Search

KR-20260064094-A - Two-stage optimal bidding method for load aggregator and apparatus thereof

KR20260064094AKR 20260064094 AKR20260064094 AKR 20260064094AKR-20260064094-A

Abstract

A two-stage optimal bidding method and apparatus for a load aggregator are disclosed. The two-stage optimal bidding method for a load aggregator comprises: (a) a step of collecting historical multivariate time series data, wherein the historical time series data includes market prices and local loads; (b) a step of preprocessing the multivariate time series data and applying it to an LSTM-autoencoder-based prediction model to predict the total load and market price; (c) a step of applying each consumer data to each agent model to statistically analyze individual total load consumption and mobile and controllable load consumption to derive individual power consumption patterns and demand response participation potential, respectively; and (d) a step of determining a bidding strategy that maximizes profit in the market using the predicted market price, total power consumption, individual power consumption patterns, and demand response participation potential.

Inventors

  • 김문겸

Assignees

  • 중앙대학교 산학협력단

Dates

Publication Date
20260507
Application Date
20241031

Claims (12)

  1. (a) A step of collecting historical multivariate time series data—the historical time series data includes market prices and local loads; (b) a step of preprocessing the above multivariate time series data and applying it to an LSTM-autoencoder-based prediction model to predict the total load and market price; (c) a step of applying each consumer data to each agent model to statistically analyze individual total load consumption and mobile and controllable load consumption to derive individual power consumption patterns and demand response participation potential, respectively; and (d) A two-step optimal bidding method for a load aggregator comprising the step of determining a bidding strategy that maximizes profit in the market using the above-mentioned predicted market price, total power consumption, individual power consumption patterns and potential for participation in demand response.
  2. In Article 1, The above step (b) is, A two-stage optimal bidding method for a load aggregator, characterized by preprocessing the above multivariate time series data to remove outliers and interpolate missing values, performing Principal Component Analysis (PCA) to extract features, and applying them to the above LSTM-autoencoder-based prediction model.
  3. In Article 1, A two-stage optimal bidding method for a load aggregator, characterized by generating a random value after Box-Cox transforming each of the above consumer data, deriving an individual total load consumption by inverting the random value, and deriving the movable and controllable load consumption by analyzing the gamma distribution.
  4. In Article 1, The above agent model calculates the potential for demand response participation for each agent's attribute by applying basic rules, but A two-stage optimal bidding method for a load aggregator, characterized by adjusting consumption patterns by reflecting the above basic rule in the price.
  5. In Article 1, A two-stage optimal bidding method for a load aggregator, characterized in that the above bidding strategy is determined by considering revenue, purchase costs, penalty costs, and power balance constraints, and is determined such that the difference between the actual load value and the expected load value is minimized.
  6. In Article 5, A two-stage optimal bidding method for a load aggregator, characterized in that the above bidding strategy is derived using the following mathematical formula. Here, represents profit, represents the purchase cost, and represents the penalty cost, represents the power balance constraint, and represents the amount of electricity sold to consumers and the retail price, and It represents the amount of electricity purchased in the market and the electricity market price from the previous day. and represents the predicted and actual values of the reference load, and and represents the price of the preliminary requirement and the penalty factor of the load.
  7. A computer-readable recording medium having recorded program code for performing a method according to any one of paragraphs 1 through 6.
  8. Memory storing at least one instruction for a two-stage bidding method; and It includes a processor that executes instructions stored in the memory above, The instructions executed by the above processor are, respectively, (a) A step of collecting historical multivariate time series data—the historical time series data includes market prices and local loads; (b) a step of preprocessing the above multivariate time series data and applying it to an LSTM-autoencoder-based prediction model to predict the total load and market price; (c) a step of applying each consumer data to each agent model to statistically analyze individual total load consumption and mobile and controllable load consumption to derive individual power consumption patterns and demand response participation potential, respectively; and (d) A computing device characterized by performing a step of determining a bidding strategy that maximizes profit in the market using the above-mentioned predicted market price, total power consumption, individual power consumption patterns and potential for participation in demand response.
  9. In Article 8, The above step (b) is, A computing device characterized by preprocessing the above-mentioned multivariate time series data to remove outliers and interpolate missing values, performing principal component analysis (PCA) to extract features, and applying them to the above-mentioned LSTM-autoencoder-based prediction model.
  10. In Article 8, A computing device characterized by generating a random value after Box-Cox transforming each of the above consumer data, deriving an individual total load consumption by inverting the random value, and deriving the movable and controllable load consumption by analyzing the gamma distribution.
  11. In Article 8, The above agent model calculates the potential for demand response participation for each agent's attribute by applying basic rules, but A computing device characterized by adjusting consumption patterns by reflecting the above basic rule in the price.
  12. In Article 8, A computing device characterized by the fact that the above bidding strategy is determined by considering revenue, purchase costs, penalty costs, and power balance constraints, and is determined such that the difference between the actual load value and the expected load value is minimized.

Description

Two-stage optimal bidding method for load aggregator and apparatus thereof The present invention relates to a two-stage optimal bidding method and apparatus for a load accumulator combining long-term memory-based prediction and an agent-based model. Demand response (DR) programs induce changes in end-user consumption patterns in response to market price signals. As a flexible demand-side resource, DR is considered an efficient tool for maintaining the reliability of the power system and addressing the issue of large-scale penetration of distributed energy resources, such as renewable energy sources. Residential consumers account for a significant proportion of electricity consumption among all end-users and can provide important opportunities for implementing DR programs. However, designing an efficient DR mechanism in the residential sector presents a significant challenge due to the large number of consumers and the minimal impact of individual consumers on the market. Despite these limitations, load aggregators have emerged as a means of participating in the wholesale electricity market on behalf of individual residential consumers. However, the emergence of DR and load aggregators has presented new challenges to the entire chain, including market operators, load aggregators, and residential consumers. Increased participation in DR programs complicates power flow in the demand-side network, affecting power consumption patterns and load forecasting. For load aggregators, accurately quantifying and estimating the impact of DR programs is crucial, as it is essential when submitting bidding strategies in the next-day market. In other words, before trading in the next-day market, load aggregators must understand the rate of change in total load due to DR; however, the behavior of residential consumers is heterogeneous. Therefore, even with similar dwelling sizes, occupancy rates, appliance sets, and geographical conditions, energy consumption can vary by up to 200%. Furthermore, sensitivity to DR signals can vary depending on consumer behavior and lifestyle. For this reason, load aggregators struggle to understand and account for these individual characteristics that determine electricity consumption. The advancement of smart meters is transforming power systems into smart power networks, and the collected data can provide important clues for understanding end-user behavior. In other words, the extensive development of smart meters enables the collection of vast amounts of granular electricity consumption data, and this high-resolution data offers rich information regarding consumers' electricity consumption patterns and lifestyles. Through data learning and analysis, this rich data provides an opportunity to objectively quantify and estimate the degree of load fluctuation under DR programs. From another perspective, existing DR research, which assumes that end users are always rational and active economic agents, leads to unexplained discrepancies between modeling results and actual observations. For example, some price-responsive loads may not consistently change their electricity consumption behavior in response to price signals. Furthermore, the degree of change may differ from expected values. Particularly in situations where consumers are autonomous, their irrational decision-making must also be taken into account when estimating the impact of DR. FIG. 1 is a flowchart illustrating a two-stage bidding method according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an LSTM autoencoder model according to an embodiment of the present invention. FIG. 3 is a diagram illustrating a D-ABM bottom-up model approach according to an embodiment of the present invention. FIG. 4 is a diagram illustrating a D-ABM approach according to an embodiment of the present invention. FIG. 5 is a diagram illustrating a daily electricity consumption profile affected by DR according to an embodiment of the present invention. FIG. 6 is a drawing illustrating a two-stage bidding framework according to an embodiment of the present invention. FIG. 7 is a diagram illustrating the interaction relationship between a load aggregator and each business entity according to an embodiment of the present invention. FIG. 8 is a block diagram schematically illustrating the internal configuration of a computing device according to one embodiment of the present invention. As used in this specification, singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "composed" or "comprising" should not be interpreted as necessarily including all of the various components or steps described in the specification, and should be interpreted as meaning that some of the components or steps may be excluded, or that additional components or steps may be included. Furthermore, terms such as "...part," "module," etc., as used in the specification refer to a unit