Background & Summary

Distributed energy systems present opportunities for localized and flexible energy production and distribution, which have the potential to deliver various benefits, including primary energy saving, enhanced resiliency, and carbon emission reduction1,2. Their significance is particularly noteworthy within the context of the ongoing transition towards decentralized and decarbonized energy systems3,4. The use of distributed power generators and absorption chillers in meeting high energy sufficiency and shaving peak load through waste heat recovery is gaining increasing attention in building energy systems, especially in the face of a changing climate5,6.

District energy systems are usually designed to meet diverse energy services, including electricity, heating, cooling, and hot water demand. District building energy demands exhibit temporal fluctuations, characterized by significant seasonal and daily variations7,8. Additionally, energy performances of district building energy systems also present a close relationship with the weather conditions, operational strategy, transmission pipeline design and implementation9,10. Those factors pose challenges for a fair comparison of decentralized and centralized energy supply systems in real-world projects11,12. Several works have even argued the energy efficiency of distributed energy systems13,14. Most of the currently available open data includes only a limited number of years of monitoring, which is insufficient to cover the entire lifespan of distributed energy generators. Additionally, long-term datasets are typically monitored at a low resolution.

Against this background, monitored datasets with long-duration periods are crucial to accurately understand the value of distributed energy systems in real-world projects. To the best of our knowledge, there exists no descriptive analysis of the actual operational performances of district energy systems with and without on-site distributed energy generators throughout their lifespan. This knowledge gap hinders our understanding of the long-term sustainability and adaptability of distributed energy systems in face of evolving challenges.

This work collected and described the 20-year dataset of high-resolution electricity, gas, and heating energy consumptions of a district building energy system in Japan. The analysis of long-term data enables researchers and designers to identify patterns, and areas for improvements and assess the effectiveness of different distributed energy technologies. This enables them to make well-informed decisions regarding the electrification and operation of district energy systems. Additionally, the weather variables including air temperature, humidity, and solar irradiance are also collected at 60-minute intervals, which could be useful for assessing the flexibility and resilience value of distributed energy resources in the face of increasing extreme weather events. This work also includes an analysis of the data quality and provides guidelines on how to effectively reuse the dataset. While the dataset may not extensively represent newer technologies, it provides a foundational reference for understanding the historical evolution of integrated building energy systems, aiding researchers in examining transitions to sustainable and advanced energy solutions.

The dataset can be used for various purposes, such as:

  • Assessing performances of energy sector coupling and different energy storage systems in improving energy flexibility15,16.

  • Evaluating the levelized energy costs of distributed energy resources within their lifespan17.

  • Valuing demand response and energy resilience of district building energy systems considering extreme events, such as the Covid-19 pandemic, heat wave, and cold snap events7,18.

  • Building energy demand pattern analysis and time-series prediction through machine learning algorithms19,20,21.

  • Applying data-driven techniques such as model predictive control and deep reinforcement learning for district energy system management22,23.

Methods

The dataset was collected from the district building energy system situated in Kitakyushu Science and Research Park in Kyushu, Japan24. Detailed structure and energy flows of the energy station are depicted in Fig. 1. This energy system was implemented in 2001, consisting of various components, namely rooftop solar photovoltaic (PV) panels, fuel cell, gas engine, gas boiler, and absorption chillers. The gas engine has a nominal power capacity of 160 kW, with a nominal power generation efficiency of 28.7%. The fuel cell has a power capacity of 200 kW, with a nominal power production efficiency of 40%. Fuel cell runs continuously for 24 hours every day, gas engine operates from 8:00 am to 10:00 pm throughout the year. The fuel cell and gas engine operate in a combined heat and power (CHP) mode to supply simultaneous heat and power energy. The nominal capacity of the solar photovoltaic system is 153 kW and 13% module efficiency. All power generators operate in off-grid mode, PV generation is self-consumed, the on-site generators are operated to fulfill the baseload requirements, while grid import is utilized to cover the remaining electricity load. The absorption chillers have cooling and heating generating capacities of 2110 kW and 1780 kW, respectively. The nominal energy conversion efficiencies for cooling and heating supply are 1.1 and 0.93, respectively. The recovered waste heat from the CHP generators, along with city gas consumption, drives the operations of the absorption chiller systems for building space heating and cooling purposes. The gas boiler has a hot water generating capacity of 80 kW, it is responsible for meeting the daily hot water demand when the recovered waste heat is insufficient to fulfill the requirements. The fuel cell was operational from 2001 to 2011, while the gas engine operated from 2001 to 2016. To provide a more comprehensive description and understanding of the changes in electricity and gas consumption at the energy station with different energy supply structures, this study divides the dataset into two distinct periods: the first ten years and the last ten years.

Fig. 1
figure 1

Schematic of the energy supply system and exteriors of buildings in Kitakyushu Science and Research Park.

Full size image

The district energy system provides electricity, hot water, space heating, and cooling services to a total of six buildings within the park. These buildings encompass two offices (a), a laboratory (b), a gymnasium (c) a library (d), a classroom and office complex building (e) as shown in Fig. 1.

Data Records

All data is recorded in Excel worksheets (xlsx format) and organized into compressed folders, which are accessible to the public via Figshare (https://doi.org/10.6084/m9.figshare.24978645)25. The choice of the xlsx format was due to its exceptional compatibility with a wide array of spreadsheet software, including Microsoft Excel, Google Sheets, and LibreOffice Calc, as well as its seamless integration with popular programming tools like Python and Java. The pressure and temperature of gas and hot water flows are monitored to convert standard flow volumes. A schematic of the original data collection system is illustrated in Fig. 2.

Fig. 2
figure 2

Overview of data collection of city gas and heat energy consumption.

Full size image

Table 1 provides a comprehensive overview of the datasets within each folder. Each folder is dedicated to a specific type of information, encompassing power generation, gas usage, heat recovery, loads for supplied buildings, and weather data. The measurement period spans from June 1, 2001, to October 31, 2022. The collected dataset includes information on district electricity consumption, space heating, space cooling, hot water, city gas consumption, and dynamic supply-demand energy balance. With the exception of data gaps in March 2011 resulting from an earthquake, the dataset remains complete for all other periods. Each Excel file is appropriately named to reflect the respective year, resulting in 22 datasets per folder.

Table 1 Overview of datasets in each folder.
Full size table

The energy Station provides data on the primary energy supplies they provide to the campus community. The dataset is divided into four categories: (1) measured electricity sources, (2) measured city gas usage, (3) measured energy supplies, (4) measured indoor temperature, and (5) weather conditions. Table 2 provides an overview of the primary data for the period of data measurements. Among the energy station equipment, the Gas engine was operational until July 2016, while the Fuel cell remained in operation until November 2011. However, the remaining equipment maintained continuous operation throughout the entire test period.

Table 2 Test data output overview.
Full size table

The data are recorded on an hourly basis throughout the measurement period, with each column representing a distinct variable of the time series. The unit for each variable is denoted in the table head, facilitating easy interpretation. The record of each dataset is highly complete, except for the absence of March 2011 due to the great earthquake. These data are available for download on Figshare25.

Technical Validation

The following subsections present a data visualization to show the quality and technical validity of the datasets. The reliability of measurements can be assessed by the value range and variability of loads at different time scales, from hourly to annual intervals. All the visualization results below are accessible via the notebook “Scientific data_Figure.ipynb”26. Furthermore, the dataset on Figshare25 also presents the “raw_data” folder, where readers can find the initial and the output data generated from the subsequent sections. To enhance clarity regarding the operational disparities of the energy station when using different equipment, the 20-year operation data of the station is presented by dividing it into two periods: the first ten years and the last ten years. This distinction is necessary due to variations in equipment lifespan, with certain components lasting 10 years and others lasting 20 years.

Time-series energy loads

Firstly, we checked the hourly pattern of the dataset25. Using the data from sheet ‘All load1_2004&2018’ on Figshare25, the time-series variations of the electricity, heating, cooling, and hot water loads with respect to the outdoor air temperature can be visualized, as shown in Fig. 3. The electricity load of 2014 and 2018 show similar patterns, with close peak and valley values. The average difference in electricity load between the two years is within a narrow margin of 0.3%. Additionally, the space heating and cooling loads display sensitivity to ambient temperature changes. In the testing dataset, the heating and cooling area index of the classroom and office complex buildings are about 90 W/m2 and 60 W/m2, respectively. These observations are in line with the local technical manual for district heating and cooling in Japan, which suggests the heating area index of approximately 110 W/m2 and the cooling area index of around 80 W/m2. These consistent findings further validate the reliability of the dataset.

Fig. 3
figure 3

Time-series profiles of various energy loads and outdoor air temperature in 2014 and 2018.

Full size image

We plot Fig. 4 to check the average weekly distribution of electricity flows during the summer (June to August 2004) and winter (December 2004 to February 2005) periods. The daily electricity supply shows regular periodical changes that are closely related to occupancy times. Specifically, more electricity is consumed on weekdays compared to weekends, and this difference is more significant in winter. This descriptive analysis can provide more comprehensive insight into district energy system operations. The measured maximum hourly power outputs of fuel cell and gas engine systems are 195 kW and 161 kW, respectively. These values closely align with their nominal power capacities, thereby providing additional verification of the data reliability.

Fig. 4
figure 4

Average hourly and daily distributions of electricity resources in summer and winter weeks.

Full size image

Gas consumption pattern

This part presents the seasonal and daily gas consumption patterns. Using gas usage sheets, Fig. 5 can be plotted to show the gas consumption characteristics of different seasons of the first and second decades, with the following seasonal time frames: Spring (March to May), Summer (June to August), Autumn (September to November), and Winter (December to February of the following year). The energy station showed reasonable increases in city gas consumption in the summer and winter seasons due to increased space heating and cooling demand.

Fig. 5
figure 5

The seasonal variability of gas consumption, 2002–2011 (left), 2012–2021 (right).

Full size image

Similar to electricity consumption, we plot Fig. 6 to show the weekly pattern of gas usage in the summer (June to August 2004) and winter (December 2004 to February 2005) periods. Since the absorption chillers serve as the main cooling and heating source, it’s reasonable to observe from the data that the primary consumption of gas is taken by absorption chillers, while the gas engine and the fuel cell take less significant amount. Similar to electricity, gas consumption is considerably higher on working days compared to non-working days, which meets the working schedule on campus.

Fig. 6
figure 6

Average hourly and daily distributions of gas usage in summer and winter weeks.

Full size image

Indoor temperature distribution

To validate the reliability of our dataset, we conducted detailed analyses of room air temperature, supply/return water temperature and flow, and wall temperature of the research building for both a summer week and a winter week in 2004, as illustrated in Fig. 7. These parameters were chosen due to their critical role in evaluating the demand response of the HVAC system and thermal comfort to seasonal variations. For the summer week, the data displayed consistent diurnal patterns, with room air temperatures peaking during midday and dipping at night, closely following the supply/return water temperature and flow trends. Wall temperatures also exhibited expected thermal inertia, lagging slightly behind the air temperature changes. The coherence between these parameters indicates the effective operation of the HVAC system in maintaining indoor comfort. During the winter week, the room air temperature remained stable, reflecting the heating system’s capacity to counteract external cold. The supply/return water flow showed regular fluctuations corresponding to the heating cycles, while the wall temperature changes were more gradual, again demonstrating thermal inertia. The consistency and logical variation of these temperatures across different seasons confirm the dataset’s accuracy and reliability. Additionally, the data aligns well with expected HVAC performance metrics, providing further validation. These analyses support the authenticity of our long-term data collection and its utility for future research on building energy management.

Fig. 7
figure 7

Hourly and daily distributions of indoor temperature & supply/return water temperature and flow in summer and winter weeks.

Full size image

Energy consumption and supply over the 20-year period

We used the sheet of ‘Purchased grid and gas’ to draw the barplot of Fig. 8b, which can help to observe the cumulative electricity import and natural gas procurement. Further, we plot Fig. 8c,d to check the detailed electricity generation and gas consumption of by equipment. Using the sheet of ‘Electricity sources’, Fig. 8c is plotted to show sources of electricity supply, including PV, gas engine, fuel cell, and electricity imported. Electricity from PV generation contributes a relatively small proportion (2.3%), and the plotted data meets the fact of the limited life span of PV (15 years). The gas engine operated from 2002 to 2016 for a lifespan of 15 years, accounting for 8.3% of the total electricity load. The fuel cell operated from 2002 to 2011, contributing to 21.7% of the total electricity load. The remaining electricity load is fulfilled by the grid, constituting 67.8% of the total electricity load.

Fig. 8
figure 8

Graphical representations of various energy loads, electricity sources, and gas consumption over the 20-year period.

Full size image

Using the sheet of ‘Gas usage’, Fig. 8d can be plotted, which illustrates the distribution of gas consumption by devices, including boilers, gas engine, fuel cell, and absorption chillers. The fuel cell and gas engine are also utilized for space heating and domestic hot water through waste heat recovery. The gas consumptions for boilers and absorption chillers are observed to increase in the second decade, which meets the fact that the heat recovery from the fuel cell and gas engine was reduced due to equipment degradation.

Electricity resources disaggregation

Figure 9 describes the pattern and variability of daily electricity consumption and the electricity imported from the grid at 60-minute intervals during the first and second decades. The solid red line and the dotted green line in the figure represent the average electricity consumption and the average electricity imported from the grid, respectively. The color bands surrounding the lines indicate their standard deviations at specific times. It is evident that the maximum, minimum, and average hourly electricity consumption during the first and second decades remained almost at the same level. Notably, the cessation of operation for certain devices, namely the gas engine and fuel cell, in the second decade led to a substantial increase in imported electricity from the grid compared to the first decade. This finding supports the reliability of the dataset by illustrating the influence of specific operational changes on electricity consumption and the reliance on the grid.

Fig. 9
figure 9

Distributions of daily electricity load (2002–2011 (a), 2012–2021 (b)) and grid imported electricity (2002–2011 (c), 2012–2021 (d)).

Full size image

Usage Notes

Readers can initially download the datasets from Figshare25. As the size of each xlsx file is relatively modest, users can conveniently review the datasets in Excel, gaining a comprehensive understanding of the data. However, for in-depth analysis spanning over 20 years, it is recommended to utilize a programming language such as Python for enhanced data visualization and manipulation capabilities. The code necessary for processing and visualizing the data is accessible on GitHub (https://github.com/XiaoyuJIN97/Scientific_data_work_Kitakyushu_campus_building).