Industrial electricity consumption prediction method based on improved multi-storage pool echo state network

Abstract

The invention discloses an industry electricity consumption prediction method based on an improved multi-storage pool echo state network, which expands the number of storage pool layers in the middle of an ESN to L (L = 1, 2, 3,..., M) to obtain an MRESN, thereby improving the accuracy of mass data prediction and enhance the feature extraction capability. In addition, for the defects of an original BAS algorithm, an IBAS algorithm is provided, wherein optimization of a longhorn beetle population is adopted, and an elite selection and retention strategy is added into iteration of longhorn beetle population search; and then a Levy flight strategy is introduced into a longhorn beetle search algorithm, and a self-adaptive step length strategy is adopted to update the current position of longhorn beetles. Therefore, the method can obtain a better data prediction effect than an original ESN network, and is more beneficial to the accurate prediction of the industry electricity consumption.

Description

technical field [0001] The invention relates to a network model prediction method, in particular to an industry power consumption prediction method based on an improved multi-storage pool echo state network. Background technique [0002] As electricity trading enters the marketization process, various industries with large electricity consumption directly participate in electricity market transactions in the form of "annual long-term association" and "monthly bidding". The electricity consumption situation of each industry is different and there is a deviation assessment of electricity consumption declaration. Whether it is "annual long-term association" or "monthly bidding", all industries must face the problem of electricity consumption declaration. At present, all provinces, autonomous regions and municipalities across the country have introduced systems to clarify the scope of deviation assessment for declared electricity. Accurate prediction of industry power consumpti...

AI Technical Summary

Problems solved by technology

[0004] However, the original ESN network still has some practical problems in load forecasting: since the storage pool in the original ESN network is randomly generated, the difference in the internal structure of the storage pool will greatly affect the performance of the prediction model; the key to the original ESN network The parameters are generally selected according to experience, which will limit the prediction performance of the model to a certain extent, and may not be able to adapt to the actual load forecasting problem

[0005] The original BAS algorithm also has defects. The selection of the search step size in the algorithm has a great impact on the optimization efficiency of the algorithm. As the number of iterations reaches a certain number, it is easy to fall into a local optimum

Images