Indoor substation ventilation air conditioner design load prediction method
By establishing physical models and multilayer perceptron models, and combining energy conservation constraints to optimize the indoor substation ventilation and air conditioning system, the shortcomings of static optimization in the design phase of existing technologies have been solved, achieving accurate load prediction and energy consumption reduction.
Patent Information
- Application Number
- CN202511065837.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-31
- Publication Date
- 2025-12-16
AI Technical Summary
Existing technologies have limitations in the static optimization of ventilation and air conditioning system design for indoor substations, lacking in-depth consideration of dynamic operating characteristics, which leads to increased energy consumption and inaccurate load calculations.
A physical model was built using the DesignBuilder platform, and parametric simulations were performed using the EnergyPlus parametric model and meteorological data. Global sensitivity analysis and pruning optimization were conducted using a multilayer perceptron model. Energy conservation constraints were introduced to optimize the ventilation design in order to improve prediction accuracy and system energy-saving effect.
It enables accurate load prediction for indoor substation ventilation and air conditioning systems, improves system operating efficiency, reduces energy consumption, and provides universal design optimization suggestions.
Smart Images
Figure CN121146136A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of air conditioning design load prediction, and particularly relates to an indoor substation ventilation air conditioning design load prediction method. BACKGROUND
[0002] With the rapid development of the power system, the number of indoor substations is increasing. The operation of a large number of electrical equipment in the substation will generate a large amount of heat, which will cause the indoor temperature to rise significantly if it cannot be discharged in time. The excessively high temperature will not only increase the resistance of electronic components and accelerate the aging of insulation, but also may cause a fire. In addition, frequent large temperature changes will also shorten the service life of the equipment. Therefore, the reasonable design of the ventilation and air conditioning system is crucial to maintaining the stability of the temperature and humidity in the substation.
[0003] At present, the energy saving and temperature control technology of the indoor substation mainly focuses on the optimization of the static building physical layer, for example, improving the heat dissipation effect by optimizing the building interface and the design of the enclosure structure, such as external shading, window-wall ratio, shape coefficient, etc. In addition, some researches use computational fluid dynamics (CFD) simulation means to compare and optimize the ventilation scheme of the substation, or combine the particle swarm optimization algorithm to study the collaborative design method of the size and position of the ventilation opening, so as to reduce the ventilation energy consumption and noise.
[0004] However, these methods are mostly limited to static optimization in the design stage, and lack of deep research on the dynamic operation characteristics of the ventilation and air conditioning system and the accurate calculation of the design load of the ventilation and air conditioning system in the design stage. This results in low operation efficiency of the ventilation and air conditioning system and increased energy consumption. SUMMARY
[0005] The present application aims to provide an indoor substation ventilation air conditioning design load prediction method, which realizes the prediction of the air conditioning design load of various indoor substations, solves the problems of static modeling limitation and insufficient energy consumption concern in the prior art, improves the prediction accuracy and system energy saving effect, and solves the common problems of design load calculation difficulty and low load calculation accuracy of the existing research methods, and the problem of improper equipment selection based on the prediction results.
[0006] To achieve the above-mentioned purpose, the present application provides an indoor substation ventilation air conditioning design load prediction method, which comprises the following steps: establishing a physical model of the indoor substation and exporting an EnergyPlus parameterized model through a DesignBuilder building simulation platform; based on the EnergyPlus parameterized model, combining meteorological data to perform parameterized simulation, and generating a plurality of sets of parameterized simulation data; performing global sensitivity analysis on the parameterized simulation data to screen key parameters affecting the design load prediction; Based on the selected key parameters, a multilayer perceptron model is constructed, and an energy conservation constraint is introduced into the loss function. Pruning optimization was performed on the multilayer perceptron model; The design load is predicted using a multilayer perceptron model, and the ventilation design of indoor substations is optimized and improved based on the prediction results. Output the final design load prediction results and optimization suggestions.
[0007] The process involves using the EnergyPlus parametric model and meteorological data to perform parametric simulations, generating multiple sets of parametric simulation data. Specific steps include: The EnergyPlus parameterized model is processed by the IDF parser to generate IDF objects; The meteorological files are processed by the EPW parser to generate EPW objects. The IDF and EPW objects are merged into a parameterized job object, and multiple sets of parameterized simulations are run to generate multiple sets of parameterized simulation data.
[0008] Among these, a global sensitivity analysis is performed on the parametric simulation data to screen key parameters affecting design load forecasting. The Sobol index method was used to conduct a global sensitivity analysis, calculate the variance contribution rate of each parameter, and screen out the parameters with high sensitivity.
[0009] Based on the selected key parameters, a multilayer perceptron model is constructed, and an energy conservation constraint is introduced into the loss function. The loss function is: Loss = MSE + λ·|Q 预测 −(Q 设备 +Q 太阳辐射 -Q 通风 )∣.
[0010] Among them, pruning optimization is performed on the multilayer perceptron model, in which: Redundant neurons and connections in the multilayer perceptron model are identified through SHAP value analysis, and structured pruning optimization is implemented.
[0011] Among these methods, a multilayer perceptron model is used to predict design loads, and the ventilation design and improvement of indoor substations are optimized based on the prediction results. The optimization strategies include optimizing the spatial layout and shading design of indoor substations, improving the thermal performance of the building envelope, and assessing the heat dissipation requirements of new equipment.
[0012] Based on the selected key parameters, a multilayer perceptron model is constructed, and an energy conservation constraint is introduced into the loss function. The multilayer perceptron model was pre-trained using a general dataset generated by EnergyPlus and then fine-tuned for a specific substation scenario.
[0013] This invention discloses a method for predicting the design load of ventilation and air conditioning (VAC) systems in indoor substations. First, a physical model of the indoor substation is established, and an EnergyPlus parametric model is exported using the DesignBuilder building simulation platform. Then, based on the EnergyPlus parametric model and meteorological data, parametric simulation is performed to generate multiple sets of parametric simulation data. Next, a global sensitivity analysis is conducted on the parametric simulation data to screen key parameters affecting the design load prediction. Then, based on the screened key parameters, a multilayer perceptron (MLP) model is constructed, and energy conservation constraints are introduced into the loss function to perform pruning optimization on the MLP model. Finally, load prediction is performed using the MLP model, and the energy-saving design strategy of the VAC system is optimized based on the prediction results, outputting the final design load prediction results and optimization suggestions. This method addresses the limitations of existing methods, which are limited to static optimization during the building design phase and lack in-depth consideration of the dynamic operating characteristics of the VAC system and accurate calculation of the design load. It conducts in-depth research on energy consumption prediction for the VAC system itself. It optimizes substation expansion schemes based on MLP prediction results and provides universally applicable load prediction for all types of substations, improving the operating efficiency of the VAC system and reducing energy consumption. Attached Figure Description
[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.
[0015] Figure 1 This is a physical model diagram of the present invention.
[0016] Figure 2 This is a flowchart of the parameterized simulation process of the present invention.
[0017] Figure 3 This is a flowchart of the sensitivity analysis of the present invention.
[0018] Figure 4 This is a flowchart of the MLP model establishment process of the present invention.
[0019] Figure 5 This is a flowchart of the indoor substation ventilation and air conditioning design load prediction method of the present invention.
[0020] Figure 6 This is a flowchart of the steps in the indoor substation ventilation and air conditioning design load prediction method of the present invention. Detailed Implementation
[0021] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.
[0022] The first embodiment of this application is as follows: Please see Figures 1 to 6 ,in, Figure 1 This is a physical model diagram of the present invention. Figure 2 This is a flowchart of the parameterized simulation process of the present invention. Figure 3 This is a flowchart of the sensitivity analysis of the present invention. Figure 4 This is a flowchart of the MLP model establishment process of the present invention. Figure 5 This is a flowchart of the indoor substation ventilation and air conditioning design load prediction method of the present invention. Figure 6 This is a flowchart of the steps in the indoor substation ventilation and air conditioning design load prediction method of the present invention.
[0023] This invention provides a method for predicting the design load of ventilation and air conditioning in indoor substations, comprising the following steps: S101: Establish a physical model of the indoor substation and export the EnergyPlus parametric model through the DesignBuilder architectural simulation platform; S102: Based on the EnergyPlus parametric model and combined with meteorological data, parametric simulation is performed to generate multiple sets of parametric simulation data; Specifically, based on on-site surveys and review of relevant design parameters, raw data about the substation was obtained. An indoor substation physical model was constructed using the DesignBuilder building simulation platform, and then an EnergyPlus parametric model was exported for parametric simulation. The obtained EnergyPlus parametric model was used as a seed model and processed by the IDF parser to generate an IDF object. The weather file for the substation's location was processed by the EPW parser to generate an EPW object. The IDF and EPW objects were merged to generate a ParametricJob object. In the ParametricJob object, the measurement parameters were applied using the `$apply_measure()` method. Specific parameters included Parameter_1 to Parameter_n, each with a series of time steps (e.g., Parameter_1 = t1_1, t1_2, t1_3, ...). Based on these parameter settings, multiple Parametric Models were generated, including IDF1, IDF2, IDF3, and so on, up to IDFn. These parameterized models are run using the `$run()` method. During model execution, R is invoked for computation. After execution, a series of results are generated, including `$errors()` (error messages), `$output_dir()` (output directory), `$locate_output()` (locate output), and a series of data reading and reporting methods, such as `$read_table()`, `$read_rdd()`, `$read_mdd()`, `$report_data_dict()`, `$reprot_data()`, and `$tabular_data()`. Finally, the results are stored in the `Results` object, which can be stored in multiple SQL databases.
[0024] It achieves fully automated parametric simulation from physical modeling to data storage, solving the problems of low efficiency and fragmented data management in traditional manual modeling.
[0025] S103: Perform global sensitivity analysis on parametric simulation data to screen key parameters that affect design load forecasting; Specifically, sensitivity analysis is performed on the generated parametric simulation dataset. The n sets of samples generated by the parametric simulation, containing m input parameters (such as the heat transfer coefficient of the building envelope, transformer power, meteorological data, etc.) and 1 output index (such as total annual energy consumption), are subjected to global sensitivity analysis based on the Sobol exponent method. The simulation is automatically run using R language to calculate the parameter variance contribution rate, and highly sensitive parameters (such as transformer power and ventilation opening size) are screened to reduce model complexity. Finally, a high-quality simulation dataset covering key parameter combinations is generated to provide training and validation samples for subsequent modeling.
[0026] By optimizing input features through a data-driven approach, efficient and accurate training samples can be provided for subsequent modeling.
[0027] S104: Based on the selected key parameters, construct a multilayer perceptron model and introduce energy conservation constraints into the loss function; S105: Pruning and optimization of the multilayer perceptron model; S106: Use a multilayer perceptron model to predict design loads and optimize and improve the ventilation design of indoor substations based on the prediction results; S107: Output the final design load prediction results and optimization suggestions.
[0028] Specifically, in the multilayer perceptron (MLP) modeling stage, guided by the results of sensitivity analysis, highly sensitive physical parameters (such as the heat output Q of the main equipment) are selected. 设备 This is used as an input node to enhance feature interpretability, and an energy conservation equation is introduced into the loss function as a regularization term (Loss=MSE+λ·∣Q). 预测 −(Q 设备 +Q 太阳辐射 -Q 通风 The constraint model prediction conforms to thermodynamic laws. Here, MSE (Mean Squared Error) measures the average squared error between the model's predicted and actual values; it is a fundamental component of the loss function and reflects the degree of deviation between the predicted and actual loads. λ (Regularization Coefficient) balances the weights of the MSE and regularization terms, controlling the strength of the energy conservation constraint, preventing overfitting, and ensuring that the model satisfies physical constraints while fitting the data. 预测 Q: The indoor substation ventilation and air conditioning load value predicted by the multilayer perceptron (MLP) model, which is the target variable to be predicted in the model output, representing the energy consumption of the ventilation and air conditioning system required by the substation to maintain indoor temperature and humidity. 设备 The heat generated by electrical equipment during operation in indoor substations is one of the key highly sensitive physical parameters affecting the load of ventilation and air conditioning equipment, such as the heat released during the operation of equipment like transformers.太阳辐射 Solar radiation heat entering the interior through the building envelope (such as walls and windows) is an important input factor in environmental parameters. Its intensity is affected by geographical location, season, weather, and other conditions, and will directly lead to an increase in indoor temperature, thereby increasing the air conditioning load. 通风 Heat expelled or brought into the room through the ventilation system (including natural and mechanical ventilation) is affected by factors such as the size and location of the vents and the ventilation rate. A well-designed ventilation system can effectively reduce indoor heat accumulation and decrease the air conditioning load. Loss: Represents the model's total loss function. By forcibly constraining the model's predictions to conform to fundamental thermodynamic laws, the predicted values are ensured to be consistent with the actual physical processes, improving the model's prediction accuracy and reliability. Simultaneously, SHAP value analysis identifies neurons and connections with less influence in the model, implementing pruning optimization to improve computational efficiency. This reduces the number of model parameters while maintaining prediction accuracy, lowers computational complexity, and improves model operating efficiency, solving the problems of high computational time and resource consumption in complex neural networks, thus achieving a lightweight model. The model simulates the long-term impact of different control strategies on energy consumption and equipment lifespan using the MLP model. Based on the MLP prediction results, it optimizes substation expansion schemes (such as assessing the heat dissipation requirements of new equipment). Furthermore, based on the general dataset of EnergyPlus, it trains and generalizes the MLP model through parametric simulation, fine-tunes the MLP model to specific substation scenarios, and provides universal design load prediction for all types of substations. This breaks through the limitation of traditional models that are only applicable to a single scenario, and achieves a combination of the versatility and scenario adaptability of the prediction method.
[0029] This research addresses the limitations of existing methods, which are confined to static optimization during the building design phase and lack in-depth consideration of the dynamic operating characteristics of ventilation and air conditioning (VAC) systems and accurate calculation of their design loads. It conducts in-depth research on energy consumption and energy-saving prediction for VAC systems themselves, and resolves the issues of insufficient data-driven application and inaccurate load calculations in existing methods. Based on MLP prediction results, it optimizes substation expansion schemes and provides universally applicable design load predictions for all types of substations, which helps improve the operating efficiency of VAC systems and reduce energy consumption.
[0030] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.
Claims
1. A method for predicting the design load of ventilation and air conditioning in an indoor substation, characterized in that, Includes the following steps: A physical model of the indoor substation was established, and an EnergyPlus parametric model was exported using the DesignBuilder architectural simulation platform. Based on the EnergyPlus parametric model and combined with meteorological data, parametric simulation is performed to generate multiple sets of parametric simulation data. Perform global sensitivity analysis on parametric simulation data to screen key parameters that affect design load forecasting; Based on the selected key parameters, a multilayer perceptron model is constructed, and an energy conservation constraint is introduced into the loss function. Pruning optimization was performed on the multilayer perceptron model; The design load is predicted using a multilayer perceptron model, and the ventilation design and improvement of indoor substations are optimized based on the prediction results. Output the final design load prediction results and optimization suggestions.
2. The method for predicting the design load of ventilation and air conditioning in indoor substations as described in claim 1, characterized in that, Based on the EnergyPlus parametric model and combined with meteorological data, parametric simulations are performed to generate multiple sets of parametric simulation data. The specific steps include: The EnergyPlus parameterized model is processed by the IDF parser to generate IDF objects; The meteorological files are processed by the EPW parser to generate EPW objects. The IDF and EPW objects are merged into a parameterized job object, and multiple sets of parameterized simulations are run to generate multiple sets of parameterized simulation data.
3. The method for predicting the design load of ventilation and air conditioning in indoor substations as described in claim 2, characterized in that, A global sensitivity analysis was performed on the parametric simulation data to identify key parameters affecting design load forecasting. The Sobol index method was used to conduct a global sensitivity analysis, calculate the variance contribution rate of each parameter, and screen out the parameters with high sensitivity.
4. The method for predicting the design load of ventilation and air conditioning in indoor substations as described in claim 3, characterized in that, Based on the selected key parameters, a multilayer perceptron model is constructed, and an energy conservation constraint is introduced into the loss function. The loss function is: Loss = MSE + λ·|Q 预测 −(Q 设备 +Q 太阳辐射 -Q 通风 )∣.
5. The method for predicting the design load of ventilation and air conditioning in indoor substations as described in claim 4, characterized in that, Pruning optimization of the multilayer perceptron model, in part: Redundant neurons and connections in the multilayer perceptron model are identified through SHAP value analysis, and structured pruning optimization is implemented.
6. The method for predicting the design load of ventilation and air conditioning in indoor substations as described in claim 5, characterized in that, A multilayer perceptron model was used to predict design loads, and the ventilation design and improvement of indoor substations were optimized based on the prediction results. The optimization strategies include optimizing the spatial layout and shading design of the substation, improving the thermal performance of the building envelope, and assessing the heat dissipation requirements of new equipment.
7. The method for predicting the design load of ventilation and air conditioning in indoor substations as described in claim 6, characterized in that, Based on the selected key parameters, a multilayer perceptron model is constructed, and an energy conservation constraint is introduced into the loss function. The multilayer perceptron model was pre-trained using a general dataset generated by EnergyPlus and then fine-tuned for a specific substation scenario.