Park energy management method based on DeepSeek, computer equipment and readable storage medium
The campus energy management system, built using the DeepSeek large model and multi-objective optimization algorithm, solves the problems of insufficient load forecasting accuracy and fault diagnosis in traditional systems, achieves accurate load forecasting and equipment failure warnings, optimizes energy use, reduces costs and carbon emissions, and improves management efficiency.
Patent Information
- Application Number
- CN202510508618.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-22
- Publication Date
- 2025-09-12
AI Technical Summary
Traditional campus energy management systems have limited load forecasting accuracy, making it difficult to achieve multi-objective optimization, and insufficient equipment fault diagnosis and early warning capabilities, resulting in energy waste and equipment loss.
Using a large model and multi-objective optimization algorithm based on DeepSeek, combined with genetic algorithm and particle swarm optimization algorithm, a personalized campus energy management system is built. Through deep learning and real-time data analysis, the optimal energy usage strategy is generated and fault warnings are provided.
It significantly improves load forecasting accuracy, optimizes energy usage strategies, reduces energy consumption costs and carbon emissions, reduces energy waste and equipment loss, and improves user experience and management efficiency.
Smart Images
Figure CN120634290A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of energy management technology, and in particular to a campus energy management method, computer equipment, and readable storage medium based on DeepSeek. Background Art
[0002] With the continuous increase in energy demand and rising energy costs, the intelligence of the park energy management system has become the key to improving energy utilization efficiency and reducing operating costs.
[0003] When it comes to energy load identification and forecasting, traditional park energy management systems often use methods such as linear regression and time series analysis. However, these methods struggle to accurately capture the complex dynamics of park energy consumption, resulting in limited prediction accuracy. Furthermore, park load power regulation strategies are often based on simple heuristic rules or single-objective optimization algorithms, making it difficult to achieve coordinated optimization of multiple objectives (such as cost, efficiency, and environmental protection) and lacking adaptability to actual park operations. Furthermore, while there are methods for equipment fault diagnosis and early warning based on threshold judgment or traditional machine learning models (such as decision trees and support vector machines), these models lack the ability to identify complex fault characteristics, resulting in inadequate timeliness and accuracy of early warnings, leading to energy waste and equipment loss.
[0004] Therefore, there is an urgent need to develop an energy management method based on advanced artificial intelligence technology to achieve more accurate load forecasting and more efficient energy optimization. Summary of the Invention
[0005] The technical problem to be solved by the present invention is to provide a campus energy management method, computer equipment and readable storage medium based on DeepSeek, which can significantly improve the accuracy of load forecasting.
[0006] In order to solve the above technical problems, the present invention provides a park energy management method based on DeepSeek, including: obtaining benchmark data of the park energy system, the benchmark data including historical energy consumption data, historical environmental data, real-time energy consumption data and real-time environmental data; preprocessing the benchmark data; constructing a DeepSeek large model; inputting the preprocessed historical energy consumption data and historical environmental data into the DeepSeek large model for training and verification to generate an optimized load forecasting model; inputting the preprocessed real-time energy consumption data and real-time environmental data into the optimized load forecasting model to generate energy consumption and load forecasting results.
[0007] As an improvement to the above-mentioned solution, the DeepSeek-based campus energy management method also includes: constructing optimization objectives, which include reducing energy consumption costs, improving energy utilization efficiency and reducing carbon emissions; constructing operating constraints, which include equipment capacity limitations, energy price fluctuations and environmental protection policy requirements; inputting the optimization objectives, operating constraints and energy load forecast results into a pre-trained multi-objective DeepSeek model to generate an optimal energy utilization strategy; controlling the campus energy system according to the optimal energy utilization strategy to adjust the equipment scheduling and energy distribution of the campus in real time.
[0008] As an improvement to the above-mentioned scheme, the training steps of the multi-objective DeepSeek model include: obtaining multi-source data of the park energy system, the multi-source data including historical energy consumption data, historical environmental data, real-time energy consumption data, real-time environmental data and equipment operation status data; preprocessing the multi-source data; inputting the preprocessed multi-source data into the DeepSeek large model, performing dimensionality reduction and feature extraction processing on the multi-source data and generating initial optimization features; screening and optimizing the initial optimization features to generate target optimization features; fusing the multi-objective optimization algorithm with the DeepSeek large model to generate a multi-objective DeepSeek model; inputting the target optimization features into the multi-objective DeepSeek large model for training and verification to generate an optimized multi-objective DeepSeek large model; during the training and verification process, an initial energy optimization strategy set is generated through the iteration of the multi-objective optimization algorithm, and the optimal energy usage strategy is selected from the initial energy optimization strategy set through the DeepSeek large model.
[0009] As an improvement to the above scheme, the step of fusing the multi-objective optimization algorithm with the DeepSeek large model to generate the multi-objective DeepSeek large model includes: using a genetic algorithm to construct a genetic optimization population, and using a particle swarm optimization algorithm to construct a particle optimization population; embedding the DeepSeek large model into the multi-objective optimization algorithm to optimize the multi-objective optimization algorithm and generate the multi-objective DeepSeek large model. During the optimization process, the genetic optimization population and the particle optimization population share information and collaborate in evolution.
[0010] As an improvement to the above-mentioned solution, the DeepSeek-based campus energy management method also includes: acquiring real-time energy data of the campus; preprocessing the real-time energy data; extracting energy features from the preprocessed real-time energy data, and converting the energy features into a target format; inputting the energy features converted into the target format into the DeepSeek large model to perform an in-depth analysis of the energy features and generate analysis results; parsing and processing the analysis results through the energy intelligent flow engine technology to generate real-time energy optimization recommendations and control strategies.
[0011] As an improvement to the above-mentioned solution, the DeepSeek-based campus energy management method also includes: obtaining the historical fault data, historical normal operation data and real-time operation data of the campus; inputting the historical fault data and historical normal operation data into the DeepSeek large model to construct a fault prediction model; inputting the real-time operation data of the campus into the fault prediction model to generate fault warning information.
[0012] As an improvement of the above solution, the DeepSeek-based campus energy management method also includes: obtaining query information; querying the historical energy consumption data and / or the optimal energy usage strategy based on the query information; and generating an energy report based on the historical energy consumption data and / or the optimal energy usage strategy.
[0013] As an improvement to the above scheme, the step of preprocessing the benchmark data includes: cleaning the benchmark data; normalizing the cleaned benchmark data; and performing feature extraction on the normalized benchmark data to generate key features, wherein the key features include peak load and seasonal changes.
[0014] Correspondingly, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the above-mentioned DeepSeek-based campus energy management method when executing the computer program.
[0015] Correspondingly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the above-mentioned DeepSeek-based campus energy management method are implemented.
[0016] The implementation of the present invention has the following beneficial effects:
[0017] The DeepSeek-based campus energy management method of the present invention significantly improves the accuracy of load forecasting by deploying a localized DeepSeek large model and training personalized domain models.
[0018] Furthermore, the DeepSeek-based campus energy management method of the present invention can utilize a multi-objective optimization algorithm and the DeepSeek large model to generate an optimal energy usage strategy under the premise of considering the optimization objectives, operating constraints, and energy load forecast results, thereby improving the scientific nature of the energy optimization strategy.
[0019] More preferably, the present invention realizes the intelligent management of the park energy system, reduces energy consumption costs and carbon emissions, and improves energy utilization efficiency; at the same time, the present invention also provides fault diagnosis and early warning functions, reducing energy waste and equipment loss; in addition, the present invention also improves user experience and management efficiency through intelligent interactive interfaces and personalized reports. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 This is a flow chart of a first embodiment of the DeepSeek-based campus energy management method of the present invention;
[0021] Figure 2 4 is a flow chart of a second embodiment of the campus energy management method based on DeepSeek of the present invention;
[0022] Figure 3 This is a flow chart of the third embodiment of the campus energy management method based on DeepSeek of the present invention. DETAILED DESCRIPTION
[0023] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be described in further detail below with reference to the accompanying drawings.
[0024] See also Figure 1 , Figure 1 A flowchart of a first embodiment of a campus energy management method based on DeepSeek is shown, which includes:
[0025] S101, obtain baseline data of the park energy system;
[0026] The benchmark data includes historical energy consumption data (such as electricity consumption data, gas consumption data, water consumption data, etc.), historical environmental data (such as temperature data, humidity data, weather data, etc.), real-time energy consumption data and real-time environmental data;
[0027] Accordingly, the benchmark data may be collected by sensors, smart meters and other collection devices, but is not limited thereto.
[0028] S102, preprocessing the benchmark data;
[0029] Furthermore, the steps of preprocessing the benchmark data include:
[0030] (1) Clean the benchmark data;
[0031] (2) normalizing the cleaned benchmark data;
[0032] (3) Perform feature extraction on the normalized benchmark data to generate key features;
[0033] It should be noted that key features include, but are not limited to, peak load and seasonal variations. Therefore, by cleaning, normalizing, and extracting features from the baseline data, a high-quality dataset can be generated and stored in a database for easy subsequent retrieval.
[0034] S103, building a DeepSeek large model;
[0035] Locally deploy the DeepSeek large model.
[0036] S104: Input the pre-processed historical energy consumption data and historical environmental data into the DeepSeek large model for training and verification to generate an optimized load forecasting model;
[0037] It should be noted that this embodiment can train and verify the DeepSeek large model through cross-validation to improve prediction accuracy. The specific steps include:
[0038] (1) Input a portion of pre-processed historical energy consumption data and historical environmental data into the DeepSeek large model for training to generate a load forecasting model;
[0039] By inputting the pre-processed historical energy consumption data and historical environmental data into the DeepSeek large model and performing personalized model training, a trained load forecasting model can be obtained.
[0040] (2) Inputting another portion of pre-processed historical energy consumption data and historical environmental data into the load forecasting model for verification to optimize the load forecasting model.
[0041] During the validation process, the prediction accuracy can be optimized by evaluating the model performance and adjusting the model parameters.
[0042] S105: Input the pre-processed real-time energy consumption data and real-time environmental data into the optimized load forecasting model to generate energy consumption and load forecasting results.
[0043] By using real-time energy consumption data and real-time environmental data as input to the load forecasting model, the future energy consumption load forecast results can be obtained.
[0044] Therefore, the campus energy management method based on DeepSeek of the present invention can utilize the powerful data processing and prediction capabilities of the DeepSeek large model to build a personalized campus energy load prediction model, and achieve high-precision prediction of future energy loads by training the model with historical data.
[0045] See also Figure 2 , Figure 2 A flow chart of a second embodiment of the DeepSeek-based campus energy management method of the present invention is shown, which includes:
[0046] S201, obtaining baseline data of the park energy system;
[0047] S202, preprocessing the benchmark data;
[0048] S203, building a DeepSeek large model;
[0049] S204: Input the pre-processed historical energy consumption data and historical environmental data into the DeepSeek large model for training and verification to generate an optimized load forecasting model;
[0050] S205: Input the pre-processed real-time energy consumption data and real-time environmental data into the optimized load forecasting model to generate energy consumption and load forecasting results.
[0051] S206, constructing an optimization target;
[0052] Optimization goals include, but are not limited to, reducing energy costs, improving energy efficiency, and reducing carbon emissions;
[0053] S207, constructing operation constraints;
[0054] Constraint adjustments include, but are not limited to, equipment capacity limitations, energy price fluctuations, and environmental protection policy requirements;
[0055] S208, inputting the optimization objectives, operation constraints and energy load forecast results into a pre-trained multi-objective DeepSeek model to generate an optimal energy use strategy;
[0056] It should be noted that the multi-objective DeepSeek model combines the global search capability of the genetic algorithm and the local convergence characteristics of the particle swarm optimization algorithm, and can efficiently find the optimal solution under complex constraints.
[0057] Accordingly, the training steps of the multi-target DeepSeek model include:
[0058] (1) Obtain multi-source data on the park’s energy system;
[0059] Multi-source data includes historical energy consumption data, historical environmental data, real-time energy consumption data, real-time environmental data, and equipment operating status data;
[0060] (2) Preprocessing of multi-source data;
[0061] Similarly, the steps for preprocessing multi-source data include:
[0062] (2.1) Cleaning and processing multi-source data;
[0063] (2.2) Normalizing the cleaned multi-source data;
[0064] (2.3) Perform feature extraction on the normalized multi-source data to generate optimized features.
[0065] (3) Input the preprocessed multi-source data into the DeepSeek large model, perform dimensionality reduction and feature extraction on the multi-source data, and generate initial optimized features;
[0066] By utilizing the feature extraction capability of the DeepSeek large model, high-dimensional optimization features are reduced in dimensionality to extract initial optimization features related to energy optimization.
[0067] (4) Screening and optimizing the initial optimization features to generate target optimization features;
[0068] Combining domain knowledge and expert experience, the initial optimized features are screened and optimized to further improve the quality and relevance of the data.
[0069] (5) Fusion of the multi-objective optimization algorithm with the DeepSeek large model to generate a multi-objective DeepSeek model;
[0070] Specifically, the steps of fusing the multi-objective optimization algorithm with the DeepSeek large model to generate the multi-objective DeepSeek large model include:
[0071] (5.1) Genetic algorithm is used to construct genetic optimization population, and particle swarm optimization algorithm is used to construct particle optimization population;
[0072] This method uses a genetic algorithm and a particle swarm optimization algorithm to initialize two populations, each representing a possible energy optimization strategy. The genetic algorithm continuously optimizes individuals within a population by simulating the selection, crossover, and mutation operations of biological evolution. The particle swarm optimization algorithm simulates the social behavior of flocks of birds or fish, leveraging information sharing between particles to find the optimal solution.
[0073] (5.2) embedding the DeepSeek large model into the multi-objective optimization algorithm to optimize the multi-objective optimization algorithm and generate the multi-objective DeepSeek large model;
[0074] The DeepSeek large model is embedded in the multi-objective optimization algorithm, leveraging its ability to learn and understand large amounts of historical energy management data to provide guidance for the optimization algorithm.
[0075] Specifically, the DeepSeek big model can model and analyze the potential relationship between energy usage patterns and optimization objectives, generating a fitness value for each individual (energy optimization strategy) that reflects the individual's comprehensive performance on multiple optimization objectives (such as cost, efficiency, and environmental protection).
[0076] During the optimization process, the genetic optimization population and the particle optimization population share information and co-evolve.
[0077] In other words, the individuals of the genetic algorithm can pass on excellent genes to their offspring through crossover and mutation operations, and can also learn some excellent characteristics from the individuals of the particle swarm optimization algorithm; the particles of the particle swarm optimization algorithm can adjust their flight direction and speed according to the individual distribution of the genetic algorithm, so as to converge to the optimal solution more quickly.
[0078] (6) Inputting the target optimization features into the multi-target DeepSeek large model for training and verification to generate an optimized multi-target DeepSeek large model;
[0079] During the training process, the multi-objective DeepSeek large model can be trained using the training data set. By adjusting the model parameters (such as the selection probability, crossover probability and mutation probability of the genetic algorithm, the learning factor and inertia weight of the particle swarm optimization algorithm, etc.), the model can better fit the data and improve the optimization performance.
[0080] During the verification process, cross-validation and other methods can be used to verify and evaluate the performance of the trained multi-objective DeepSeek large model. The evaluation indicators include prediction accuracy, optimization effect, convergence speed, etc. By comparing with traditional multi-objective optimization algorithms or other single algorithms, the advantages and effectiveness of the proposed method in generating energy optimization strategies are verified.
[0081] At the same time, during the training and verification process, the initial energy optimization strategy set can be generated through the iteration of the multi-objective optimization algorithm, and the optimal energy usage strategy can be selected from the initial energy optimization strategy set through the DeepSeek large model.
[0082] It's important to note that through the continuous iteration of the multi-objective optimization algorithm, a series of non-dominated solutions, known as the Pareto optimal solution set, is generated. These solutions represent different energy optimization strategies that achieve the best balance between multiple optimization objectives. The DeepSeek large-scale model can select the most appropriate energy optimization strategy from the Pareto optimal solution set based on the actual needs and preferences of the park, or provide decision makers with multiple alternative options to help them make more informed decisions.
[0083] Accordingly, the DeepSeek large-scale model generates an optimal energy usage strategy based on a multi-objective optimization algorithm, effectively balancing cost, efficiency, and environmental protection requirements. This optimal energy usage strategy can include, but is not limited to, equipment scheduling and energy allocation.
[0084] S209, controlling the park energy system according to the optimal energy usage strategy to adjust the equipment scheduling and energy distribution of the park in real time.
[0085] and Figure 1 Unlike the first embodiment shown, in this embodiment, the optimization objectives, operating constraints and energy load forecast results are input into the DeepSeek large model to generate the optimal energy usage strategy, and then the optimal energy usage strategy is applied to the park energy system to adjust the equipment scheduling and energy distribution of the park in real time.
[0086] Furthermore, this embodiment can also optimize and dynamically adjust the optimal energy usage strategy in real time:
[0087] (1) Real-time data processing and analysis: During the actual operation of the park energy management system, new multi-source data is collected in real time, and the real-time data analysis capabilities of the DeepSeek large model are used to quickly process and analyze the multi-source data to timely capture changes and anomalies in energy usage patterns.
[0088] (2) Dynamic adjustment and optimization strategy update: Based on the results of real-time data analysis and combined with a multi-objective optimization algorithm, the energy optimization strategy is dynamically adjusted. If it is found that the current strategy deviates from the expected optimization goals or new energy demand changes occur, the optimization process can be restarted in time to generate a new energy optimization strategy. This strategy can then be applied to the campus energy management system to achieve real-time optimization and dynamic adjustment of energy use.
[0089] Therefore, the campus energy management method based on DeepSeek of the present invention can utilize the DeepSeek large model to learn a large amount of multi-source data, accurately extract the potential relationship between energy usage patterns and optimization goals, and provide intelligent guidance for the multi-objective optimization algorithm, so that the generated strategy is more in line with the actual operation needs of the campus. Compared with the traditional single optimization algorithm, it can better balance multiple objectives such as cost, efficiency and environmental protection, and provide more innovative and adaptable energy usage strategies.
[0090] See also Figure 3 , Figure 3 A flowchart of a third embodiment of the DeepSeek-based campus energy management method of the present invention is shown, which includes:
[0091] S301, obtaining baseline data of the park energy system;
[0092] Benchmark data includes historical energy consumption data, historical environmental data, real-time energy consumption data, real-time environmental data, and equipment operating status data;
[0093] S302, preprocessing the benchmark data;
[0094] S303, building a DeepSeek large model;
[0095] S304: Input the pre-processed historical energy consumption data and historical environmental data into the DeepSeek large model for training and verification to generate an optimized load forecasting model;
[0096] S305: Input the pre-processed real-time energy consumption data and real-time environmental data into the optimized load forecasting model to generate energy consumption and load forecasting results.
[0097] S306, constructing optimization objectives;
[0098] S307, constructing operation constraints;
[0099] S308, inputting the optimization objectives, operation constraints and energy load forecast results into the pre-trained multi-objective DeepSeek model to generate the optimal energy use strategy;
[0100] S309, controlling the park energy system according to the optimal energy usage strategy to adjust the equipment scheduling and energy distribution of the park in real time.
[0101] S310, obtains real-time energy data of the park;
[0102] Real-time campus energy data is collected from a variety of data sources, including sensor networks, smart electricity meters, water meters, temperature and humidity sensors, light sensors, leak detection sensors, pipeline pressure sensors, and other devices. These data sources generate large, continuous data streams, and the Energy Intelligent Flow Engine technology ensures uninterrupted data collection and transmission.
[0103] S311, preprocessing real-time energy data;
[0104] Remove noisy data and outliers to improve data quality.
[0105] For example, filtering algorithms can be used to remove noise from sensor data, or interpolation methods can be used to fill in missing data points.
[0106] S312, extracting energy features from the preprocessed real-time energy data and converting the energy features into a target format;
[0107] Energy characteristics related to energy management (such as power, energy consumption, temperature, etc.) are extracted from real-time energy data and converted into a format suitable for analysis.
[0108] For example, extract periodic features, trend features, etc. from time series data, or perform structured processing, normalization, and other operations on the data to ensure that the data meets the input requirements of the model.
[0109] S313, inputting the energy characteristics converted into the target format into the DeepSeek large model to perform in-depth analysis on the energy characteristics and generate analysis results;
[0110] The converted energy characteristics are input into the DeepSeek large model, and its powerful computing power and ability to recognize complex patterns are used to conduct in-depth analysis of real-time data (for example, calculating real-time energy consumption, power changes and other indicators).
[0111] The conversion process may involve complex computing tasks (such as time window aggregation, association analysis, etc.), which can process and respond to data in real time.
[0112] DeepSeek's large model can identify hidden patterns and trends in data, providing a basis for energy optimization strategies.
[0113] S314, analyzes and processes the analysis results through the energy intelligent flow engine technology to generate real-time energy optimization suggestions and control strategies.
[0114] The output of the DeepSeek large model is parsed and processed by the Energy Intelligence Flow Engine technology, generating real-time energy optimization recommendations and control strategies. These strategies are fed back to the energy management system in real time to adjust energy distribution and equipment operating status, achieving dynamic energy optimization and real-time adjustment. Therefore, with the continuous input of data streams and the continuous operation of the model, the combination of engine technology and the DeepSeek large model continuously accumulates operational data and experience. Furthermore, by analyzing this data and retraining the model, the performance of this domain intelligent model and the accuracy of its strategies can be continuously optimized, enabling the energy management system to better adapt to changes in the campus' energy needs.
[0115] In other words, this invention combines data stream processing technology with the DeepSeek large-scale model to collect campus energy data in real time. It then performs rapid preprocessing and feature extraction, converting the data stream into a format suitable for large-scale model analysis. It then leverages the efficient computing architecture and parallel processing capabilities of the DeepSeek large-scale model to analyze the data in real time and generate optimization recommendations. Compared to traditional methods, this invention can process large amounts of real-time data in seconds, promptly capturing subtle changes in energy demand, enabling more accurate and rapid energy strategy adjustments, and effectively improving the real-time responsiveness and operational efficiency of the energy management system.
[0116] Therefore, the present invention can dynamically optimize the optimal energy usage strategy to adapt to changes in energy demand through the real-time data analysis capabilities of the DeepSeek large model.
[0117] S315, obtaining historical fault data, historical normal operation data and real-time operation data of the park.
[0118] S316, inputting historical fault data and historical normal operation data into the DeepSeek large model to construct a fault prediction model;
[0119] S317, inputting the real-time operation data of the park into the fault prediction model to generate fault warning information.
[0120] For fault diagnosis, a localized DeepSeek large-scale model conducts in-depth learning on massive amounts of historical fault data and normal operation data, automatically extracting characteristic information from energy system equipment and building a precise fault pattern library. During real-time monitoring, the trained domain model can quickly compare current operating data with the characteristics in the fault pattern library, promptly identifying potential fault signs. Furthermore, multi-source data fusion analysis combined with sensor network data further improves the accuracy of fault diagnosis.
[0121] Furthermore, an intelligent early warning system can be established to detect and deal with problems in advance, reducing energy waste and equipment loss.
[0122] Compared with traditional fault diagnosis methods based on thresholds or machine learning models based on artificial features, the present invention can discover the early characteristics of equipment failures, provide early warning of potential risks of equipment at various points in the park, effectively reduce energy waste and equipment loss caused by failures, and ensure the stable operation of the park's energy system.
[0123] Therefore, the present invention can monitor the operating status of the park energy system in real time, identify potential faults and provide early warnings through the fault diagnosis capabilities of the DeepSeek large model.
[0124] S318, obtaining query information;
[0125] It should be noted that users can utilize an intelligent interactive interface based on natural language processing to upload query information via voice or text, thereby realizing an intelligent user interactive service mode.
[0126] S319, querying historical energy usage data and / or optimal energy usage strategy based on the query information;
[0127] S320: Generate an energy report based on historical energy usage data and / or optimal energy usage strategy.
[0128] Therefore, the present invention can help users better manage energy consumption by automatically generating energy reports.
[0129] In summary, the DeepSeek-based campus energy management method of the present invention significantly improves the accuracy of load forecasting and the scientific nature of energy optimization strategies by deploying a localized DeepSeek large model and training personalized domain models; further, the present invention realizes the intelligent management of the campus energy system, reduces energy consumption costs and carbon emissions, and improves energy utilization efficiency; at the same time, the present invention also provides fault diagnosis and early warning functions, reducing energy waste and equipment loss; in addition, the present invention also improves user experience and management efficiency through an intelligent interactive interface and personalized reports.
[0130] Correspondingly, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the above-mentioned DeepSeek-based campus energy management method when executing the computer program.
[0131] At the same time, the present invention also discloses a computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the above-mentioned DeepSeek-based campus energy management method are implemented.
[0132] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.
Claims
1. A campus energy management method based on DeepSeek, characterized in that: include: Obtaining baseline data for the park's energy system, including historical energy usage data, historical environmental data, real-time energy usage data, and real-time environmental data; Preprocessing the benchmark data; Build a large DeepSeek model; Input the pre-processed historical energy consumption data and historical environmental data into the DeepSeek large model for training and verification to generate an optimized load forecasting model; The pre-processed real-time energy consumption data and real-time environmental data are input into the optimized load forecasting model to generate energy consumption load forecast results.
2. The campus energy management method based on DeepSeek according to claim 1, characterized in that: Also includes: Establishing optimization goals, including reducing energy consumption costs, improving energy efficiency, and reducing carbon emissions; Establishing operational constraints, including equipment capacity limitations, energy price fluctuations, and environmental policy requirements; Input the optimization objectives, operating constraints, and energy load forecast results into a pre-trained multi-objective DeepSeek model to generate an optimal energy usage strategy; The park energy system is controlled according to the optimal energy utilization strategy to adjust the equipment scheduling and energy distribution of the park in real time.
3. The DeepSeek-based campus energy management method according to claim 2, characterized in that: The training steps of the multi-target DeepSeek model include: Acquire multi-source data of the park energy system, including historical energy consumption data, historical environmental data, real-time energy consumption data, real-time environmental data, and equipment operating status data; Preprocessing the multi-source data; Input the pre-processed multi-source data into the DeepSeek large model, perform dimensionality reduction and feature extraction on the multi-source data and generate initial optimized features; Screening and optimizing the initial optimization features to generate target optimization features; Fusion of the multi-objective optimization algorithm with the DeepSeek large model to generate a multi-objective DeepSeek model; Inputting the target optimization features into the multi-target DeepSeek large model for training and verification to generate an optimized multi-target DeepSeek large model; During the training and verification process, an initial energy optimization strategy set is generated through iteration of the multi-objective optimization algorithm, and the optimal energy use strategy is selected from the initial energy optimization strategy set through the DeepSeek large model.
4. The DeepSeek-based campus energy management method according to claim 3, characterized in that: The step of fusing the multi-objective optimization algorithm with the DeepSeek large model to generate the multi-objective DeepSeek large model includes: Genetic algorithm is used to construct genetic optimization population, and particle swarm optimization algorithm is used to construct particle optimization population; The DeepSeek large model is embedded in a multi-objective optimization algorithm to optimize the multi-objective optimization algorithm and generate the multi-objective DeepSeek large model. During the optimization process, information sharing and collaborative evolution are performed between the genetic optimization population and the particle optimization population.
5. The DeepSeek-based campus energy management method according to claim 2, characterized in that: The DeepSeek-based campus energy management method further includes: Obtain real-time energy data for the park; Preprocessing the real-time energy data; extracting energy features from the preprocessed real-time energy data and converting the energy features into a target format; Inputting the energy signature converted into the target format into the DeepSeek large model to perform in-depth analysis on the energy signature and generate analysis results; The analysis results are parsed and processed through energy intelligent flow engine technology to generate real-time energy optimization suggestions and control strategies.
6. The DeepSeek-based campus energy management method according to claim 1, characterized in that: The DeepSeek-based campus energy management method further includes: Obtain historical fault data, historical normal operation data and real-time operation data of the park; Inputting the historical fault data and the historical normal operation data into the DeepSeek large model to construct a fault prediction model; The real-time operation data of the park is input into the fault prediction model to generate fault warning information.
7. The DeepSeek-based campus energy management method according to claim 1 or 2, characterized in that: Also includes: Get query information; querying the historical energy usage data and / or the optimal energy usage strategy according to the query information; An energy report is generated based on the historical energy usage data and / or the optimal energy usage strategy.
8. The campus energy management method based on DeepSeek according to claim 1, characterized in that: The step of preprocessing the benchmark data comprises: performing cleaning processing on the benchmark data; performing normalization processing on the cleaned benchmark data; The normalized benchmark data is subjected to feature extraction to generate key features, wherein the key features include peak load and seasonal variation.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the DeepSeek-based campus energy management method described in any one of claims 1 to 8 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the DeepSeek-based campus energy management method according to any one of claims 1 to 8 are implemented.