Forecasting based on the second generation artificial neural network for decision support in especially significant situations

Nowadays, specialized system models implemented on the basis of decision support in exceptional (emergency) situations (states) using machine learning, artificial intelligence (including using neural networks) to reproduce, predict and prevent (or minimize risk) consequences) in exceptional situations are useful and becoming increasingly popular. Floods also fall under such exceptional situations and states. Therefore, there arises the problem of early forecasting of an exceptional situation using the example of rising water levels at stationary hydrological posts in order to prevent (or minimize the risk) the transition of the system under consideration into an exceptional state (emergency situation). To solve this problem, the authors propose a decision support system for early forecasting water rise levels. It is based on a neural network (intelligent) analysis of retrospective data (code of a stationary hydrological station / automatic station, date, water level at a stationary hydrological station / automatic station, atmospheric pressure, wind speed, snow cover thickness, amount of precipitation, time and air temperature) in order to calculate the values of water levels for 5 days in advance. The artificial neural network itself is based on the freely distributed TensorFlow machine learning software library; a modified backpropagation method is used as training. Its main difference is the addition of an artificial neural network (ANN) learning rate increase factor. An analysis of the effectiveness of the proposed solution in the framework of forecasting the flood situation has shown high accuracy in calculating the forecast values of water levels: the difference between the real and predicted values is 2.10 %. This will allow specialized services to carry out specialized anti-flood measures in advance (5 days in advance). Thus, information support during special situations is an absolute (not relative) indicator of data quality that allows developing and making decisions in the framework of predicting possible critical situations and preventing the transfer of the state of the territory management system to critical situations.