YUSHU-AI/HVAC-simulate-time-series-data

---
license: mit
---
## Dataset Details

### Dataset Description

The dataset is constructed based on the data dictionary released by Persagy in 2021 ( https://buildingdtp.com ), which defines objects and relationships in building operation systems. This dictionary includes 58 building types, 353 space types, 85 categories of mechanical and electrical systems, 558 equipment types, 95,000 attributes, and 81 relationship types. The dataset primarily focuses on HVAC-related objects and consists of both real-world empirical data and synthetic data generated through a simulation environment.

Simulation Data: The main dataset currently provided is based on 12 real-world project EnergyPlus simulation models. These simulations systematically vary 18 key influencing factors of building systems, including building location (city), building orientation, wall thermal transmittance, operating hours, source-side equipment capacity, distribution equipment capacity, and terminal equipment capacity, among others. This process generates 5,000 simulation scenario cases. We currently provide a subset of 50 simulation cases as an open dataset.

Real Project Data: The dataset includes operational data from 1,045 real-world building projects, covering subsystems such as energy management, security surveillance, facility management, and building automation. Among them, 508 projects contain relatively comprehensive information. Due to privacy concerns, the real-world dataset requires data anonymization , and a subset of the dataset will be open-sourced in the future.

This dataset supports machine learning applications in building operation systems, enabling research in energy efficiency optimization, predictive maintenance, and intelligent control strategies.

## Dataset Structure

### Data Organization
```
data/
├── sample_0001/
│ ├── graph/
│ │ ├── ACATACATFC-serve-building.csv
│ │ └── ...
│ └── timeseries/
│ ├── ACATACATAH.parquet
│ └── ...
├── sample_0002/
│ ├── graph/
│ └── timeseries/
└── ...
└── sample_0050/

- Each `sample_XXXX` folder represents an individual sample.
- Each sample contains two subdirectories:
- `graph/`: Contains CSV files representing adjacency matrices.
- File naming convention: `{source_node}-{edge_type}-{target_node}.csv`
- `timeseries/`: Contains time series data in Parquet format.
```

许可证：MIT
---
## 数据集详情

### 数据集描述
本数据集基于Persagy公司2021年发布的数据字典（https://buildingdtp.com）构建，该字典定义了建筑运维系统中的对象与关联关系。该字典涵盖58种建筑类型、353种空间类型、85类机电系统、558种设备类型、95000个属性以及81种关联关系。本数据集主要聚焦于暖通空调（HVAC）相关对象，包含真实实测数据与仿真环境生成的合成数据两类。

#### 仿真数据
当前提供的主数据集基于12个真实建筑项目的EnergyPlus仿真模型构建。本次仿真系统地调整了建筑系统的18项关键影响因素，涵盖建筑所在地（城市）、建筑朝向、墙体传热系数、运行时长、冷热源侧设备容量、输配侧设备容量以及末端设备容量等。该过程共生成5000个仿真场景案例，本次开源仅开放其中50个仿真案例子集。

#### 真实项目数据
本数据集包含1045个真实建筑项目的运维数据，覆盖能源管理、安防监控、设施管理以及建筑自动化等子系统。其中508个项目的信息相对完备。考虑到隐私保护需求，真实数据集需进行数据匿名化处理，未来将开源该数据集的子集。

本数据集可支撑建筑运维系统中的机器学习应用，助力能效优化、预测性维护以及智能控制策略等方向的研究。

## 数据集结构

### 数据组织

data/
├── sample_0001/
│ ├── graph/
│ │ ├── ACATACATFC-serve-building.csv
│ │ └── ...
│ └── timeseries/
│ ├── ACATACATAH.parquet
│ └── ...
├── sample_0002/
│ ├── graph/
│ └── timeseries/
└── ...
└── sample_0050/

- 每个`sample_XXXX`文件夹代表一个独立样本。
- 每个样本包含两个子目录：
- `graph/`：存储表示邻接矩阵的CSV文件。
- 文件命名规则：`{源节点}-{边类型}-{目标节点}.csv`
- `timeseries/`：存储Parquet格式的时序数据。