Distribution transformer load forecasting method based on mxnet framework deep neural network

Abstract

The invention discloses a distribution transformer load forecasting method based on a mxnet framework deep neural network, and relates to a distribution transformer load forecasting method. At present, there is no uniform method to characterize and describe the distribution transformer and a worker cannot fully understand the distribution transformer and cannot accurately forecast the load changetrend. The method includes the following steps: acquiring system internal data and external data; purifying the acquired data to obtain load-related index data and historical load data; and separatelyfitting a long-term load forecasting model and a short-term load forecasting model in the load by using an optimal combination prediction model and an artificial neural network algorithm and using acourt as a unit; according to a load forecasting result and an index dimension, extracting a load-related label system and building a view of the court; displaying portrait view on a human-machine interface. The method establishes a load forecasting assessment model to monitor load fluctuations, achieve continuity of forecasting, grasp the important characteristic-load dynamic change process of acommon transformer, and can forecast the load change trend.

Description

technical field [0001] The invention relates to a distribution transformer load forecasting method, in particular to a distribution transformer load forecasting method based on an mxnet framework deep neural network. Background technique [0002] Power load forecasting is of great significance to the production of power systems and the safe operation of power grids. Accurate load forecasting is an important basis for scheduling dispatch plans, power supply plans, and transaction plans in a market environment. [0003] Traditional load forecasting research often faces problems such as general objects, single data sources, and traditional forecasting methods. In the past, the load forecasting object was often an area with a large load value, and the random relationship differences between individuals would be offset by each other, so the fluctuations were relatively stable; the data source was single, and in the past it mainly relied on historical load data, or Combined with ...

AI Technical Summary

Problems solved by technology

[0003] Traditional load forecasting research often faces problems such as general objects, single data sources, and traditional forecasting methods.

In the past, the load forecasting object was often an area with a large load value, and the random relationship differences between individuals would be offset by each other, so the fluctuations were relatively stable; the data source was single, and in the past it mainly relied on historical load data, or Combined with meteorological data, the factors of electricity users are usually not considered; the prediction methods are biased towards more traditional methods, such as time series method, regression analysis method, etc., all have certain defects, and time series methods cannot consider complex factors such as meteorological data Under the influence of load fluctuations, the regression analysis method has the problem of how to determine the appropriate regression equation, which cannot deal with the unbalanced transient state between climate variables and loads; there are also studies on short-term load forecasting through neural network algorithms, but are limited The impact of computing power is usually a single-layer neural network, which does not give full play to the advantages of machine learning in effect, which leads to insufficient prediction accuracy of traditional load forecasting methods, and the accuracy often reaches the bottleneck and cannot be broken through

[0004] There are many related objects involved in the distribution change, and the data are scattered in different system environments. At present, there is no unified method to describe and describe the distribution change. The staff cannot fully understand the distribution change, and the hidden characteristics of the distribution change It is also impossible to understand and master, and it is impossible to accurately predict the trend of load changes

Images