In recent years, the number of user-side energy storage systems has significantly increased, thanks to supportive policies for energy storage development. This growth has prompted a new area of research focused on how to effectively identify existing energy storage users and implement detailed management as part of load control in the new power system. This article aims to identify commercial and industrial user-side energy storage in the context of peak and valley arbitrage models, combining energy storage operation strategies with changes in load curves to establish a typical characteristic index system. It employs MiniBatch K-Means clustering, random forest feature selection, and multidimensional cross-iteration models to accurately identify existing energy storage users, thus aiding grid companies in the comprehensive management of user-side energy storage resources.
Challenges in Identifying User-Side Energy Storage
In April 2024, the National Energy Administration issued the notice titled “On Promoting the Grid Connection and Dispatching of New Energy Storage” (Guoneng Fa Keji [2024] No. 26). This aims to regulate the grid connection management of new energy storage and optimize its operational mechanisms, thereby enhancing the role of new energy storage in supporting the construction of a new power system. User-side energy storage offers significant cost reduction and efficiency improvement benefits. By shifting consumption to off-peak periods, it can effectively reduce electricity costs, and when combined with photovoltaics, it increases the uptake of renewable energy during the day while lowering carbon emissions. This not only promotes environmental protection and sustainable development but also balances grid loads and reduces the costs associated with expanding power supply capacity, offering practical value and significant potential for promotion.
However, early investments in user-side energy storage were primarily driven by enterprises, without mandatory grid management requirements. This has led to a lack of information for power companies regarding operational energy storage users, making it challenging to mobilize energy storage resources for grid interaction. Users are also unable to adjust their energy storage operation strategies in response to grid shortages and subsidy policies, limiting the economic value of energy storage. To reduce the difficulty of information verification, it is essential to analyze changes in users’ electricity consumption behaviors before and after the installation of energy storage systems and construct a model for identifying existing energy storage users. This will assist grassroots operations in efficiently conducting energy storage user investigations and support the grid in implementing detailed management of energy storage load resources.
Current Issues in Energy Storage User Identification
- Scarcity of Research Samples: The early rough management approach limited power companies’ control over energy storage users. On-site verification requires user cooperation and is time-consuming, leading to insufficient sample data for analysis. This complicates pattern recognition and affects the accuracy and generalization ability of the identification model.
- Complexity of Electricity Consumption Characteristics: Large users experience significant fluctuations in electricity load due to their business operations. Additionally, various factors interfere with electricity consumption characteristic analyses. For instance, when energy storage capacity is low, the load adjustment effects generated by energy storage can be overwhelmed by the users’ inherent consumption variations. Many users have already implemented demand response strategies to reduce electricity costs, making it challenging to distinguish between the load adjustment effects of energy storage and those of demand response. Furthermore, a significant proportion of large users have installed photovoltaic systems, which can lead to significant fluctuations in daytime loads and complicate the analysis of electricity consumption characteristics.
- High Flexibility of Energy Storage Charging and Discharging: Energy storage devices have controllable input and output power, but they may not always charge and discharge at a constant power level, displaying various charge and discharge profiles. Currently, power companies lack detailed monitoring of internal electricity consumption within enterprises, making typical load separation analysis infeasible. Thus, a comprehensive approach that considers both business operations and data performance is needed to develop effective solutions.
Proposed Solutions for Identifying User-Side Energy Storage
This article proposes a robust model for identifying commercial and industrial user-side energy storage based on a data amplification typical sample screening method. The specific solutions are as follows:
- Data Amplification and Typical Sample Screening: To address the issue of missing device-level load data, this strategy combines typical sample screening with incremental optimization to continuously expand the energy storage sample database, dynamically adjusting characteristic rules and thresholds to enhance the accuracy of the energy storage load identification model.
- User Clustering Using MiniBatch K-Means: By conducting detailed clustering analysis of sample users’ load curves, this method identifies typical user groups, providing direction for building a user-side energy storage identification index system and improving the accuracy and relevance of the analysis.
- Feature Selection and Optimization with Random Forest: Utilizing a random forest model, this approach selects and optimizes features based on a unified data source, extracting key indicators and establishing a comprehensive index system to enhance the model’s adaptability and generalization capability.
- Multidimensional Cross-Iteration Model Optimization: This model continuously optimizes the combination of indicators and thresholds based on identification results, enhancing recognition accuracy through iterative improvements and making the model more applicable for broader implementation.
Conclusion
This article focuses on the precise identification of user-side energy storage by employing clustering algorithms and machine learning classification techniques to accurately identify energy storage users outside the current regulatory framework. Through in-depth analysis of users’ electricity load curves, an energy storage user identification model has been established to improve identification accuracy, allowing power companies to gain a comprehensive understanding of user-side energy storage distribution. Future work will concentrate on developing models for potential energy storage investors and calculating returns, analyzing existing user characteristics and profit models to identify high-yield potential users and providing configuration recommendations to expand the energy storage market.
