporation of timed calculations of each hosting capacity section, and then combined with the TGOS Address Geocoding Service and the generic version of the electronic maps to construct the visualized distribution feeder hosting capacity system. The system provides the public or the industry to conduct a location query by means of an address to achieve the purpose of visualizing the distribution feeder hosting capacity.

2. System Architecture Description

   

   The system is divided into two parts: the internal system and the external system. In order to allow external users to have a more intuitive understanding on the possible hosting capacity of feeder line, this project will regularly extract the topology information of existing distribution equipment from each business branch of Taipower by using the Extract-Transform-Load (ETL) technology. This project then simplifies it to become the information structure required to calculate the possible hosting capacity and subsequently summarize and send it to the information center as the basic information for computation. We tested several computing cores and compared their benefits. Analyze its accuracy and calculation speed, and choose the best one.

   Calculation results will integrate to the Geographic Information System (GIS) technology and displaying the possible hosting capacity of each feed line section on the map with colors. Furthermore, to solve the performance issue that may be caused by external users downloading a large number of map data, this project constructs a feeder hosting capacity visualization system by using Map Tile display technology. It was thereby achieving the establishing hosting capacity visualization of the feeder.

3. Power System Impact Analysis

   

   This system integrates data from several systems automatically, such as Distribution Mapping Management System, Renewable Energy Management Systems, Distribution Planning Information Systems, etc. Input data mainly include wire parameters, transformer parameters, electrical connectivity, power distribution equipment data, transformer data, load data, renewable energy data, user data, coordinate points, etc. Comprehensively organize all the materials and build a complete feeder input file. The schematic diagram of power system impact analysis is shown in the right picture.

   Taipower has about 10,000 total feeders, and the number of calculations for each joint is very large. Considering the overall load of the mainframe and the accuracy of the data, the core of the most cost-saving calculation is adopted. Each joint will be inversely calculated one by one, and get the maximum hosting capacity finally. It will be the basis of updating the externally displayable hosting capacity.

4. System introduction

   

   In order to allow the user to clearly view the hosting capacity of each section of feeder, we indicate the hosting capacity of each section by means of color display. This referred to the webpages which are built by the foreign electricity utility.

   The hosting capacity is divided into six levels and six colors from small to large, and display on the geographic map.

   Feeder hosting capacity visualization system for power distribution system:

   https://service.taipower.com.tw/hcweb/geohc/

5. Future Development

   The system is designed to collect data, calculate hosting capacity and release map tiles automatically. Through the artificial elimination of abnormalities, it is expected to improve the system's capabilities.

   Compared with the relevant open systems of advanced countries, this system is more advanced, integrated and flexible in development. It can provide consistency comprehensive information for the government, the electricity industry, the industry, and users about renewable energy build land and hosting capacity. It may help to achieve 20% renewable energy government policy before 2025.

   In the future, more information will be collected to strengthen the system impact analysis and reduce the need for human intervention. We will continue to study and modify the logic of constructing feeder and make the feeder closer to the actual situation.

Renewable Energy Monitoring

1. Introduction

   TaiPower company had 174 onshore wind turbines in 18 sites and 21 offshore turbines in 1 site in Taiwan, including 6 brands and 9 models. Each brand of manufacturer has developed and designed its monitoring system. To facilitate the unified and centralized management of real-time operation information of wind power stations scattered across Taiwan and outlying island, a complete integrated monitoring platform system has been established since 2011. After the compiled program interface determines the status of various types of wind turbines, it will organize the operation status in real-time and display on the website.

2. Architecture of the Monitoring platform system

   Due to the difference in the communication system of wind turbines, we need to integrate different wind turbine operation information. We set up a system provided with an interface computer beside the original monitoring computer of each wind power plant, connect with the OPC communication protocol to retrieve the real-time operation information. The data is transmitting from the regeneration office via of Taipower company through the communications network to monitoring center server room in office (PISRV1，main server)and Onshore Wind Power Plant computer server room (PISRV2，backup). Then the original data of PISRV1 will transmit to another server (PISRV3) on the company intranet through an isolator, finally established the data structure through a logical layer computer (PIAF01). The web host (PIWEB) then retrieves the data and builds a website for users to connect with the browser to improve network.

   

   Each wind power station system is roughly similar. Taking the〝Datan wind power station〞as an example, it was established in three-phase with a total of 7 wind turbines, 3 different brands of the manufacturer, including 2 GE’s、2 Enercon’s and 3 Vestas’s wind turbines. Therefore, it is more complicated to retrieve data in the interface. The data transmission structure is as shown below.

   The wind turbine operation data of Datan wind power station was retrieved from the original computer SCADA data and directly transmit to the main engineering room in Taichung control center office through the interface of Datan power plant microwave computer room.

   

3. Security and Convenience

   In response to the intermittent renewable energy power generation, Taipower currently collects information on renewable energy power generation systems and developed a 48-hour renewable energy power generation forecasting system through algorithms. It can help to optimize the dispatch and be used to arrange the unit's scheduled maintenance as a massive amount of renewable energy feed-in the grid in the future.