การพัฒนาระบบสนับสนุนการตัดสินใจเพื่อการจัดการน้ำท่วมในลุ่มน้ำกว๊านพะเยา

Abstract

The objectives of the study are to develop DSS (Decision Support System) to solve flooding in the area of Kwan Phayao by using water balance. The data is collected by using a rainfall-runoff model to measure the amount of water flowing in Kwan Phayao. Furthermore, the study aims to forecast the watershed runoff around the risk area by using an artificial neural network. The case focused on Mae Tum; the main river that flows through the municipal area and ends up in Kwan Phayao. IF AS program on two layers tank model involves the development of DSS for measuring the amount of water from tributary entered to Kwan Phayao. The main input is the amount of rainfall and physical area. According to the study, Kwan Phayao River has split into six sub-basins. The runoff data from Ing river, Mae Tam, and Mae Tum; the branches near a gauging station, is used to compare with the IF AS program on two layers tank model. The Parameter value is adjusted depending on the nature of the land; for example, SKF (Final Infiltration Capacity), SNF (Roughness Coefficient of Ground Surface), and AUD (Regulation Coefficient of Rapid Intermediate Outflow).The proper parameter is considered by R2 (or 0.931, 0.896, and 0.875). Due to the development of DSS (Decision Support System), there are four plans of solutions in the condition of the physical environment of Kwan Phayao and the amount of water calculated via the model. Furthermore, flood forecasting in the risk area by using an artificial neural network gathers data through runoff and rainfall from Mae Tum's gauging station. In addition, The time-lapse of rate of flow in the departments results as 4-6 hours. Thus, the input focus on the rate of flow in the previous four hours and the amount of rainfall by using two hidden layers with Sigmoid Function and Rectified linear unit Function. The model has diverged into two types. There are the model that only uses the input data is the runoff and the models that use the input data are runoff and rainfall which 1-4 hours in advance shows the best result. The four hours model that use the input data are runoff and rainfall with 12 units input and a hidden layer Sigmoid Function represent MSE; 0.2295.