This paper proposes an electricity price prediction methodology and computational framework that integrates time series prediction algorithms and cloud native architecture for the multi-dimensional characteristics of the electricity spot market. The operation mechanism and tariff formation mechanism of three types of electricity market models, namely, pool-type, bilateral-type and hybrid-type, are systematically analyzed, and the normal distribution test is used to quantify the tariff distribution law. The dynamic similar subsequence prediction model is proposed, and the error correction mechanism is constructed by the time window optimization algorithm. We design a cloud-native edge computing framework for the Internet of Things (IoT), which solves the real-time computing and data security problems in resource-constrained scenarios at the edge in a multi-dimensional way. Building the arithmetic example, the proposed model improves significantly in prediction accuracy compared with traditional ARIMA and ANN methods, and the average absolute error is reduced to 0.249, which has the best prediction results.
