TL;DR
This paper proposes a novel method to assess noise discomfort between neighbors in apartment buildings by analyzing electricity consumption data. By clustering residents based on their appliance usage patterns and modeling the time differences in their electricity use, the study offers a data-driven approach to quantify and potentially mitigate a common source of urban living annoyance.
What Problem Does This Paper Address?
Living in close proximity, especially in multi-unit dwellings like apartment buildings, often leads to inter-neighbor conflicts arising from noise. A significant portion of this noise originates from household electrical appliances (e.g., hair dryers, vacuum cleaners, refrigerators). While the impact of noise on well-being is recognized, accurately quantifying noise discomfort (ND) between neighbors, particularly from appliance usage, has been challenging. Traditional methods like surveys or interviews are subjective, prone to recall bias, and time-consuming. This paper aims to bridge this gap by leveraging readily available electricity consumption data from advanced metering infrastructure (AMI) to objectively assess ND.
Methods
The proposed method involves four key steps:
- Data Collection and Preprocessing: Electricity consumption data (time, active power) from individual residential units are collected and cleaned. This involves identifying and excluding "empty" units or those with residents moving in/out during the study period. Data are aggregated into hourly intervals to capture usage patterns.
- Residential Units Clustering: A time series K-means clustering algorithm is employed to group units with similar electricity consumption patterns. This step identifies distinct lifestyle groups within the building.
- Noise Discomfort (ND) Modeling: A model is developed based on three key features extracted from the aggregated consumption data for each cluster: peak time (when consumption is highest), start time (when consumption begins to rise), and end time (when consumption returns to baseline). The model quantifies ND by summing the time differences in these features between neighboring units. Initial assumptions simplify the model by considering only adjacent units and assuming uniform ND propagation.
- ND Evaluation: The developed model is used to calculate the ND level for each unit and for the entire building. The paper also explores variations of the model by relaxing initial assumptions (e.g., considering diagonal neighbors or varying ND impact based on direction) to assess their influence on the results.
Datasets
The study utilizes electricity consumption data collected from 75 units in a 15-story campus apartment building at POSTECH, South Korea. The data were collected over four months (February 18 to June 7, 2019), focusing on weekday hourly data. After preprocessing, 55 units were used for the analysis. No new benchmark datasets were introduced; the focus was on applying the method to a real-world scenario.
Key Insights and Ideas
- Electricity data as a proxy for noise: The core idea is that the timing of electricity consumption directly reflects the usage of electrical appliances, which are a primary source of household noise. By analyzing these patterns, we can infer potential noise-related conflicts.
- Clustering reveals lifestyle groups: Grouping residents by their electricity usage patterns effectively captures different lifestyle habits, which are crucial for understanding how appliance usage might overlap and cause discomfort.
- Time differences are key: The model's novelty lies in quantifying ND based on the differences in appliance usage times between neighbors, rather than just the presence of noise. Larger time gaps in peak, start, or end times of appliance usage are hypothesized to lead to greater ND.
- Model flexibility: The paper demonstrates that the ND assessment model can be adapted by relaxing assumptions, allowing for more nuanced assessments that account for factors like diagonal neighbors or directional noise propagation.
Key Results
- The 55 analyzed units were clustered into three groups with distinct electricity consumption patterns.
- Under the original assumptions, the total daily ND for the building was calculated to be 304. The average ND per household was 5.53.
- The largest ND level experienced by a single unit was 16 per day.
- Relaxing Assumption 1 (considering diagonal neighbors) significantly increased the total ND to 528 per day.
- Relaxing Assumption 2 (allowing for directional weighting of ND) reduced the total ND to 246.4 per day, suggesting that noise impact can vary based on direction (e.g., noise from above/below might be less impactful than from adjacent sides).
- The study also explored an optimization scenario for household assignment to units, demonstrating that strategic placement could potentially reduce total ND to zero.
Limitations and Future Work
- Lack of empirical validation: The authors acknowledge that the primary limitation is the absence of empirical validation. The ND assessed by the model has not been directly compared with residents' subjective evaluations or actual noise measurements. Future work should include follow-up studies with questionnaires and experimental setups.
- Advanced data aggregation and clustering: The paper suggests exploring more advanced time series clustering methods and diverse aggregation intervals (e.g., 1-min to 1-month) to capture more granular consumption patterns and seasonality.
- Incorporating appliance-specific data: The current model relies on general electricity usage. Future work could integrate non-intrusive appliance load monitoring to identify specific appliances and their noise levels, leading to more accurate ND assessment.
- Quantitative noise data: Including actual noise decibel data alongside electricity consumption would significantly strengthen the model's accuracy.
- Reward models for alleviation: The paper proposes combining ND assessment with reward systems (e.g., reduced electricity bills) to incentivize residents to adjust their appliance usage patterns and reduce noise.
- Dynamic ND assessment: The model assumes linear increases in ND with time differences. Future research could explore non-linear relationships, especially considering that noise late at night might be more disruptive.
Summary updated