Geothermal cascade heating decision optimization system based on deep learning

Abstract

The invention discloses a geothermal cascade heating decision optimization system based on deep learning, and relates to the technical field of heating systems, comprising the following steps: S1, data acquisition and preprocessing; s1-1, a timing data and real-time data acquisition module; s1-2, a data preprocessing module; s2, pre-estimating and evaluating; s2-1, a building heating load prediction module; s2-2, a geothermal resource evaluation module; s3, decision optimization is carried out; s3-1, a deep learning optimization model module; s3-2, a multi-objective decision module; s4, executing and feeding back; s4-1, a strategy execution and monitoring module; and S4-2, a feedback and iteration module. According to the geothermal cascade heating decision optimization system based on deep learning, geothermal resource fluctuation and heating load change can be predicted 12-24 hours ahead of time based on historical data and real-time data through a time sequence prediction capability relying on deep learning, so that judgment can be made in advance before decision making, and the decision making can be planned.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of heating systems, in particular to a geothermal cascade heating decision optimization system based on deep learning. BACKGROUND

[0002] The geothermal cascade heating decision optimization system is a comprehensive solution combining geothermal energy cascade utilization technology and intelligent decision algorithm. Its core function is to achieve efficient, economic and safe operation of the geothermal heating system through precise control and dynamic optimization. It dynamically adjusts the heating output based on user heating behavior data and weather forecasts to reduce over-heating or insufficient heating. In use, traditional heating decision systems usually rely on experience or fixed rules provided by historical data when making decisions. However, these data and rules cannot adapt to changes in working conditions during long-term operation, and the decision-making effect will decay over time, leading to data bias in decisions provided by historical data or fixed rules.

[0003] To overcome the above-mentioned defects, the existing technology one (Chinese patent with publication number CN119962739A and publication date 2025-05-09) is a wisdom heating system optimization scheduling platform considering human-machine collaboration, which includes: a digital twin unit for establishing a heating system digital twin model using mechanism modeling and data identification methods; a heating AI unit including a machine decision module for outputting a machine decision-based heating system optimization scheduling strategy; a manual analysis unit including a model manual verification module for dispatchers to construct a dispatching business model performance evaluation index to evaluate and verify the dispatching business model; further including a manual decision module for dispatchers to output a manual decision-based heating system optimization scheduling strategy; a human-machine collaborative decision unit for obtaining a human-machine collaborative decision-based heating system optimization scheduling strategy; further for state perception and intention understanding of dispatchers through multi-channel human-machine interaction technology to assist the heating AI unit in decision-making; a mode switching unit for switching the heating system optimization scheduling mode between manual decision and machine decision; Prior art two (the patent number of China is CN119862470B, the announcement date is 2025-05-27) a kind of based on artificial intelligence big model's heating intelligent decision system, comprising: data layer, for storing the source network load storage multi-type data of heating system;Knowledge base establishment and big model training layer, for establishing the diversified task knowledge base of heating intelligent decision and establishing each task big model;Man-machine enhancement optimization big model layer, for making improvement suggestion to big model, optimization generated answer, and the multiple answers generated by big model are evaluated and the big model parameter optimization training is carried out;Agent decision layer, for integrating each task big model into each agent to carry out environmental perception, collaborative interaction and autonomous adaptability decision-making;Simulation simulation layer, for providing simulation running environment, user interface and scene dynamic parameter simulation;Agent evaluation optimization layer, for constructing the performance index of evaluating each agent, the behavior and decision of each agent are evaluated and optimized adjustment.

[0004] The above mechanism realizes the optimization of decision-making by utilizing the mutual cooperation between man and machine, but in the process of actual use, due to the non-transparent decision logic and the lack of basis for parameter adjustment, it is usually difficult for operation and maintenance personnel to understand, and it is more difficult to optimize and troubleshoot subsequently, affecting the stability of the subsequent decision system in use. SUMMARY

[0005] The present application aims to provide a geothermal cascade heating decision optimization system based on deep learning to solve the problem that the traditional heating decision system relies on the experience or fixed rules provided by historical data when making decisions, and these data and rules cannot adapt to the changes in working conditions generated during long-term operation, and the decision-making effect will decay over time, resulting in data deviation in the decision-making provided by historical data or fixed rules.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a geothermal cascade heating decision optimization system based on deep learning, comprising the following steps: S1, data acquisition and preprocessing; S1-1, timed data and real-time data acquisition module; S1-2, data preprocessing module; S2, estimation and evaluation; S2-1, building heating load prediction module; S2-2, geothermal resource evaluation module; S3, decision optimization; S3-1, deep learning optimization model module; S3-2, multi-objective decision module; S4, execution and feedback; S4-1, strategy execution and monitoring module; S4-2, feedback and iteration module.

[0007] Further, the S1-1 timed data and real-time data acquisition module collects various data according to the principle of full coverage of multi-source heterogeneous data, and the timed data and real-time data acquisition module includes collecting environmental and load data and geothermal system data.

[0008] Further, the environmental and load data includes outdoor temperature, humidity, solar radiation intensity, wind speed, real-time heating load and historical load data of each area in the building, and the geothermal system data includes geothermal well water temperature, flow, pressure, recharging temperature, flow, geothermal heat exchanger inlet and outlet parameters, water pump speed of each link of cascade heating, valve opening, heat exchanger energy efficiency and other operation data.

[0009] Further, the S1-2 data preprocessing module includes data cleaning, data integration and data standardization, the data cleaning includes processing missing values and removing outliers, the data integration includes associating and binding data of different sources and different formats, and the data standardization includes processing the data to standardize it to avoid the influence of data dimension difference on the deep learning optimization model.

[0010] Further, the S2-1 building heating load prediction module includes input of preprocessed environmental data and load data, model training and prediction, and result verification, the model training and prediction includes training the model through historical data, and after completion, inputting future environmental prediction data and outputting future heating load prediction curve.

[0011] Further, the S2-2 geothermal resource assessment module includes data input, resource potential assessment and output result, the data input includes calling preprocessed geothermal well water temperature, flow, recharging temperature, flow, geologic heat conductivity, underground water level and other data, and future load demand output by the load prediction module.

[0012] Further, the S3-1 deep learning optimization model module includes determining optimization variables and constraints, optimization variables, constraint conditions, model training and optimization, and outputting preliminary optimization scheme, the optimization variables include controllable parameters of each link of cascade heating, and the constraint conditions include geothermal resource constraints, heating quality constraints and equipment safety constraints.

[0013] Further, the S3-2 multi-objective decision module includes target weight setting, scheme evaluation and optimal scheme determination, the target weight setting needs to determine the weight of each target according to user demand or entropy weight method, the scheme evaluation scores the preliminary scheme in terms of energy consumption, cost, geothermal resource utilization rate, indoor temperature compliance rate and other indicators, and determines the optimal scheme after calculating the comprehensive score.

[0014] Further, the S4-1 strategy execution and monitoring module includes instruction issuing, real-time monitoring and exception handling, the instruction issuing converts the operation parameters in the decision strategy into device control signals through the industrial control system, the real-time monitoring collects device operation state and heating effect in real time through sensors, and data is uploaded to the monitoring platform in real time.

[0015] Further, the S4-2 feedback and iteration module includes effect evaluation, model iteration and strategy adjustment, the effect evaluation includes comparing the expected effect of the decision strategy with the actual operation effect, calculating the error, analyzing the error reason, the model iteration includes supplementing the actual operation data to the historical database, retraining the load prediction model and the deep learning optimization model, adjusting the model parameters and improving the model precision.

[0016] Compared with the prior art, the present application has the following beneficial effects: 1. Based on the time series prediction ability of deep learning, the fluctuation of geothermal resources and the change of heating load can be predicted 12-24 hours in advance based on historical data and real-time data, so that the planning can be made in advance before the decision is made.

[0017] Further, through the multi-objective optimization algorithm driven by deep learning, the core objectives of energy utilization rate, user comfort, cost control and equipment life are balanced.

[0018] Further, through the fault diagnosis and dynamic reconstruction ability of deep learning, downtime maintenance can be avoided when a fault occurs, and self-compensation of interference is realized, so that the robustness exceeds that of the prior art; the module receives device sensor data in real time, through the pre-trained deep learning fault diagnosis model, the fault type can be identified in a short time, and the solution can be automatically generated, the load of the fault stage can be distributed to the standby heat exchanger, or the gap can be compensated by adjusting the flow of other stages to avoid system paralysis; when disturbed, when the external working condition changes suddenly, the module combines real-time data and short-term prediction to prioritize core demand, and calls backup energy to supplement the gap to ensure heating continuity.

[0019] 2. By connecting the system monitoring platform, the stage distribution scheme and optimization target are displayed in the form of charts, and the operation and maintenance personnel can clearly understand the decision reason, result tracing and analysis: the module automatically records the input data, decision process and execution result of each decision, when a problem occurs, the historical data can be traced back to locate the cause, and the optimization can be quickly optimized.

[0020] Further, the explainability design is integrated into the deep learning framework, and the monitoring platform is connected to realize full-process transparency, which greatly reduces the operation and maintenance difficulty, the decision logic can be traced back, and the module records the reason of each optimization, and the operation and maintenance personnel can check the complete link from input to decision to output through the platform, and understand the data support behind the decision; compared with the traditional blind search operation and maintenance, the module can shorten the fault positioning time, improve the efficiency of daily parameter adjustment, and greatly reduce the experience dependence of the operation and maintenance personnel. BRIEF DESCRIPTION OF DRAWINGS

[0021] Fig. 1 The present application is a system flowchart.

[0022] Fig. 2 Fig. 1 is a schematic diagram of a system architecture according to the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0024] Embodiment one: as Figs. 1-2The technical scheme is shown, in order to solve the problem that the traditional heating decision-making system usually relies on the experience or fixed rules provided by the historical data when making decisions, and these data and rules cannot adapt to the changes in working conditions generated in long-term operation, and the decision-making effect will decay over time, and then the decision-making provided by the historical data or fixed rules will deviate from the data: the geothermal cascade heating decision-making optimization system based on deep learning discloses a real-time model parameter updating module, including the following steps: S1, data acquisition and preprocessing; S1-1, timing data and real-time data acquisition module; S1-2, data preprocessing module; S2, estimation and evaluation; S2-1, building heating load prediction module; S2-2, geothermal resource evaluation module; S3, decision optimization; S3-1, deep learning optimization model module; S3-2, multi-objective decision-making module; S4, execution and feedback; S4-1, strategy execution and monitoring module; S4-2, feedback and iteration module; the S1-1 timing data and real-time data acquisition module collects various data according to the multi-source heterogeneous data full coverage principle, the timing data and real-time data acquisition module includes collecting environmental and load data and geothermal system data; the environmental and load data includes outdoor temperature, humidity, solar radiation intensity, wind speed, real-time heating load and historical load data of each area in the building, and the geothermal system data includes geothermal well water temperature, flow, pressure, recharge temperature, flow, geothermal heat exchanger inlet and outlet parameters, water pump speed of each link of cascade heating, valve opening, heat exchanger energy efficiency and other operation data; the S1-2 data preprocessing module includes data cleaning, data integration and data standardization, the data cleaning includes processing missing values and removing outliers, the data integration includes associating and binding data of different sources and different formats, and the data standardization includes processing the data to standardize it, avoiding the influence of dimension difference on the deep learning optimization model; the S2-1 building heating load prediction module includes inputting preprocessed environmental data and load data, model training and prediction, and result verification, the model training and prediction includes training the model through historical data, inputting future environmental prediction data after completion, and outputting future heating load prediction curve; the S2-2 geothermal resource evaluation module includes data input, resource potential evaluation and output result, the data input includes calling preprocessed geothermal well water temperature, flow, recharge temperature, flow, geology thermal conductivity, underground water level and other data, and future load demand output by the load prediction module.

[0025] In this example, S1 data acquisition and preprocessing is the data layer of the system, responsible for collecting original data of the whole chain of geothermal heating, providing high-quality data input for the deep learning model, and is the basis for decision optimization, predicting future heating load demand for a period of time through the deep learning model, and providing basis for heat distribution of cascade heating.

[0026] Traditional tiered heating decisions often rely on fixed rules or human experience, failing to respond to dynamic changes in heating demand and geothermal resources. By leveraging the time-series prediction capabilities of deep learning, it is possible to predict geothermal resource fluctuations and heating load changes 12 to 24 hours in advance based on historical and real-time data, enabling advance assessment and planning before making decisions. Through deep learning-driven multi-objective optimization algorithms, the core objectives of energy utilization, user comfort, cost control, and equipment lifespan are simultaneously balanced.

[0027] Example 2: Figs. 1-2 The technical solution presented addresses the problem that underfloor heating systems are susceptible to external interference and internal faults during practical use, leading to system downtime or inefficient operation. This deep learning-based geothermal cascade heating decision optimization system discloses interference self-compensation. The S3-1 deep learning optimization model module includes determining optimization variables and constraints, defining the variables and constraints, model training and optimization, and outputting a preliminary optimization scheme. The optimization variables include controllable parameters for each stage of the cascade heating system, and the constraints include geothermal resource constraints, heating quality constraints, and equipment safety constraints. The S3-2 multi-objective decision module includes setting objective weights, scheme evaluation, and determining the optimal scheme. Setting objective weights requires combining user needs or using the entropy weight method to determine the weights of each objective. The scheme evaluation scores the preliminary scheme based on indicators such as energy consumption, cost, geothermal resource utilization rate, and indoor temperature compliance rate, and calculates a comprehensive score to determine the optimal scheme.

[0028] In this example, the geothermal resource dynamic assessment module is the resource layer of the system. It is responsible for assessing the real-time availability of geothermal resources, avoiding over-extraction leading to resource depletion, or under-extraction leading to heat waste, and providing constraints on the upper limit of heat sources for cascade heating. By calculating the maximum exploitable amount of the current geothermal well in real time, and combining it with reinjection efficiency to judge resource sustainability, it tracks changes in water temperature and water quality, corrects the calculation of usable heat, and predicts the geothermal resource decline trend in the next 1 to 5 years based on historical mining data and geological models, thus supporting the long-term planning of the system.

[0029] The cascade heating decision optimization module is the system's decision-making layer. Combining load forecast results and geothermal resource assessment results, it outputs the optimal heating operation plan through deep learning and optimization algorithms. The core is to realize the cascade distribution of geothermal heat. The geothermal fluid first heats the high-temperature demand side through the primary heat exchanger, and then the cooled geothermal fluid heats the low-temperature demand side through the secondary heat exchanger. Finally, the fluid close to room temperature is reinjected into the ground, realizing the cascade utilization of heat and maximizing resource utilization.

[0030] The traditional decision system has weak response capability to external interference and internal failure, and often can only be shut down for maintenance, resulting in interruption of heating or large area inefficiency. Through the fault diagnosis and dynamic reconstruction capability of deep learning, the system can avoid shutdown maintenance when failure occurs, and self-compensate for interference, so that the robustness exceeds the existing technology. The module receives real-time device sensor data, identifies the fault type in a short time through the pre-trained deep learning fault diagnosis model, and automatically generates a solution. The fault gradient load can be distributed to the standby heat exchanger, or the gap can be compensated by adjusting the flow of other stages to avoid system paralysis. When disturbed, the module combines real-time data and short-term prediction to prioritize core demand, while calling backup energy to supplement the gap and ensure heating continuity.

[0031] In this embodiment, the system control and execution module is the execution layer of the system, which converts the scheme output by the decision optimization module into actual device operation, realizes the closed loop from decision to execution, and outputs real-time control signals. Finally, the system is fed back by the iteration module. Figs. 1-2 As shown in the technical scheme, in order to solve the problem that in actual use, due to the opaque decision logic and the lack of basis for parameter adjustment, it is difficult for operation and maintenance personnel to understand, and it is difficult to optimize and troubleshoot subsequently, affecting the stability of the subsequent decision system in use: the geothermal gradient heating decision optimization system based on deep learning discloses decision logic visualization. The S4-1 strategy execution and monitoring module includes instruction issuing, real-time monitoring and exception handling. The instruction issuing converts the operation parameters in the decision strategy into device control signals through the industrial control system. The real-time monitoring collects the device operation state and heating effect in real time through the sensor, and uploads the data to the monitoring platform in real time. The S4-2 feedback and iteration module includes effect evaluation, model iteration and strategy adjustment. The effect evaluation includes comparing the expected effect of the decision strategy with the actual operation effect, calculating the error, analyzing the error reason, the model iteration includes supplementing the actual operation data to the historical database, retraining the load prediction model and the deep learning optimization model, adjusting the model parameters, and improving the model precision.

[0032] In this example, the system control and execution module is the execution layer of the system, which converts the scheme output by the decision optimization module into actual device operation, realizes the closed loop from decision to execution, and outputs real-time control signals. Finally, the system is fed back by the iteration module.

[0033] The decision process visualization is visually displayed in the form of charts through the system monitoring platform, and the hierarchical allocation scheme and optimization target are intuitively displayed. The operation and maintenance personnel can clearly understand the decision reason, result traceability and analysis: the module automatically records the input data, decision process and execution result of each decision. When a problem occurs, the historical data can be traced back to locate the cause, so as to facilitate rapid optimization; the explainability design is integrated into the deep learning framework, and the whole process is transparent through the monitoring platform, which greatly reduces the operation and maintenance difficulty, the decision logic is traceable, and the module records the reason of each optimization. The operation and maintenance personnel can view the complete link from input to decision to output through the platform, and understand the data support behind the decision; compared with the traditional blind search operation and maintenance, the module can shorten the fault positioning time, improve the efficiency of daily parameter adjustment, and greatly reduce the experience dependence of the operation and maintenance personnel.

[0034] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A deep learning-based decision optimization system for geothermal cascade heating, characterized in that: Includes the following steps: S1. Data Acquisition and Preprocessing; S1-1, Timed data and real-time data acquisition module; S1-2, Data Preprocessing Module; S2. Forecasting and Assessment; S2-1, Building Heating Load Prediction Module; S2-2, Geothermal Resource Assessment Module; S3, Decision Optimization; S3-1, Deep Learning Optimization Model Module; S3-2, Multi-objective decision-making module; S4. Execution and Feedback; S4-1, Strategy Execution and Monitoring Module; S4-2, Feedback and Iteration Module.

2. The geothermal cascade heating decision optimization system based on deep learning according to claim 1, characterized in that: The S1-1 timed data and real-time data acquisition module collects various data according to the principle of full coverage of multi-source heterogeneous data. The timed data and real-time data acquisition module includes collecting environmental and load data and geothermal system data.

3. The geothermal cascade heating decision optimization system based on deep learning according to claim 2, characterized in that: The environmental and load data include outdoor temperature, humidity, solar radiation intensity, wind speed, real-time heating load and historical load data for each area of ​​the building, and the geothermal system data includes geothermal well outlet water temperature, flow rate, pressure, reinjection temperature, flow rate, geothermal heat exchanger inlet and outlet parameters, pump speed, valve opening, heat exchanger energy efficiency and other operating data for each stage of the cascade heating system.

4. The geothermal cascade heating decision optimization system based on deep learning according to claim 3, characterized in that: The S1-2 data preprocessing module includes data cleaning, data integration, and data standardization. Data cleaning includes handling missing values ​​and removing outliers. Data integration includes associating and binding data from different sources and in different formats. Data standardization includes processing the data to standardize it, so as to avoid the impact of differences in data units on the deep learning optimization model.

5. The deep learning-based geothermal cascade heating decision optimization system according to claim 4, characterized in that: The S2-1 building heating load prediction module includes inputting preprocessed environmental data and load data, model training and prediction, and result verification. The model training and prediction includes training the model using historical data, inputting future environmental prediction data after completion, and outputting future heating load prediction curves.

6. The geothermal cascade heating decision optimization system based on deep learning according to claim 5, characterized in that: The S2-2 geothermal resource assessment module includes data input, resource potential assessment, and output results. The data input includes calling up pre-processed data such as geothermal well outlet water temperature, flow rate, reinjection temperature, flow rate, geological thermal conductivity, and groundwater level, as well as future load demand output by the load prediction module.

7. The geothermal cascade heating decision optimization system based on deep learning according to claim 6, characterized in that: The S3-1 deep learning optimization model module includes determining optimization variables and constraints, optimization variables, constraint conditions, model training and optimization, and outputting a preliminary optimization scheme. The optimization variables include controllable parameters of each link in the cascade heating system, and the constraint conditions include geothermal resource constraints, heating quality constraints, and equipment safety constraints.

8. The geothermal cascade heating decision optimization system based on deep learning according to claim 7, characterized in that: The S3-2 multi-objective decision-making module includes objective weight setting, scheme evaluation, and optimal scheme determination. The objective weight setting needs to combine user needs or entropy weight method to determine the weight of each objective. The scheme evaluation scores the preliminary scheme for indicators such as energy consumption, cost, geothermal resource utilization rate, and indoor temperature compliance rate, and calculates the comprehensive score to determine the optimal scheme.

9. A deep learning-based geothermal cascade heating decision optimization system according to claim 8, characterized in that: The S4-1 strategy execution and monitoring module includes instruction issuance, real-time monitoring, and anomaly handling. The instruction issuance transforms the operating parameters in the decision-making strategy into equipment control signals through the industrial control system. The real-time monitoring collects the equipment operating status and heating effect in real time through sensors, and the data is uploaded to the monitoring platform in real time.

10. A deep learning-based geothermal cascade heating decision optimization system according to claim 9, characterized in that: The S4-2 feedback and iteration module includes effect evaluation, model iteration, and strategy adjustment. The effect evaluation includes comparing the expected effect of the decision strategy with the actual operating effect, calculating the error, and analyzing the cause of the error. The model iteration includes supplementing the actual operating data into the historical database, retraining the load prediction model and the deep learning optimization model, adjusting the model parameters, and improving the model accuracy.

Images