The realization of short-term load forecasting is the basis of system planning and decision-making, and it is an important index to evaluate the safety and economy of power grid.In order to accurately predict the power load under the influence of many factors, a new short-term power load prediction method based on fuzzy support vector machine and similar daily linear extrapolation is proposed, which combinesthe method of fuzzy support vector machine and linear extrapolation of similar days. The method first selects similar days according to the effect of integrated weather and time on load. Then the fuzzy membership of the training sample is obtained by the normalization processing, and the daily maximum and minimum load is predicted by the fuzzy support vector machine. Finally, the load prediction value is obtained by combining the load trend curve obtained by the similar daily linear extrapolation method. and this method is feasible and effective for short-term forecasting of power load.

Through the effective word vector training method, we can obtain semantic-rich word vectors and can achieve better results on the same task. In view of the shortcomings of the traditional skip-gram model in coding and modeling the processing of context words, this study proposes an improved word vector-training method based on skip-gram algorithm. Based on the analysis of the existing skip-gram model, the concept of distribution hypothesis is introduced. The distribution of each word in the word context is taken as the representation of the word, the word is put into the semantic space of the word, and then the word is modelled, which is better modelled by the smoothing of words and the semantic space of words. In the training process, the random gradient descent method is used to solve the vector representation of each word and each Chinese character. The proposed training method is compared with skip gram, CWE + P , and SEING by using word sense similarity task and text classification task in the experiment. Experimental results showed that the proposed method had significant advantages in the Chinese-word segmentation task with a performance gain rate of about 30%. The method proposed in this study provides a reference for the in-depth study of word vector and text mining.

A deep learning speech enhancement algorithm based on dynamic hybrid feature and adaptive mask and DSP implementation is proposed in this paper, which solves the problem of feature loss and improves the performance of speech enhancement. The dynamic features incorporate the log Mel power spectrum, Mel cepstral coefficients, and Multiresolution Auditory Cepstral Coefficients (MRACC) and capture the speech transient information by deriving the derivatives to comprehensively represent the nonlinear structure of speech and reduce distortion. To make the system improve the speech quality while reducing the speech distortion as much as possible, a soft mask that can be adaptively adjusted considering the signal-to-noise ratio information is proposed, which can be automatically adjusted according to the different speech signal-to-noise ratio information to obtain the mask value under the corresponding signal-to-noise ratio conditions, and phase difference information that can improve the speech intelligibility is incorporated in it. Then, an improved deep neural network model is designed to effectively improve the speech enhancement performance. Finally, the hardware and algorithm software design of the DSP-based speech enhancement system is given. Experimental simulations are carried out for multiple voices in different noise backgrounds. The experimental results indicate that the performance indexes of the proposed method are significantly improved compared with the existing speech enhancement methods, which verifies the feasibility and superiority of the proposed method.

For complexity and efficiency of the multi-objective optimization, we proposed the mobile distance field-driven adaptive crowd optimization algorithm. In space, we modify the surface parameters based on the corresponding changes of the distance field to obtain the moving target's moving track and moving surface. When the curve of the moving track is changed, the x axis and the y axis of the moving track are adjusted adaptively. In this paper, the moving process is divided into three processes: the target dynamic crowd control, the crowd model algorithm, and the predictive control of linear time domain based on the moving target prediction and crowd control algorithm. Then, the multi-objective optimization algorithm of moving objects is proposed by using the crowd model to predict the status and the position of the target. The experimental results show the high accuracy, low complexity, and high efficiency of the proposed optimization algorithm.