Building Service Control Using Random Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building management systems rely on limited sensor data, often using a single type of sensor, which can lead to inefficient energy control and increased costs, as they fail to account for the complexities of various environmental conditions within and outside buildings.
Innovation Solution
A controller utilizing multiple random neural networks processes a variety of input variables, including environmental and weather data, to generate control parameters for building services such as heating, ventilation, and air conditioning, allowing for tailored control strategies specific to different parts of a building based on their unique properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single sensor or limited set of sensors is used in building management systems, then the system complexity is reduced, but the energy control efficiency deteriorates due to insufficient environmental data
Solution Approach 1:
The building is divided into multiple zones or areas, each equipped with its own sensors and controlled by dedicated neural network models. This segmentation allows comprehensive environmental monitoring without requiring a single complex centralized system, thus maintaining manageable complexity while improving energy control efficiency through localized optimization.
Solution Approach 2:
Neural network models are introduced as intermediary components that process sensor data and generate control decisions. These intermediaries transform raw environmental data into actionable control parameters, enabling efficient energy management without directly increasing physical sensor quantity or system complexity.
2Productivity
If multiple random neural networks are used to process various input variables, then the energy management efficiency is improved, but the device complexity increases
Solution Approach 1:
Multiple random neural networks are designed with a unified architecture and common training methodology, allowing them to serve different zones or functions while maintaining structural consistency. This universality enables the system to handle complex multi-variable control tasks without proportionally increasing overall system complexity, as the same computational framework is reused across different applications.
Solution Approach 2:
The system adjusts neural network parameters such as learning rates, network depths, and input variable selections based on specific zone requirements and environmental conditions. This parameter adaptation allows the same neural network framework to efficiently manage diverse energy control scenarios without requiring fundamentally different system architectures for each case.
3Device complexity
If traditional control methods are used, then the system is simpler to implement, but the calculation time and computational efficiency are insufficient compared to random neural networks
Solution Approach 1:
Neural networks are trained offline in advance using historical environmental data and control outcomes. This preliminary training phase prepares the networks with pre-learned patterns and relationships, enabling them to make rapid real-time control decisions without performing complex calculations during actual operation. The computationally intensive work is done beforehand, reducing real-time calculation time while maintaining sophisticated control capabilities.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A controller for controlling at least one service apparatus for a building or one or more parts of a building and an associated method, the controller having at least one input interface for receiving a plurality of input variables associated with the building or the one or more parts of the building;at least one output interface for interfacing with the at least one service apparatus; and processing apparatus configured to process the plurality of input variables using one or more random neural networks in order to generate one or more control parameters for controlling the at least one service apparatus. Preferably, the plurality of input variables include one or more environmental conditions associated with the building or the one or more parts of the building. Optionally, the controller or processing apparatus implements a plurality of random neural networks and the controller or processing apparatus implements a selector for selecting a random neural network from the plurality of neural networks dependent on the part of the building associated with the input variables.