Application of IBM SPSS Modeler in power load forecasting

Power load forecasting is one of the important tasks in the management of power system dispatching, electricity, plan and planning. To improve the level of load forecasting is beneficial to plan power management, to rationally arrange power system operation mode and unit overhaul plan, to benefit coal, fuel economy and reduce generation cost, to make reasonable power grid construction plan, and to improve economic and social benefits of power systems. Therefore, the load forecasting has become one of the important contents to realize power system management and modernization.

Introduction to IBM SPSS Modeler

IBM SPSS Modeler is a set of data mining tools that can be used to quickly build predictive models using computer technology and apply them to business activities to improve the decision-making process.

IBM SPSS Modeler provides a variety of modeling methods that rely on machine learning, artificial intelligence, and statistics. By modeling the methods in the palette, you can generate new information based on data and develop forecast models. Each method has its own strengths and is suitable for solving specific types of problems.

CRISP-DM Process Model

IBM SPSS Modeler is designed by reference to the industry standard CRISP-DM model to support the entire data mining process from data to better business results.

The common CRISP-DM process model includes six phases to solve the problem of data mining. These six phases are fitted in a circular process designed to apply data mining to larger business practices.

Business Understanding : Identify business objects, assess situations, identify data-mining objectives, and develop engineering plans.

Data Understanding : Collecting initial data, describing data, exploring data and validating data quality.

Data Preparation : Selecting, cleaning, building, integrating data, and formatting data.

Modeling: Selecting modeling techniques, generating test designs, and building and evaluating models.

Evaluation : Evaluate the results, view the data mining process, and determine the next steps.

Deployment : Plan deployment, monitor and maintain, generate final reports, and review the project.

Figure 1. CRISP-DM model

IBM's rich model of SPSS Modeler provides support for power load forecasting

IBM SPSS Modeler Data Mining tool, which provides a variety of data mining algorithms, supporting the complete process of data mining, to power load forecasting, can effectively improve the accuracy and timeliness of load forecasting.

Time Series model

A time series is an ordered set of measured values collected at regular intervals, such as daily stock prices or weekly sales figures. The time series modeling method assumes that history always repeats itself--even if it's not exactly the same--that it's very close enough to make better decisions about the future by studying the past.

The time series model can be divided into exponential smoothing model and synthetic autoregressive moving average (ARIMA).

Exponential Smoothing Model : A predictive method that uses the weighted values of previous sequence observations to predict future values. Therefore, the exponential smoothing is not based on the theoretical understanding of the data.

ARIMA Model : Provides a more mature approach to modeling trends and seasonal components than an exponential smoothing model, and in particular, increases the advantages that can be included in the model with independent variables (predictive variables). This includes the explicit designation of the autoregressive order and the moving average order and the difference number. You can include predictor variables and define transformation functions for any or all predictive variables and specify automatic detection or precise settings for outlier values.