A tensor completion algorithm for missing user data in spot trading of electricity market

The electricity spot market, a lack of electricity data disrupts the balance between supply and demand and makes it difficult to plan generation and supply. To solve this problem, this paper presents a tensor complementation algorithm that uses time series decomposition and considers the high dimensionality and significant fluctuations of electricity consumption data in the spot market. The method starts with the decomposition of time series data for individual users, followed by the construction of a Hankel tensor. A tensor regularization model based on parallel factorization is developed and solved using hierarchical alternating least squares (HALS) with gradient normalization to reduce computation time. The experiments were conducted using three different datasets. Using the relative recovery error as the evaluation metric, the results show a 12.7 % improvement in accuracy compared to tensor CP factorization for data with 60 consecutive missing entries, providing enhanced support for electricity spot trading decisions.