City short-term water consumption prediction method based on least square support vector machine model

Abstract

The invention provides a city short-term water consumption prediction method based on a least square support vector machine model. The method comprises the following steps that preprocessing is carried out on historical water consumption; correlation analysis is carried out; a least square support vector machine method is adopted for setting up a city short-term water consumption predicting model, and time sequence combinations of the historical water consumption with correlation coefficients larger than set values are selected to serve as a training sample set for training; the city short-term water consumption predicting model is adopted for carrying out prediction in real time; prediction errors are calculated, and if the prediction errors do not meet the prediction accuracy requirement, the city short-term water consumption predicting model is improved. According to the city short-term water consumption prediction method, preprocessing is carried out on the historical water consumption, an original change law is kept as much as possible, and therefore the prediction accuracy can be improved; as the least square support vector machine method is adopted, the problem of nonlinearity of a water supply system and the problem that an accurate model can not be set up are solved; weather data and / or holiday factors are considered comprehensively, and the prediction accuracy is improved.

Description

technical field [0001] The invention relates to a method for predicting urban water consumption, in particular to a method for predicting urban short-term water consumption based on a least square support vector machine model. Background technique [0002] With the acceleration of urbanization and the continuous increase of city size and population, the demand for production water and residential water is increasing. According to statistics from the Ministry of Water Resources, about two-thirds of China's 660 cities are facing a water resource crisis. It is estimated that by 2030, the country will consume 750 billion cubic meters of water resources, accounting for about 90% of the total available water resources. At the same time, China is also facing serious water pollution problems. According to the report of the China Geological Survey Bureau, 90% of the groundwater resources in the country are polluted, and 60% of them are seriously polluted. How to effectively guide an...

AI Technical Summary

Problems solved by technology

The time series forecasting method predicts future water consumption based on historical water volume data, which has obvious hysteresis and cannot reflect the influence of meteorological factors (temperature, precipitation, etc.)

When sudden changes in weather and other factors lead to large fluctuations in water consumption, it is impossible to track changes in water volume in time, resulting in poor forecasting results

However, the artificial neural network has the following disadvantages: 1) there is a risk of falling into a local extremum, resulting in a large training error; 2) the algorithm learning is based on the experience with the smallest risk, which makes the algorithm's training error the smallest, so the generalization ability of the established network model is not strong ; 3) The structural design of the network (such as the number of hidden layers and the number of nodes) depends on prior knowledge and experience

However, the existing forecasting techniques and methods have not been able to effectively solve the above problems.

Images