HEBR

This project implements the method of hierarchical electricity-theft behavior recognition proposed in [1], which is an algorithm for detecting electricity theft.

Testing

This project is implemented in Python 3.6

Dependency:

Python 3.6. Version 3.6.2 has been tested.
Tensorflow. Version 1.2.0 has been tested. Note that the GPU support is encouraged as it greatly boosts training efficiency.
Other Python modules. Some other Python module dependencies are listed in requirements.txt, which can be easily installed with pip pip install -r requirements.txt

Testing the Project:

python run.py -f ./data/ -o ./data/result.csv

Usage

Given an array of multi-source time series data, including users' electricity records, non-technical losses in transformer area, temperature records, this program can be used to estimate the electricity-theft probabilities for users.

Input Format

The input files are expected to be four parts:

(1) Electricity Usage Records: a CSV file, which contains daily records for every user: #userID, #areaID, #date, #total electricity, #top electricity, #on-peak electricity, #flat electricity, #off-peak electricity.

(2) Non-Technical Loss Records: a CSV file, which contains daily records for every area: #areaID, #date, #cost electricity, #billed electricity, #lost electricity. Here, the areaID corresponds to the areaID in (1)

(3) Temperature Records: a CSV file, which records daily weather, containing #areaID, #data, #high temperature, #low temperature. We can spider the data by the code in ./data_factory/temperature_spider.py

(4) Labels of Electricity Thieves: a CSV file, which records the time that the electricity thieves are caught, containing #userID, #areaID, #date.

Output Format

The program outputs to a file named result.csv which contains the results of electricity-theft probabilities estimated by HEBR.

Main Script

The help of main script can be obtained by excuting command:

python run.py -h
usage: run.py [-h] [-f DATA_FILE] [-o OUTPUT_FILE] [-hl HISTORY_LENGHT]
              [-b BATCH_SIZE] [-e NUM_EPOCH] [-n CPU_JOBS] [-g GPU_ID]
              [-l LEARNING_RATE] [-iu USER_DIMS] [-il NTL_DIMS]
              [-ie CLIMATE_DIMS] [-iul USER_NTL_DIMS] [-iue USER_CLIMATE_DIMS]
              [-iule USER_NTL_CLIMATE_DIMS]

optional arguments:
  -h, --help            show this help message and exit
  -f DATA_FILE, --data_file DATA_FILE
                        path of input file
  -o OUTPUT_FILE, --output_file OUTPUT_FILE
                        path of output file
  -hl HISTORY_LENGHT, --history_lenght HISTORY_LENGHT
                        the historical length for observed data, default value
                        is 180
  -b BATCH_SIZE, --batch_size BATCH_SIZE
                        the number of samples in each batch, default value is
                        1000
  -e NUM_EPOCH, --num_epoch NUM_EPOCH
                        number of epoch, default value is 100
  -n CPU_JOBS, --cpu_jobs CPU_JOBS
                        number of cpu jobs, default value is maximum number of
                        cpu kernel
  -g GPU_ID, --gpu_id GPU_ID
                        index of gpu, default value is 0
  -l LEARNING_RATE, --learning_rate LEARNING_RATE
  -iu USER_DIMS, --user_dims USER_DIMS
                        dimension of micro-level memory matrix, default value
                        is 16
  -il NTL_DIMS, --ntl_dims NTL_DIMS
                        dimension of meso-level memory matrix, default value
                        is 4
  -ie CLIMATE_DIMS, --climate_dims CLIMATE_DIMS
                        dimension of macro-level memory matrix, default value
                        is 8
  -iul USER_NTL_DIMS, --user_ntl_dims USER_NTL_DIMS
                        dimension of user-area memory matrix, default value is
                        64
  -iue USER_CLIMATE_DIMS, --user_climate_dims USER_CLIMATE_DIMS
                        dimension of user-climate memory matrix, default value
                        is 64
  -iule USER_NTL_CLIMATE_DIMS, --user_ntl_climate_dims USER_NTL_CLIMATE_DIMS
                        dimension of user_ntl_climate memory matrix, default
                        value is 256

Reference

[1] Wenjie, H; Yang, Y; Jianbo, W; Xuanwen, H and Ziqiang, C, 2020, Understanding Electricity-Theft Behavior via Multi-Source Data, In WWW, 2020

 @inproceedings{hu2020theft, 
    title={Understanding Electricity-Theft Behavior via Multi-Source Data},
    author={Wenjie Hu and Yang Yang and Jianbo Wang and Xuanwen Huang and Ziqiang Cheng},
    booktitle={Proceedings of WWW},
    year={2020}
    }

zjunet / HEBR