Oil chromatography data together with its variation trend provide the key basis for evaluation of transformer health state. The existing studies on deep belief network (DBN) and support vector machine (SVM) have reported a few results in field oil prediction. However, above-mentioned methods are proposed classification problem, so there is no theoretical applying to time-series wrong usage limits accuracy predicted results, which observed as obvious "time-shift" error prediction curve,...

In this paper, a Comprehensive Diagram Method (CDM) for Multi-Layer Perceptron Neuron Network (MLPNN) is proposed to realize natural gas energy metering using temperature, pressure, and the speed of sound from an ultrasonic flowmeter. Training testing MLPNN model were performed on basis 1003 real data points describing compression factors (Z-factors) calorific values three main components in Sichuan province, China. Moreover, 20 days tests conducted verify measurements' accuracy adaptability...

The stochastic characteristics of wind power and photovoltaic (PV) make the resource allocation system difficult. Therefore, it is necessary to consider correlation between generation PV stations avoid waste guarantee supply, while traditional analysis method cannot accurately describe multiscale time-varying correlation. In this article, based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), time series was decomposed into components. Moreover, time-dependent...

Daily electricity consumption forecasting is a classical problem. Existing algorithms tend to have decreased accuracy on special dates like holidays. This study decomposes the daily series into three components: trend, seasonal, and residual, constructs two-stage prediction method using piecewise linear regression as filter Dilated Causal CNN predictor. The specific steps involve setting breakpoints time axis fitting model with one-hot encoded information such month, weekday, For challenging...

With the development of economy and improvement living standards people, electricity consumption around world has increased dramatically. However, load that contains many components with different characteristics is affected by factors. How to classify extract improve accuracy forecasting become a focus attention. Based on this, this paper proposes using Complete Ensemble Empirical Mode Decomposition Adaptive Noise (CEEMDAN) decompose get multiple time scales. The hidden each are analyzed...

Power transformer is one of the key electrical apparatus in power system, fault could be warned and identified based on time series dissolved gas oil acquired by on-line monitoring technology Dissolved Gas Analysis (DGA) real time. More or less data will cause result inaccurate, when using DGA to analysis. Therefore, this paper proposed an optimal length selection method phase space reconstruction. Firstly, Lyapunov exponent used verify whether chaotic, delay coordinate reconstruct for...

This study aims at improving the forecast accuracy of primary energy consumptions in China, Japan and South Korea verifying correlation among neighboring countries. Considering diversity composition, this selects 6 components energy, including oil, coal, natural gas, nuclear hydropower renewable as characteristic variables. A collaborative prediction model based on SVR for consumption is proposed to explore three countries Korea. The results show that there a strong between when multiple...

With the rapid development of renewable energy such as wind power and solar popularization electric vehicles, operation planning active distribution network needs to consider consequent uncertainties. To solve this problem, a generation algorithm typical scenario based on Wasserstein probability distance is proposed. The first transforms continuous density functions / vehicle output at single time into discrete quantiles containing precise information through index. Then, considering amount...

Aiming at the problem of low utilization efficiency energy storage system in renewable power station, an optimal dispatching strategy station is proposed based on combination Long Short-Term Memory (LSTM) neural network and multi-stage decision-making optimization. Specifically, LSTM used to find regularity long time series data, rolling forecasting intraday electricity price market carried out solve long-term dependency traditional algorithm forecasting. On this basis, scheduling decomposed...

In order to improve the economy of electricity sales photovoltaic power plants equipped with energy storage system, this paper proposes an optimization sale strategy which takes into account prediction price peak time and fluctuation. First, according historical data, is predicted. Then, a long short term memory neural network used predict at off-peak times. With goal maximizing revenue, method are established. Finally, generation data plant in Australia conduct simulation test. After...

Abstract To improve the utilization of new energy sources and reduce economic costs, a micro-grid dispatching model based on prediction with P2G is proposed. This paper consists two stages. The first stage to optimize support vector machine using particle swarm algorithm predict data wind power photovoltaic generation several influence factors. second use PV output in put them into this for optimal dispatching. solved by CPLEX+YALMIP divided three scenarios comparative analysis. conclusion...