Shelter power demand response method and system based on artificial intelligence algorithm

Abstract

The invention provides a shelter power demand response method and system based on an artificial intelligence algorithm. The method comprises the following steps: collecting real-time power consumption data of each electric device in the shelter, and generating an original data set; performing feature selection based on an artificial intelligence algorithm, and generating a comprehensive power demand data set; on the basis of the comprehensive power demand data set, demand fluctuation is predicted by using a space-time diagram neural network structure search algorithm, result uncertainty is evaluated by using a Monte Carlo simulation technology, and a prediction evaluation result is generated; based on the prediction evaluation result, generating an optimized power utilization strategy scheme by applying a self-adaptive chaos search algorithm and combining a dynamic planning technology optimization strategy; and based on the optimized power utilization strategy scheme, monitoring actual deviation, recording environmental parameters for feedback, and generating a power demand response model. According to the technical scheme provided by the invention, the utilization efficiency of electric power resources can be improved, and intelligent management is realized.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of smart grid technology, and more particularly to a method and system for responding to electricity demand in a modular cabin based on an artificial intelligence algorithm. Background Art

[0002] With the widespread application of shelters in the fields of medical treatment, disaster relief, etc., the complexity and importance of internal power demand management have become increasingly prominent. As a temporary or emergency living and working place, shelters require a method that can monitor and predict power demand in real time and optimize power usage strategies accordingly to ensure the stability and efficiency of power supply. Traditional power management systems are often unable to cope with such complex and changing demands. Therefore, there is an urgent need for a power demand response method based on advanced artificial intelligence algorithms to improve the utilization efficiency of power resources, ensure user comfort, and achieve intelligent management.

[0003] At present, the power demand management of shelters mainly relies on manual adjustments and simple automated control systems. These systems usually monitor power consumption and environmental parameters through sensors based on preset rules and thresholds, and then start and stop the equipment according to preset logic. Although these methods can meet basic power needs to a certain extent, their response speed and accuracy are insufficient when faced with complex and changing environments and needs.

[0004] Traditional methods rely on fixed rules and thresholds, and are unable to adjust power consumption strategies in real time, making it difficult to cope with sudden changes in power demand. Existing systems lack effective forecasting models and are unable to accurately predict future fluctuations in power demand, resulting in irrational resource allocation and serious energy waste. Traditional optimization methods usually only consider a single goal, such as energy efficiency ratio or user comfort, and are difficult to achieve comprehensive optimization of multiple goals, resulting in poor overall system performance. Existing power management systems mostly rely on manual intervention, lack intelligent self-learning and optimization capabilities, and are unable to adapt to complex and changing environmental requirements. Summary of the invention

[0005] The embodiments of the present application provide a method and system for responding to electricity demand in a cabin based on an artificial intelligence algorithm, so as to solve the problems of insufficient response speed and accuracy in the prior art.

[0006] In a first aspect, an embodiment of the present application provides a method for responding to electricity demand in a shelter based on an artificial intelligence algorithm, comprising:

[0007] Collect the real-time power consumption data of each electrical device in the shelter and combine it with the environmental parameter data to generate the original data set;

[0008] Based on artificial intelligence algorithms, the original data set is preprocessed to perform feature selection and dimensionality reduction processing to generate a comprehensive power demand data set;

[0009] Based on the comprehensive power demand data set, a spatiotemporal graph neural network structure search algorithm is used to predict future power demand fluctuations, and a Monte Carlo simulation technique is used to evaluate the uncertainty of the prediction results to generate a prediction evaluation result;

[0010] Based on the prediction and evaluation results, an adaptive chaos search algorithm is used in combination with dynamic programming technology to optimize the power consumption strategy inside the shelter, balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy plan;

[0011] Based on the optimized electricity consumption strategy plan, the deviation between actual electricity consumption and expected effects is monitored, environmental parameters are recorded during the implementation of the plan, feedback and iterative optimization are performed, and an electricity demand response model is generated.

[0012] Optionally, based on the comprehensive power demand data set, using a spatiotemporal graph neural network structure search algorithm to predict future power demand fluctuations, using a Monte Carlo simulation technique to evaluate the uncertainty of the prediction results, and generating a prediction evaluation result, including:

[0013] Based on the comprehensive power demand data set, the time series data and spatial distribution information are obtained, and the spatiotemporal graph structure is generated by combining the spatiotemporal relationship of power consumption inside the shelter;

[0014] Based on the spatiotemporal graph structure, a spatiotemporal graph neural network structure search algorithm is used to automatically search for the optimal network structure, maximize the ability to predict future power demand fluctuations, and generate spatiotemporal fluctuation prediction results;

[0015] Based on the spatiotemporal fluctuation prediction results, the Monte Carlo simulation technique is used to perform multiple random samplings to evaluate the uncertainty range of the prediction results and generate uncertainty assessments;

[0016] Based on the uncertainty assessment, historical data and current environmental parameters are combined to identify potential influencing factors and generate predictive assessment results.

[0017] Optionally, based on the spatiotemporal graph structure, a spatiotemporal graph neural network structure search algorithm is used to automatically search for the best network structure, maximize the ability to predict future power demand fluctuations, and generate spatiotemporal fluctuation prediction results, including:

[0018] Based on the space-time graph structure, the basic architecture of the space-time graph neural network is initialized, the initial network layer configuration is performed, and an initialized space-time graph neural network is generated;

[0019] Based on the initialized spatiotemporal graph neural network, define the network structure search space, set the number of network layers and activation function type, and generate the search space configuration;

[0020] Based on the search space configuration, the space-time graph neural network structure search algorithm is used to automatically search for the best network structure, maximize the ability to predict future power demand fluctuations, and gradually optimize the network structure through repeated training and verification to generate an optimized space-time graph neural network;

[0021] Based on the optimized spatiotemporal graph neural network, deep learning processing is performed on the spatiotemporal graph structure to extract the future trend of electricity demand changes and generate spatiotemporal fluctuation prediction results.

[0022] Optionally, based on the search space configuration, a spatiotemporal graph neural network structure search algorithm is used to automatically search for the best network structure, maximize the ability to predict future power demand fluctuations, and gradually optimize the network structure through repeated training verification to generate an optimized spatiotemporal graph neural network, including:

[0023] Initializing a network structure based on the search space configuration;

[0024] Using forward propagation to input the spatiotemporal graph structure, obtaining the predicted label of each sample, and recording the network weights and bias items to generate a fitness score of the network structure;

[0025] The fitness score of the network structure is calculated through the following steps:

[0026]

[0027] Among them, A(W,X) is the fitness score of the network structure; W is the network weight matrix; X is the input spatiotemporal graph structure; y i is the true label of the i-th sample; is the predicted label of the i-th sample; w j is the weight of the jth layer; K is the number of bias terms; b k is the kth bias term; λ is the regularization parameter; i is the index of the training sample, from 1 to N; N is the number of training samples; j is the index of the network layer, from 1 to L; L is the number of network layers; k is the index of the bias term, from 1 to K; K is the number of bias terms;

[0028] Based on the fitness score of the network structure, a nonlinear transformation optimization mechanism is introduced, and an optimization objective function is constructed by combining the network prediction error and the weight size to generate an optimization objective function value;

[0029] The optimization objective function value is calculated using the following formula:

[0030]

[0031] Among them, O(W,X,t) is the optimization objective function value; W is the network weight matrix; X is the input spatiotemporal graph structure; A(W,X) is the fitness score of the network structure; yi is the true label of the i-th sample; is the predicted label of the i-th sample; w j is the weight of the jth layer; K is the number of bias terms; b k is the kth bias term; α is the chaotic search strength coefficient; β is the error sensitivity coefficient; γ is the weight regularization strength coefficient; t is the current iteration number; i is the index of the training sample, from 1 to N; N is the number of training samples; j is the index of the network layer, from 1 to L; L is the number of network layers; k is the index of the bias term, from 1 to K; K is the number of bias terms;

[0032] Based on the optimization objective function value, through multiple rounds of iterative optimization, the network weight matrix and bias terms are gradually adjusted, and the network structure is updated to minimize the optimization objective function value and generate an optimized spatiotemporal graph neural network.

[0033] Optionally, the method of performing multiple random samplings based on the spatiotemporal fluctuation prediction results using Monte Carlo simulation technology to evaluate the uncertainty range of the prediction results and generate an uncertainty assessment includes:

[0034] Based on the spatiotemporal fluctuation prediction results, identifying probability distribution of prediction model output variables and generating a prediction value probability distribution model;

[0035] Based on the predicted value probability distribution model, the number of Monte Carlo simulation iterations is set, random sampling is performed to ensure the randomness and representativeness of the input parameters, and random samples are generated;

[0036] Based on the random samples, the prediction model is iterated multiple times, all iteration results are collected, and a prediction result set is generated;

[0037] Based on the set of prediction results, statistical key features are identified, the uncertainty range of the set of prediction results is analyzed, and an uncertainty assessment is generated.

[0038] Optionally, based on the prediction and evaluation results, an adaptive chaos search algorithm is used in combination with dynamic programming technology to optimize the power consumption strategy inside the shelter, balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy solution, including:

[0039] Based on the forecast evaluation results, analyze the time distribution characteristics, evaluate the impact of uncertainty, determine the key influencing factors, and generate basic input data;

[0040] Based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the electricity consumption strategy inside the shelter, and a multi-stage multi-objective optimization process is performed to generate a preliminary electricity consumption strategy set;

[0041] Based on the preliminary power consumption strategy set, combined with dynamic programming technology, the premise of ensuring that power demand is met, balancing energy efficiency maximization and comfort optimization, and generating an optimized power consumption strategy set;

[0042] Based on the optimized electricity consumption strategy set, a multi-criteria decision analysis method is used for screening, electricity consumption flexibility is comprehensively considered, and an optimized electricity consumption strategy solution is generated.

[0043] Optionally, based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the electricity consumption strategy inside the shelter, perform multi-stage multi-objective optimization processing, and generate a preliminary electricity consumption strategy set, including:

[0044] Based on the basic input data, an adaptive model of the optimization problem is constructed, the optimization objectives and constraints are defined, and an adaptive optimization model is generated;

[0045] Based on the adaptive optimization model and in combination with the operating parameter range of each electrical equipment in the shelter, a preset number of individual samples are randomly selected to generate an initial population individual set;

[0046] Based on the initial population individual set, an adaptive chaos search algorithm is used to dynamically adjust the search strategy, simulate the dynamic behavior of the chaotic system, enrich the optimal solution space, and generate a new population individual set;

[0047] Based on the new population individual set, through multi-stage iterative processing, the fitness of candidate strategies is traversed and evaluated, and combined with the Pareto optimization technology, the search scope is gradually narrowed to generate a preliminary electricity consumption strategy set.

[0048] Optionally, based on the initial population individual set, an adaptive chaos search algorithm is used to dynamically adjust the search strategy, simulate the dynamic behavior of the chaotic system, enrich the optimal solution space, and generate a new population individual set, including:

[0049] Based on the initial population individual set, evaluating the fitness value of each individual;

[0050] By analyzing the distribution of individuals in the population, determining the center point of the current iteration, obtaining the average fitness value of the current population, and setting parameters related to population diversity, a search strategy is generated;

[0051] The search strategy is calculated using the following formula:

[0052]

[0053] Where S(t+1) is the search strategy of the t+1th iteration; S(t) is the search strategy of the tth iteration; C(t) is the center point of the tth iteration; η is the learning rate; σ is the chaotic search intensity coefficient; t is the current iteration number; T is the total iteration number; ω is the nonlinear weight coefficient; k is the nonlinear adjustment coefficient; F i (t) is the fitness value of the i-th individual at the t-th iteration; F avg (t) is the average fitness value of the tth iteration;

[0054] Based on the search strategy, combined with the current individual position, nonlinear adjustment items and search strategy influencing items are introduced to enhance the dynamic adjustment capability, ensure the robustness and effectiveness of the optimization process, and generate the population fitness value;

[0055] The population fitness value is calculated using the following formula:

[0056]

[0057] Among them, F i (t+1) is the population fitness value of the i-th individual at the t+1th iteration; F i (t) is the population fitness value of the i-th individual at the t-th iteration; F best (t) is the optimal population fitness value of the tth iteration; α is the step size factor; β is the nonlinear adjustment coefficient; X ij (t) is the position of the i-th individual in the j-th dimension; X best,j (t) is the position of the optimal solution in the jth dimension; t is the current iteration number; S(t+1) is the search strategy for the t+1th iteration; θ is the influence coefficient of the search strategy; j is the index of the individual dimension, from 1 to D; D is the number of individual dimensions; S(t+1) is the search strategy for the t+1th iteration;

[0058] Based on the population fitness value, combined with the Pareto optimization technology, the search range is gradually narrowed, and the population diversity is enriched through crossover and mutation operations to converge the optimization process and generate a new set of population individuals.

[0059] Optionally, the optimized power consumption strategy scheme, monitoring the deviation between actual power consumption and expected effect, recording environmental parameters during the implementation of the scheme, performing feedback and iterative optimization, and generating a power demand response model include:

[0060] Based on the optimized power consumption strategy, adjust the power consumption strategy inside the shelter, monitor the actual power consumption, record the environmental parameters during the implementation, and generate actual operation parameter records;

[0061] Based on the actual operating parameter records, the actual power consumption is compared with the expected effect, multi-dimensional deviation results are calculated, and a deviation analysis report is generated;

[0062] Based on the deviation analysis report, feedback and iterative optimization are performed, the electricity consumption strategy inside the shelter is evaluated and optimized multiple times, the actual operating parameters are gradually adjusted, and the optimization iterative results are generated;

[0063] Based on the optimization iteration results, effective strategies in the optimization iteration process are integrated to improve the accuracy of the prediction model and the effectiveness of the response measures, and generate a power demand response model.

[0064] In a second aspect, the embodiment of the present application provides a shelter power demand response system based on an artificial intelligence algorithm, comprising:

[0065] The collection module is used to collect the real-time power consumption data of each electrical device in the shelter and generate the original data set by combining it with the environmental parameter data;

[0066] A processing module, used to pre-process the original data set based on an artificial intelligence algorithm, perform feature selection and dimensionality reduction processing, and generate a comprehensive power demand data set;

[0067] A prediction module, which is used to predict future power demand fluctuations based on the comprehensive power demand data set by using a spatiotemporal graph neural network structure search algorithm, and to use a Monte Carlo simulation technique to evaluate the uncertainty of the prediction results and generate a prediction evaluation result;

[0068] An optimization module is used to optimize the power consumption strategy inside the shelter based on the prediction and evaluation results, using an adaptive chaos search algorithm combined with dynamic programming technology, balancing energy efficiency and comfort, and generating an optimized power consumption strategy plan;

[0069] The monitoring module is used to monitor the deviation between actual power consumption and expected effect based on the optimized power consumption strategy plan, record environmental parameters during the implementation of the plan, perform feedback and iterative optimization, and generate a power demand response model.

[0070] In an embodiment of the present application, real-time power consumption data of each electrical equipment inside the cabin is collected and combined with environmental parameter data to generate an original data set; based on an artificial intelligence algorithm, the original data set is preprocessed, feature selection and dimensionality reduction are performed, and a comprehensive power demand data set is generated; based on the comprehensive power demand data set, a spatiotemporal graph neural network structure search algorithm is used to predict future power demand fluctuations, and a Monte Carlo simulation technique is used to evaluate the uncertainty of the prediction results to generate a prediction evaluation result; based on the prediction evaluation result, an adaptive chaos search algorithm is used in combination with dynamic programming technology to optimize the power consumption strategy inside the cabin, balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy plan; based on the optimized power consumption strategy plan, the deviation between the actual power consumption and the expected effect is monitored, the environmental parameters during the implementation of the plan are recorded, feedback and iterative optimization are performed, and a power demand response model is generated.

[0071] The technical solution of this application has the following beneficial effects:

[0072] By accurately predicting future fluctuations in electricity demand and optimizing electricity use strategies with an adaptive chaos search algorithm, the efficiency of electricity resource utilization in the shelter is effectively improved, and unnecessary energy waste is reduced. Based on the application of the spatiotemporal graph neural network structure search algorithm, the system can flexibly respond to various environmental changes and uncertainties in electricity demand, enhancing the adaptability and stability of the system. By optimizing the electricity use strategy inside the shelter, while ensuring the energy efficiency ratio, the user's demand for comfort is also taken into account, achieving a good balance between energy conservation and comfort, and improving user satisfaction. The entire process, from data collection, preprocessing, demand prediction to strategy optimization and feedback iteration, uses advanced AI technology to achieve intelligent management of shelter electricity demand response and reduce the complexity and cost of manual intervention. The system helps promote the concept of green buildings and sustainable development, reduces carbon emissions through intelligent control and optimization, and has positive significance for the long-term operation of facilities such as shelters.

[0073] Furthermore, the optimal network structure is automatically found through the spatiotemporal graph neural network structure search algorithm, which significantly improves the prediction accuracy of future electricity demand fluctuations and provides a reliable basis for the rational allocation of power resources; the Monte Carlo simulation technology is used to evaluate the uncertainty range of the prediction results through multiple random sampling to ensure the reliability and stability of the prediction results and reduce the risks caused by prediction errors; the historical data and current environmental parameters are combined to identify the potential factors affecting the fluctuation of electricity demand, which provides a scientific basis for formulating more accurate electricity consumption strategies and enhances the adaptability and flexibility of the system; through accurate prediction and uncertainty assessment, the power resource management inside the cabin is optimized, unnecessary energy waste is reduced, and energy utilization efficiency is improved; on the basis of ensuring the stability of power supply, the actual needs of users are taken into consideration to improve the living comfort and satisfaction of users; this method helps to promote the concept of green buildings and sustainable development, and reduces carbon emissions through intelligent control and optimization, which has positive significance for the long-term operation of facilities such as cabins.

[0074] Furthermore, the power consumption strategy inside the shelter is optimized through adaptive chaos search algorithm and dynamic programming technology, which significantly improves the energy efficiency ratio, reduces energy waste, and improves energy utilization efficiency. In the process of optimizing the power consumption strategy, the user's demand for comfort is fully considered to ensure that while maximizing the energy efficiency ratio, good user comfort is maintained, thereby improving the user experience. Based on the prediction and evaluation results, the time distribution characteristics and uncertainty impact are analyzed, the key influencing factors are determined, and basic input data is generated, thereby improving the accuracy and reliability of the prediction. Through the multi-criteria decision analysis method, the power consumption flexibility is comprehensively considered to generate an optimized power consumption strategy plan, so that the system can flexibly respond to different power demands and environmental changes. The optimized power consumption strategy plan ensures the satisfaction of power demand, while balancing the energy efficiency ratio and comfort, and improving the management efficiency and economic benefits of power resources inside the shelter. This method helps to promote the concept of green buildings and sustainable development, and reduces carbon emissions through intelligent control and optimization, which has positive significance for the long-term operation of facilities such as shelters.

[0075] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0077] Figure 1A flowchart of a method for responding to electricity demand in a shelter based on an artificial intelligence algorithm provided in an embodiment of the present application;

[0078] Figure 2 A schematic diagram of the structure of a cabin power demand response system based on an artificial intelligence algorithm provided in an embodiment of the present application. DETAILED DESCRIPTION

[0079] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0080] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0081] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0082] Figure 1 A flowchart of a method for responding to electricity demand in a shelter based on an artificial intelligence algorithm is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0083] 101. Collect the real-time power consumption data of each electrical device in the shelter and combine it with the environmental parameter data to generate the original data set;

[0084] The real-time power consumption data of electrical equipment includes the real-time power consumption of each electrical equipment (such as air conditioners, lighting systems, medical equipment, etc.) inside the cabin, which is usually collected through smart meters or sensors to understand the power usage of each device; the environmental parameter data includes environmental parameters such as temperature, humidity, and light intensity inside the cabin, which are used to evaluate the impact of environmental conditions on power demand; the real-time power consumption data of electrical equipment and the environmental parameter data are integrated together to form a multidimensional data set containing timestamps for subsequent data processing and analysis.

[0085] First, smart meters and environmental sensors are installed inside the cabin to collect power consumption data and environmental parameter data of each electrical device in real time. Secondly, the data is transmitted to the central data processing system through the wireless communication module. The system performs preliminary cleaning and formatting of the data to ensure the integrity and accuracy of the data. Finally, a structured original data set is generated.

[0086] Assume that in a quarantine shelter, smart meters and environmental sensors are installed at various key locations, and power consumption data needs to be obtained; smart meters are installed on each electrical device, including air conditioners, lighting systems, medical equipment, etc.; smart meters record power consumption data every 15 minutes; environmental sensors, including temperature sensors, humidity sensors, and light intensity sensors, are installed in different areas inside the shelter; environmental sensors record environmental parameter data every 30 minutes; smart meters and environmental sensors transmit the collected data to the central data processing system in real time through wireless communication modules; after receiving the data, the central data processing system stores it in the database to ensure the integrity and security of the data; the system performs preliminary cleaning on the collected data to remove outliers and missing values; if a sensor does not return data for a period of time, the system will mark it as a missing value and process it; ensure that the timestamps of all data are consistent for subsequent data integration and analysis; integrate the power consumption data and environmental parameter data of the electrical equipment to form a multidimensional data set containing timestamps; each row of the data set represents a record at a point in time, including timestamp, device ID, power consumption, temperature, humidity, light intensity and other information.

[0087] Through the above steps, the generated original data set provides a basis for subsequent data processing and analysis, ensuring the integrity and accuracy of the data.

[0088] 102. Based on an artificial intelligence algorithm, preprocess the original data set, perform feature selection and dimensionality reduction processing, and generate a comprehensive power demand data set;

[0089] Artificial intelligence algorithms include machine learning and data mining technologies, which are used to process and analyze raw data sets and extract useful information; feature selection is to select the most influential features for predicting electricity demand from the raw data set, reduce the dimension of the data, and improve the training efficiency and prediction accuracy of the model; dimensionality reduction processing is to convert high-dimensional data into low-dimensional data through methods such as principal component analysis (PCA), further reduce the complexity of the data, and improve the interpretability and computational efficiency of the model; the comprehensive electricity demand data set is a data set that has undergone feature selection and dimensionality reduction processing, which contains the key features that best reflect changes in electricity demand and is used for subsequent prediction and optimization processing.

[0090] First, after collecting the original data set, artificial intelligence algorithms are used to preprocess the data. Second, feature selection methods are used to identify the features that have the greatest impact on power demand forecasting, such as the power consumption of certain key equipment and specific environmental parameters. Dimensionality reduction techniques such as principal component analysis are used to convert high-dimensional data into low-dimensional data, reducing data redundancy and complexity. Finally, a comprehensive power demand data set is generated to retain key information and improve data processing efficiency and model prediction accuracy.

[0091] Assume that in a modular laboratory, a data acquisition system is needed to record various data in the laboratory once an hour; smart meters are installed on the main experimental equipment in the laboratory to record power consumption data once an hour; environmental sensors are installed in different areas inside the laboratory to record environmental parameters such as temperature, humidity and light intensity once an hour; all collected data are transmitted to the central data processing system in real time through wireless communication modules; the system performs preliminary cleaning on the collected data to remove outliers and missing values ​​to ensure the integrity and accuracy of the data; ensure that the timestamps of all data are consistent for subsequent data integration and analysis; use feature selection methods to Analyze and score the importance to determine the most influential features for power demand forecasting; through analysis, it is found that power consumption, room temperature and humidity have the greatest impact on power demand; select these key features to reduce the dimension of the data and improve the training efficiency and prediction accuracy of the model; apply the principal component analysis method to convert the selected key features into two principal components; convert high-dimensional data into low-dimensional data to reduce data redundancy and complexity; the generated principal component data retains the main information of the original data while simplifying the data structure; integrate the principal component data after dimensionality reduction into a new data set, where each row represents a record at a time point, including timestamp, principal component 1 and principal component 2 and other information.

[0092] Through the above steps, the generated comprehensive electricity demand dataset can simplify the data structure, retain the most critical information for electricity demand forecasting, and provide high-quality data support for subsequent model training and optimization.

[0093] 103. Based on the comprehensive power demand data set, use the spatiotemporal graph neural network structure search algorithm to predict future power demand fluctuations, use Monte Carlo simulation technology to evaluate the uncertainty of the prediction results, and generate prediction evaluation results;

[0094] The comprehensive electricity demand dataset is a dataset that has undergone feature selection and dimensionality reduction processing. It contains the key features that best reflect changes in electricity demand and is used to predict future electricity demand. The spatiotemporal graph neural network structure search algorithm is a deep learning algorithm that automatically searches for the optimal network structure by constructing a spatiotemporal graph structure to improve the accuracy of predicting future electricity demand fluctuations. The Monte Carlo simulation technology is a statistical method that uses multiple random sampling to evaluate the uncertainty range of the prediction results to ensure the reliability of the prediction. The prediction evaluation result is a comprehensive evaluation report generated by combining the prediction results and uncertainty assessment, which is used to guide subsequent optimization processing.

[0095] Firstly, based on the comprehensive electricity demand dataset, a spatiotemporal graph structure is constructed, and the spatiotemporal graph neural network structure search algorithm is used to automatically search for the optimal network structure to improve the accuracy of prediction of future electricity demand fluctuations. Subsequently, Monte Carlo simulation technology is used to conduct multiple random sampling to evaluate the uncertainty range of the prediction results and ensure the reliability of the prediction. Finally, the generated prediction evaluation results not only provide the predicted value of future electricity demand, but also evaluate the uncertainty of the prediction, providing a reliable basis for subsequent optimization processing.

[0096] Optionally, the step 103 uses a spatiotemporal graph neural network structure search algorithm based on the comprehensive power demand data set to predict future power demand fluctuations, uses a Monte Carlo simulation technique to evaluate the uncertainty of the prediction result, and generates a prediction evaluation result, including:

[0097] Based on the comprehensive electricity demand data set, time series data and spatial distribution information are obtained, and the spatiotemporal relationship of electricity consumption inside the cabin is combined to generate a spatiotemporal graph structure; based on the spatiotemporal graph structure, the spatiotemporal graph neural network structure search algorithm is used to automatically search for the optimal network structure, maximize the future electricity demand fluctuation prediction ability, and generate spatiotemporal fluctuation prediction results; based on the spatiotemporal fluctuation prediction results, the Monte Carlo simulation technology is used to perform multiple random samplings to evaluate the uncertainty range of the prediction results and generate an uncertainty assessment; based on the uncertainty assessment, historical data and current environmental parameters are combined to identify potential influencing factors and generate prediction evaluation results.

[0098] In the embodiment of the present application, based on the comprehensive power demand data set, time series data and spatial distribution information are obtained, and the spatiotemporal relationship of power consumption inside the cabin is combined to generate a spatiotemporal graph structure; the spatiotemporal graph neural network structure search algorithm is used to automatically search for the optimal network structure to improve the ability to predict future power demand fluctuations; the Monte Carlo simulation technology is continued to be used to perform multiple random samplings to evaluate the uncertainty range of the prediction results; finally, historical data and current environmental parameters are combined to identify potential influencing factors and generate prediction evaluation results to provide a reliable basis for subsequent optimization processing.

[0099] Suppose that in a square cabin hospital, it is necessary to predict the fluctuations in electricity demand in the next few days in order to optimize the allocation of electricity resources; obtain a comprehensive electricity demand data set from the central data processing system, which includes the power consumption data and environmental parameter data of each electrical equipment in the square cabin in the past month; extract time series data and spatial distribution information from the data set, the time series data is recorded once every hour, and the spatial distribution information includes the specific location of each electrical equipment in the square cabin; technicians combine the time series data and spatial distribution information to generate a spatiotemporal graph structure; this graph structure not only reflects the changes in power consumption of each electrical equipment over time, but also captures the spatial relationship between devices, helping to understand the spatiotemporal distribution of power consumption; use the spatiotemporal graph neural network structure search algorithm to automatically search for the optimal network structure. Through multiple iterations, the algorithm gradually optimizes the network structure and improves the ability to predict future fluctuations in electricity demand. The generated spatiotemporal fluctuation prediction results provide hourly electricity demand forecasts for the next few days. Monte Carlo simulation technology is used to perform multiple random samplings to evaluate the uncertainty range of the prediction results. By simulating different scenarios, multiple prediction results are generated, and the standard deviation and confidence interval of these results are calculated to evaluate the reliability of the prediction. The generated uncertainty assessment report includes the predicted value of future electricity demand and its confidence interval, helping technicians understand the credibility of the prediction. Combining historical data and current environmental parameters, potential influencing factors are identified, such as weather changes, equipment failures, etc., which may affect fluctuations in electricity demand. The generated prediction evaluation results not only provide predicted values ​​for future electricity demand, but also evaluate the uncertainty of the prediction, providing a reliable basis for subsequent optimization processing.

[0100] Through the above steps, the fluctuations in electricity demand in the Fangcang Hospital in the next few days can be accurately predicted, and the reliability of the prediction can be evaluated, providing a scientific basis for optimizing the allocation of electricity resources.

[0101] Optionally, based on the spatiotemporal graph structure, a spatiotemporal graph neural network structure search algorithm is used to automatically search for the best network structure, maximize the ability to predict future power demand fluctuations, and generate spatiotemporal fluctuation prediction results, including:

[0102] Based on the space-time graph structure, the basic architecture of the space-time graph neural network is initialized, the initial network layer configuration is performed, and an initialized space-time graph neural network is generated; based on the initialized space-time graph neural network, the network structure search space is defined, the number of network layers and the activation function type are set, and the search space configuration is generated; based on the search space configuration, the space-time graph neural network structure search algorithm is used to automatically search for the optimal network structure to maximize the ability to predict future electricity demand fluctuations, and the network structure is gradually optimized through repeated training and verification to generate an optimized space-time graph neural network; based on the optimized space-time graph neural network, the space-time graph structure is deep learned to extract future electricity demand change trends and generate space-time fluctuation prediction results.

[0103] In an embodiment of the present application, based on the space-time graph structure, the basic architecture of the space-time graph neural network is initialized and processed, the initial network layer configuration is performed, and an initialized space-time graph neural network is generated; the network structure search space is defined, the number of network layers and the activation function type are set, and the search space configuration is generated; at the same time, the space-time graph neural network structure search algorithm is used to automatically search for the optimal network structure, and through multiple training and verification, the network structure is gradually optimized to generate an optimized space-time graph neural network; finally, the optimized space-time graph neural network is used to perform deep learning processing on the space-time graph structure, extract the changing trend of future electricity demand, and generate space-time fluctuation prediction results.

[0104] Assume that in a quarantine center, it is necessary to estimate the change in electricity demand in the next few days in order to allocate electricity resources more effectively; based on the spatiotemporal graph structure, initialize the basic architecture of the spatiotemporal graph neural network; set the basic configuration of the network's input layer, hidden layer, and output layer to generate an initialized spatiotemporal graph neural network; the initialized network structure includes multiple convolutional layers and fully connected layers, which are used to process time series data and spatial distribution information in the spatiotemporal graph structure; define the network structure search space, and set the range of the number of network layers and activation function types; the number of network layers can be selected between 3 and 10 layers, and the activation function can be selected from ReLU, Sigmoid, or Tanh, etc.; the generated search space configuration provides a clear range and options for subsequent network structure searches; use the spatiotemporal graph neural network structure search algorithm to automatically search for the best network structure in the defined search space; through multiple iterations and training, gradually optimize the network structure to improve the ability to predict future electricity demand fluctuations; through repeated training and verification, evaluate the performance of different network structures, and select the network structure with the strongest prediction ability as the final optimization result; use the optimized spatiotemporal graph neural network to perform deep learning processing on the spatiotemporal graph structure. The network extracts the changing trend of future electricity demand by learning the spatiotemporal relationship in historical data; the generated spatiotemporal fluctuation prediction results provide hourly electricity demand forecast values ​​in the next few days, so as to prepare electricity resource allocation and scheduling in advance.

[0105] Through the above steps, the changing trend of electricity demand in the quarantine center in the next few days can be accurately estimated, and reliable forecast data can be produced, laying the foundation for the scientific optimization of electricity resource allocation.

[0106] Optionally, based on the search space configuration, a spatiotemporal graph neural network structure search algorithm is used to automatically search for the best network structure, maximize the ability to predict future power demand fluctuations, and gradually optimize the network structure through repeated training verification to generate an optimized spatiotemporal graph neural network, including:

[0107] Initializing a network structure based on the search space configuration;

[0108] Using forward propagation to input the spatiotemporal graph structure, obtaining the predicted label of each sample, and recording the network weights and bias items to generate a fitness score of the network structure;

[0109] The fitness score of the network structure is calculated through the following steps:

[0110]

[0111] Among them, A(W,X) is the fitness score of the network structure; W is the network weight matrix; X is the input spatiotemporal graph structure; y i is the true label of the i-th sample; is the predicted label of the i-th sample; w j is the weight of the jth layer; K is the number of bias terms; b k is the kth bias term; λ is the regularization parameter; i is the index of the training sample, from 1 to N; N is the number of training samples; j is the index of the network layer, from 1 to L; L is the number of network layers; k is the index of the bias term, from 1 to K; K is the number of bias terms;

[0112] Based on the fitness score of the network structure, a nonlinear transformation optimization mechanism is introduced, and an optimization objective function is constructed by combining the network prediction error and the weight size to generate an optimization objective function value;

[0113] The optimization objective function value is calculated using the following formula:

[0114]

[0115] Among them, O(W,X,t) is the optimization objective function value; W is the network weight matrix; X is the input spatiotemporal graph structure; A(W,X) is the fitness score of the network structure; y i is the true label of the i-th sample; is the predicted label of the i-th sample; w j is the weight of the jth layer; K is the number of bias terms; b kis the kth bias term; α is the chaotic search strength coefficient; β is the error sensitivity coefficient; y is the weight regularization strength coefficient; t is the current iteration number; i is the index of the training sample, from 1 to N; N is the number of training samples; j is the index of the network layer, from 1 to L; L is the number of network layers; k is the index of the bias term, from 1 to K; K is the number of bias terms;

[0116] Based on the optimization objective function value, through multiple rounds of iterative optimization, the network weight matrix and bias terms are gradually adjusted, and the network structure is updated to minimize the optimization objective function value and generate an optimized spatiotemporal graph neural network.

[0117] This method aims to ensure that the network structure achieves the best balance between predictive ability and generalization ability by comprehensively considering network prediction error, weight size and chaotic search mechanism through fitness scoring and optimizing objective function.

[0118] In the fitness score A(W,X) of the network structure, the network prediction error term Calculate the mean square error between the network prediction value and the true value to evaluate the prediction accuracy of the model; weight regularization term L1 regularization is used to penalize excessive weights and biases to prevent overfitting of the model and improve generalization ability.

[0119] Where W is the network weight matrix, which is obtained by initializing the network structure; X is the input spatiotemporal graph structure, which is obtained from the comprehensive power demand dataset; y i is the true label of the i-th sample, obtained from the comprehensive power demand dataset; is the predicted label of the i-th sample, obtained by forward propagation calculation; w j is the weight of F, obtained by initializing the network structure; b k is the kth bias term, obtained by initializing the network structure; λ: regularization parameter, set according to experience or cross-validation, for example, λ = 0.01; N is the number of training samples, obtained from the comprehensive power demand data set; L is the number of network layers, determined according to the network structure configuration; K is the number of bias terms, determined according to the network structure configuration;

[0120] In the optimization objective function O(W,X,t), the fitness score term A(W,X): evaluates the prediction error and weight regularization of the network to ensure the accuracy and generalization ability of the model; the chaos search intensity term Through the exponential decay function, the chaotic search intensity is introduced to dynamically adjust the search strategy to avoid falling into the local optimum; the weight regularization intensity term Through the sine function, the weight regularization strength is introduced to further optimize the weight distribution and improve the stability and generalization ability of the model;

[0121] Among them, A(W,X) is the fitness score, which is obtained by calculating the fitness score of the network structure; α is the chaotic search intensity coefficient, which is set according to experience or cross-validation, for example, α = 0.5; β: error sensitivity coefficient, which is set according to experience or cross-validation, for example, β = 0.1; γ is the weight regularization intensity coefficient, which is set according to experience or cross-validation, for example, γ = 0.2; t is the current number of iterations, which increases from 0; N: the number of training samples, which is obtained from the comprehensive power demand data set; L is the number of network layers, which is determined according to the network structure configuration; K is the number of bias items, which is determined according to the network structure configuration;

[0122] Assume that in a quarantined biological laboratory, technicians need to optimize the prediction model of future power demand fluctuations through the spatiotemporal graph neural network structure search algorithm; based on the comprehensive power demand data set, initialize the network structure, set the number of network layers L = 5, the number of bias items K = 10, and the regularization parameter λ = 0.01; Assume N = 1000, y i and Obtained from the comprehensive electricity demand dataset, w j and b k Obtained by initializing the network structure, λ = 0.01;

[0123] Fitness score:

[0124]

[0125] Optimize the objective function value:

[0126]

[0127] After 100 rounds of iterations, the final optimization objective function value O(W,X,t) is 0.035, and the fitness score A(W,X) is reduced to 0.015; assuming that the set fitness score threshold is 0.02 and the optimization objective function value threshold is 0.04, since the final fitness score of 0.015 is less than the set threshold of 0.02, the optimization objective function value of 0.035 is less than 0.04, which shows that through multiple rounds of iterative optimization, the network structure has been significantly improved and the expected optimization effect has been achieved; the network prediction error and weight regularization in the optimization process are effectively controlled, so that the model can maintain high prediction accuracy while avoiding overfitting, thereby improving the model's prediction ability for unknown data. Therefore, it can be concluded that the optimized spatiotemporal graph neural network shows excellent performance in processing future electricity demand fluctuation prediction tasks; through the above steps, technicians have successfully optimized the spatiotemporal graph neural network, significantly improving the accuracy and generalization ability of the model in predicting future electricity demand fluctuations.

[0128] Optionally, the method of performing multiple random samplings based on the spatiotemporal fluctuation prediction results using Monte Carlo simulation technology to evaluate the uncertainty range of the prediction results and generate an uncertainty assessment includes:

[0129] Based on the spatiotemporal fluctuation prediction results, the probability distribution of the prediction model output variables is identified to generate a prediction value probability distribution model; based on the prediction value probability distribution model, the number of Monte Carlo simulation iterations is set, and random sampling is performed to ensure the randomness and representativeness of the input parameters and generate random samples; based on the random samples, the prediction model is iterated multiple times, all iteration results are collected, and a prediction result set is generated; based on the prediction result set, key statistical features are identified, the uncertainty range of the prediction result set is analyzed, and an uncertainty assessment is generated.

[0130] In an embodiment of the present application, based on the spatiotemporal fluctuation prediction results, the probability distribution of the output variables of the prediction model is identified, and a predicted value probability distribution model is generated; the number of iterations of the Monte Carlo simulation is set, and random sampling is performed to ensure the randomness and representativeness of the input parameters, and a random sample is generated; then, based on the generated random sample, the prediction model is iterated multiple times, all iteration results are collected, and a set of prediction results is generated; finally, through statistical analysis of the set of prediction results, key features are identified, the uncertainty range of the prediction results is evaluated, and an uncertainty assessment report is generated.

[0131] Suppose that in a modular hospital clinic, technicians need to evaluate the uncertainty of the electricity demand forecast for the next few days to ensure the rational allocation of electricity resources; based on the spatiotemporal fluctuation forecast results, technicians identify the probability distribution of the output variables of the forecast model; for example, assuming that the electricity demand value output by the forecast model obeys the normal distribution, the mean and standard deviation can be fitted through historical data to generate a forecast value probability distribution model; the generated forecast value probability distribution model provides a basis for subsequent random sampling; the number of iterations of the Monte Carlo simulation is set, for example 1000 times; in each iteration, a random sample is generated according to the forecast value probability distribution model; the generation of random samples ensures the randomness and representativeness of the input parameters and simulates different forecast scenarios; based on the set number of iterations, random sampling is performed. In each iteration, a sample is randomly drawn from the predicted value probability distribution model to generate a random sample; the generated random sample is used for subsequent iterative processing of the prediction model; based on the generated random sample, the prediction model is iterated multiple times; in each iteration, a random sample is used as input to run the prediction model and generate a prediction result; all iterative results are collected to generate a prediction result set; the prediction result set includes 1,000 prediction values, each corresponding to one iteration; statistical analysis is performed on the prediction result set to identify key features, such as mean, standard deviation, confidence interval, etc.; the uncertainty range of the prediction results is evaluated and an uncertainty assessment report is generated; the report includes the predicted values ​​and confidence intervals of hourly electricity demand in the next few days to help technicians understand the reliability of the prediction.

[0132] Through the above steps, technicians can comprehensively evaluate the uncertainty of the electricity demand forecast of the modular laboratory in the next few days and ensure the rational allocation and scheduling of power resources.

[0133] 104. Based on the prediction and evaluation results, an adaptive chaos search algorithm is used in combination with dynamic programming technology to optimize the power consumption strategy inside the shelter, balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy plan;

[0134] The prediction and evaluation results include the predicted value of future electricity demand and its uncertainty assessment, which are used to guide the optimization process; the adaptive chaos search algorithm is an optimization algorithm that simulates the dynamic behavior of the chaotic system and dynamically adjusts the search strategy to avoid falling into the local optimal solution; the dynamic programming technology is an optimization method that decomposes the problem into sub-problems and gradually solves them to ensure the global optimal solution; the optimized electricity consumption strategy is the electricity consumption strategy generated by the optimization process, which balances the energy efficiency ratio and comfort and ensures the efficient satisfaction of electricity demand.

[0135] First, based on the prediction and evaluation results, an adaptive chaos search algorithm is used to explore the optimization space of the power consumption strategy inside the cabin and generate a preliminary set of power consumption strategies. Secondly, dynamic programming technology is combined to ensure that the energy efficiency ratio and comfort level are balanced while meeting the power demand, and to generate an optimized power consumption strategy plan. In this process, multiple iterations and evaluations are carried out to gradually optimize the power consumption strategy to ensure the efficiency and feasibility of the final plan.

[0136] Optionally, the method in step 104 uses an adaptive chaos search algorithm based on the prediction and evaluation results, combined with dynamic programming technology, to optimize the power consumption strategy inside the shelter, balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy solution, including:

[0137] Based on the prediction and evaluation results, the time distribution characteristics are analyzed, the impact of uncertainty is evaluated, the key influencing factors are determined, and the basic input data is generated; based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the power consumption strategy inside the cabin, and multi-stage and multi-objective optimization processing is performed to generate a preliminary power consumption strategy set; based on the preliminary power consumption strategy set, combined with dynamic programming technology, ensure that the premise of meeting the power demand is met, balance the maximization of energy efficiency ratio and the optimization of comfort, and generate an optimized power consumption strategy set; based on the optimized power consumption strategy set, screen through a multi-criteria decision analysis method, comprehensively consider the flexibility of power consumption, and generate an optimized power consumption strategy plan.

[0138] In the embodiment of the present application, based on the prediction and evaluation results, the time distribution characteristics are analyzed, the impact of uncertainty is evaluated, the key influencing factors are determined, and the basic input data is generated; based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the power consumption strategy inside the cabin, and multi-stage and multi-objective optimization processing is performed to generate a preliminary power consumption strategy set; then, based on the preliminary power consumption strategy set, combined with dynamic programming technology, it is ensured that the energy efficiency ratio is maximized and the comfort is optimized while meeting the electricity demand, and an optimized power consumption strategy set is generated; finally, through a multi-criteria decision analysis method, the power consumption flexibility is comprehensively considered, the optimal plan is screened out from the optimized power consumption strategy set, and the final optimized power consumption strategy plan is generated.

[0139] Suppose that in a square cabin hospital, technicians need to optimize the internal electricity consumption strategy to balance energy efficiency and comfort; based on the prediction and evaluation results, technicians analyze the time distribution characteristics of future electricity demand, evaluate the impact of uncertainty, and determine key influencing factors, such as temperature, humidity and other environmental parameters; based on these analysis results, generate basic input data, including future electricity demand forecast values, uncertainty ranges, key influencing factors, etc., for the input of the optimization algorithm; based on the generated basic input data, technicians use an adaptive chaos search algorithm to explore the optimization space of the electricity consumption strategy inside the square cabin; the algorithm simulates the dynamic behavior of the chaotic system and dynamically adjusts the search strategy to avoid falling into the local optimal solution; through multi-stage and multi-objective optimization processing, a preliminary set of electricity consumption strategies is generated. Each preliminary electricity consumption strategy takes into account the two objectives of energy efficiency and comfort. Based on the preliminary electricity consumption strategy set, combined with dynamic programming technology, it ensures that the maximum energy efficiency and the optimal comfort are balanced under the premise of meeting the electricity demand. Dynamic programming technology decomposes the problem into sub-problems and solves them step by step to ensure that the generated optimized electricity consumption strategy set is optimal in the global scope. Based on the optimized electricity consumption strategy set, through the multi-criteria decision analysis method, the electricity consumption flexibility is comprehensively considered to screen out the best solution from multiple solutions. The multi-criteria decision analysis method takes into account multiple evaluation indicators, such as energy efficiency, comfort, implementation difficulty, etc., to ensure that the final generated optimized electricity consumption strategy solution is both efficient and feasible.

[0140] Through the above steps, technicians can generate an optimized power consumption strategy plan, balance energy efficiency and comfort, and ensure that the internal power demand of the square cabin hospital is efficiently met.

[0141] Optionally, based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the electricity consumption strategy inside the shelter, perform multi-stage multi-objective optimization processing, and generate a preliminary electricity consumption strategy set, including:

[0142] Based on the basic input data, an adaptive model of the optimization problem is constructed, the optimization objectives and constraints are defined, and an adaptive optimization model is generated; based on the adaptive optimization model, a preset number of individual samples are randomly selected in combination with the operating parameter ranges of the electrical equipment inside the cabin to generate an initial population individual set; based on the initial population individual set, an adaptive chaos search algorithm is used to dynamically adjust the search strategy, simulate the dynamic behavior of the chaotic system, enrich the optimal solution space, and generate a new population individual set; based on the new population individual set, through multi-stage iterative processing, the fitness of the candidate strategies is traversed and evaluated, and the search range is gradually narrowed in combination with the Pareto optimization technology to generate a preliminary power consumption strategy set.

[0143] In the embodiment of the present application, based on the prediction and evaluation results, the time distribution characteristics are analyzed, the impact of uncertainty is evaluated, the key influencing factors are determined, and the basic input data is generated; based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the power consumption strategy inside the cabin, and multi-stage and multi-objective optimization processing is performed to generate a preliminary power consumption strategy set; based on the preliminary power consumption strategy set, combined with dynamic programming technology, it is ensured that the energy efficiency ratio is maximized and the comfort is optimized while meeting the electricity demand, and an optimized power consumption strategy set is generated; finally, through a multi-criteria decision analysis method, the power consumption flexibility is comprehensively considered, the optimal plan is screened out from the optimized power consumption strategy set, and the final optimized power consumption strategy plan is generated.

[0144] Assume that in a square cabin hospital, technicians need to optimize the internal power consumption strategy to balance energy efficiency and comfort. Based on the prediction and evaluation results, technicians analyze the time distribution characteristics of future power demand, evaluate the impact of uncertainty, and determine key influencing factors, such as temperature, humidity and other environmental parameters. Based on these analysis results, basic input data are generated, including future power demand forecast values, uncertainty ranges, key influencing factors, etc., which are used for the input of the optimization algorithm. Based on the generated basic input data, technicians use an adaptive chaos search algorithm to explore the optimization space of the power consumption strategy inside the square cabin. The algorithm simulates the dynamic behavior of the chaotic system and dynamically adjusts the search strategy to avoid falling into the local optimal solution. Through multi-stage and multi-objective optimization processing, a preliminary set of power consumption strategies is generated. Each preliminary electricity consumption strategy takes into account the two objectives of energy efficiency and comfort. Based on the preliminary electricity consumption strategy set, technical personnel combined dynamic programming technology to ensure that the maximum energy efficiency and the optimal comfort are balanced under the premise of meeting electricity demand. Dynamic programming technology breaks down the problem into sub-problems and gradually solves them to ensure that the generated set of optimized electricity consumption strategies is optimal in the global scope. Based on the optimized set of electricity consumption strategies, technical personnel use multi-criteria decision analysis methods to comprehensively consider electricity flexibility and select the optimal solution from multiple solutions. The multi-criteria decision analysis method takes into account multiple evaluation indicators, such as energy efficiency, comfort, implementation difficulty, etc., to ensure that the final generated optimized electricity consumption strategy solution is both efficient and feasible.

[0145] Through the above steps, technicians can generate an optimized power consumption strategy plan, balance energy efficiency and comfort, and ensure that the internal power demand of the square cabin hospital is efficiently met.

[0146] Optionally, based on the initial population individual set, an adaptive chaos search algorithm is used to dynamically adjust the search strategy, simulate the dynamic behavior of the chaotic system, enrich the optimal solution space, and generate a new population individual set, including:

[0147] Based on the initial population individual set, evaluating the fitness value of each individual;

[0148] By analyzing the distribution of individuals in the population, determining the center point of the current iteration, obtaining the average fitness value of the current population, and setting parameters related to population diversity, a search strategy is generated;

[0149] The search strategy is calculated using the following formula:

[0150]

[0151] Where S(t+1) is the search strategy of the t+1th iteration; S(t) is the search strategy of the tth iteration; C(t) is the center point of the tth iteration; η is the learning rate; σ is the chaotic search intensity coefficient; t is the current iteration number; T is the total iteration number; ω is the nonlinear weight coefficient; k is the nonlinear adjustment coefficient; F i (t) is the fitness value of the i-th individual at the t-th iteration; F avg (t) is the average fitness value of the tth iteration;

[0152] Based on the search strategy, combined with the current individual position, nonlinear adjustment items and search strategy influencing items are introduced to enhance the dynamic adjustment capability, ensure the robustness and effectiveness of the optimization process, and generate the population fitness value;

[0153] The population fitness value is calculated using the following formula:

[0154]

[0155] Among them, F i (t+1) is the population fitness value of the i-th individual at the t+1th iteration; F i (t) is the population fitness value of the i-th individual at the t-th iteration; F best (t) is the optimal population fitness value of the tth iteration; α is the step size factor; β is the nonlinear adjustment coefficient; X ij (t) is the position of the i-th individual in the j-th dimension; X best,j (t) is the position of the optimal solution in the jth dimension; t is the current iteration number; S(t+1) is the search strategy for the t+1th iteration; θ is the influence coefficient of the search strategy; j is the index of the individual dimension, from 1 to D; D is the number of individual dimensions; S(t+1) is the search strategy for the t+1th iteration;

[0156] Based on the population fitness value, combined with the Pareto optimization technology, the search range is gradually narrowed, and the population diversity is enriched through crossover and mutation operations to converge the optimization process and generate a new set of population individuals.

[0157] This method aims to dynamically adjust the search strategy through an adaptive chaotic search algorithm, combine nonlinear adjustment terms and search strategy influencing terms, enhance the robustness and effectiveness of the optimization process, ensure population diversity, and ultimately generate a new set of population individuals.

[0158] In the search strategy, the current population center point term η·(C(t)-S(t)): the difference between the current population center point C(t) and the current search strategy S(t), multiplied by the learning rate η, guides the search strategy to move closer to the population center point to ensure the stability of the search process; the chaotic search intensity term The chaotic search mechanism is introduced through the chaotic search intensity coefficient σ and the sine function to increase the randomness and diversity of the search process; the nonlinear weight term Through the nonlinear adjustment of the nonlinear weight coefficient ω and the fitness value, the influence of the search strategy is dynamically adjusted to enhance the robustness of the search process;

[0159] Among them, S(t) is the search strategy of the tth iteration, and the initial value can be set randomly or according to experience; C(t) is the center point of the tth iteration, which is obtained by calculating the average position of the individuals in the current population; η is the learning rate, which is set according to experience or cross-validation; σ is the chaotic search intensity coefficient, which is set according to experience or cross-validation; t is the current iteration number, which increases from 0; T is the total number of iterations, which is set according to optimization requirements; ω is the nonlinear weight coefficient, which is set according to experience or cross-validation; k is the nonlinear adjustment coefficient, which is set according to experience or cross-validation; F i (t) is the fitness value of the i-th individual at the t-th iteration, obtained by evaluating the fitness of the individual; F avg (t) is the average fitness value of the tth iteration, obtained by calculating the average fitness of all individuals in the current population;

[0160] In the population fitness value, the optimal fitness value term α·(F best (t)-F i (t)): through the optimal fitness value F best (t) and the current individual fitness value F i The difference of (t) is multiplied by the step size factor α to guide the individual to the optimal solution and ensure the effectiveness of the optimization process; the nonlinear adjustment term Through the nonlinear adjustment coefficient β and the difference between the individual position and the optimal solution position, a nonlinear adjustment term is introduced to enhance the dynamic adjustment capability; Search strategy influence term θ·S(t+1): Through the search strategy influence coefficient θ and the updated search strategy S(t+1), the influence of the search strategy on the individual fitness value is ensured, and the robustness of the optimization process is enhanced;

[0161] Among them, F i(t) is the population fitness value of the i-th individual at the t-th iteration, obtained by evaluating the fitness of the individual; F best (t) is the optimal population fitness value of the tth iteration, obtained by comparing the fitness of all individuals in the current population; α is the step factor, set according to experience or cross-validation; β is the nonlinear adjustment coefficient, set according to experience or cross-validation; X ij (t) is the position of the ith individual in the th dimension, obtained from the initial population individual set; X best,j (t) is the position of the optimal solution in the jth dimension, obtained by comparing the positions of all individuals in the current population; θ is the search strategy influence coefficient, set based on experience or cross-validation; S(t+1) is the search strategy for the t+1th iteration, obtained by calculation; D is the number of individual dimensions, set according to the dimension of the problem;

[0162] Suppose that in a shelter chemical laboratory, technicians need to optimize the internal power consumption strategy through an adaptive chaos search algorithm;

[0163] Assume that the parameters S(t) = 0.5, C(t) = 0.6, η = 0.1, σ = 0.5, t = 10, T = 100, ω = 0.2, k = 0.1, F i (t)=0.8,F avg (t) = 0.7;

[0164] Search strategy

[0165] Assume the parameter is F i (t)=0.8,F best (t)=0.9,α=0.1,β=0.2,X ij (t)=[0.5,0.6,0.7,0.8,0.9,1.0,1.1,1.2,1.3,1.4],

[0166] X best,j (t)=[0.6,0.7,0.8,0.9,1.0,1.1,1.2,1.3,1.4,1.5],θ=0.3,S(t+1)=

[0167] 0.607,D=10;

[0168] Population fitness value

[0169] Assuming that the set fitness value threshold is 1.1, since the final population fitness value of 1.1921 is greater than the set threshold, this shows that through multiple rounds of iterative optimization, the fitness values ​​of the individuals in the population have been significantly improved, and the optimization process has achieved the expected results, indicating that the optimization algorithm has effectively improved the fitness and optimization performance of the population; through the above steps, the technicians have successfully optimized the internal electricity consumption strategy of the square cabin chemical laboratory, significantly improving the robustness and effectiveness of the model.

[0170] 105. Based on the optimized electricity consumption strategy plan, monitor the deviation between actual electricity consumption and expected effects, record environmental parameters during the implementation of the plan, conduct feedback and iterative optimization, and generate an electricity demand response model.

[0171] The optimized electricity consumption strategy is a power consumption strategy generated through optimization processing, which balances energy efficiency and comfort to ensure efficient satisfaction of electricity demand; the deviation between actual electricity consumption and expected effect is the difference between actual electricity consumption data and expected effect of optimization strategy, which is used to evaluate the effectiveness of optimization strategy; environmental parameters include temperature, humidity and other environmental parameters, recording these parameters helps to analyze the impact of environmental changes on electricity demand; the electricity demand response model is a model generated through feedback and iterative optimization, which can adapt to different environmental conditions and achieve balance between electricity supply and demand.

[0172] First, based on the optimized power consumption strategy plan, these strategies are implemented inside the cabin, the actual power consumption is monitored, and the deviation between the actual power consumption and the expected effect is recorded; secondly, the environmental parameters during the implementation of the plan, such as temperature, humidity, etc., are recorded at the same time; thirdly, feedback and iterative optimization are carried out through data analysis, and the optimization strategy is gradually adjusted to generate the final power demand response model; in this process, the strategy is continuously optimized and adjusted to ensure that the model can adapt to different environmental conditions and achieve a dynamic balance between power supply and demand.

[0173] Optionally, the optimized power consumption strategy scheme in step 105 monitors the deviation between actual power consumption and expected effect, records environmental parameters during implementation of the scheme, performs feedback and iterative optimization, and generates a power demand response model, including:

[0174] Based on the optimized power consumption strategy, the power consumption strategy inside the shelter is adjusted, the actual power consumption is monitored, the environmental parameters during the implementation are recorded, and the actual operating parameter records are generated; based on the actual operating parameter records, the actual power consumption is compared with the expected effect, the multi-dimensional deviation results are calculated, and a deviation analysis report is generated; based on the deviation analysis report, feedback and iterative optimization are performed, the power consumption strategy inside the shelter is evaluated and optimized multiple times, the actual operating parameters are gradually adjusted, and the optimization iteration results are generated; based on the optimization iteration results, the effective strategies in the optimization iteration process are integrated to improve the accuracy of the prediction model and the effectiveness of the response measures, and generate a power demand response model.

[0175] In the embodiment of the present application, based on the optimized power consumption strategy plan, the power consumption strategy inside the cabin is adjusted, the actual power consumption is monitored, the environmental parameters during the implementation period are recorded, and the actual operating parameter records are generated; based on the actual operating parameter records, the actual power consumption is compared with the expected effect, the multi-dimensional deviation results are calculated, and a deviation analysis report is generated; then based on the deviation analysis report, feedback and iterative optimization are performed, the power consumption strategy inside the cabin is evaluated and optimized multiple times, the actual operating parameters are gradually adjusted, and the optimization iteration results are generated; finally, based on the optimization iteration results, the effective strategies in the optimization iteration process are integrated to improve the accuracy of the prediction model and the effectiveness of the response measures, and generate a power demand response model.

[0176] Assume that in a meteorological monitoring cabin, the technician needs to monitor the deviation between actual power consumption and expected effects through the optimized power strategy plan, record the environmental parameters during the implementation of the plan, conduct feedback and iterative optimization, and generate a power demand response model; based on the optimized power strategy plan, the technician adjusts the power strategy inside the cabin, monitors the actual power consumption in real time, and records the environmental parameters during the implementation period, such as temperature and humidity, to generate actual operating parameter records; based on the actual operating parameter records, the technician compares the actual power consumption with the expected effects, calculates multi-dimensional deviation results, including power consumption, energy efficiency ratio and comfort, and generates a deviation analysis report; based on the deviation analysis report, the technician conducts feedback and iterative optimization, evaluates and optimizes the power strategy inside the cabin multiple times, and gradually adjusts the actual operating parameters, such as adjusting the air conditioning temperature setting, optimizing the lighting system switching time, etc., to generate optimization iteration results; based on the optimization iteration results, the technician integrates the effective strategies in the optimization iteration process, improves the accuracy of the prediction model and the effectiveness of the response measures, and generates a power demand response model;

[0177] Through the above steps, technicians successfully generated an electricity demand response model, which can not only accurately predict future electricity demand, but also make dynamic adjustments based on actual operating conditions to ensure that the electricity demand of the square cabin isolation point is efficiently met while balancing energy efficiency and comfort.

[0178] Figure 2 A schematic diagram of the structure of a shelter power demand response system based on an artificial intelligence algorithm is provided for an embodiment of the present application, such as Figure 2 As shown, the device comprises:

[0179] The collection module 21 is used to collect the real-time power consumption data of each electrical device in the shelter and generate an original data set in combination with the environmental parameter data;

[0180] A processing module 22 is used to pre-process the original data set based on an artificial intelligence algorithm, perform feature selection and dimensionality reduction processing, and generate a comprehensive power demand data set;

[0181] A prediction module 23 is used to predict future power demand fluctuations based on the comprehensive power demand data set by using a spatiotemporal graph neural network structure search algorithm, and to use a Monte Carlo simulation technique to evaluate the uncertainty of the prediction results and generate a prediction evaluation result;

[0182] The optimization module 24 is used to optimize the power consumption strategy inside the cabin based on the prediction and evaluation results, use an adaptive chaos search algorithm, and combine dynamic programming technology to balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy plan;

[0183] The monitoring module 25 is used to monitor the deviation between actual power consumption and expected effect based on the optimized power consumption strategy plan, record environmental parameters during the implementation of the plan, perform feedback and iterative optimization, and generate a power demand response model.

[0184] Figure 2 The described shelter power demand response system based on artificial intelligence algorithm can execute Figure 1 The implementation principle and technical effect of the shelter power demand response method based on artificial intelligence algorithm described in the embodiment shown will not be repeated. The specific way in which each module and unit performs operations in the shelter power demand response system based on artificial intelligence algorithm in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0185] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for responding to electricity demand in a shelter based on an artificial intelligence algorithm, characterized in that: include: Collect the real-time power consumption data of each electrical device in the shelter and combine it with the environmental parameter data to generate the original data set; Based on artificial intelligence algorithms, the original data set is preprocessed to perform feature selection and dimensionality reduction processing to generate a comprehensive power demand data set; Based on the comprehensive power demand data set, a spatiotemporal graph neural network structure search algorithm is used to predict future power demand fluctuations, and a Monte Carlo simulation technique is used to evaluate the uncertainty of the prediction results to generate a prediction evaluation result; Based on the prediction and evaluation results, an adaptive chaos search algorithm is used in combination with dynamic programming technology to optimize the power consumption strategy inside the shelter, balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy plan; Based on the optimized electricity consumption strategy plan, the deviation between actual electricity consumption and expected effects is monitored, environmental parameters are recorded during the implementation of the plan, feedback and iterative optimization are performed, and an electricity demand response model is generated.

2. The method according to claim 1, characterized in that: Based on the comprehensive power demand data set, the spatiotemporal graph neural network structure search algorithm is used to predict future power demand fluctuations, and the Monte Carlo simulation technology is used to evaluate the uncertainty of the prediction results to generate prediction evaluation results, including: Based on the comprehensive power demand data set, time series data and spatial distribution information are obtained, and the spatiotemporal relationship of power consumption inside the shelter is combined to generate a spatiotemporal graph structure; Based on the spatiotemporal graph structure, a spatiotemporal graph neural network structure search algorithm is used to automatically search for the optimal network structure, maximize the ability to predict future power demand fluctuations, and generate spatiotemporal fluctuation prediction results; Based on the spatiotemporal fluctuation prediction results, the Monte Carlo simulation technique is used to perform multiple random samplings to evaluate the uncertainty range of the prediction results and generate uncertainty assessments; Based on the uncertainty assessment, historical data and current environmental parameters are combined to identify potential influencing factors and generate predictive assessment results.

3. The method according to claim 2, characterized in that Based on the spatiotemporal graph structure, the spatiotemporal graph neural network structure search algorithm is used to automatically search for the optimal network structure, maximize the future power demand fluctuation prediction capability, and generate spatiotemporal fluctuation prediction results, including: Based on the space-time graph structure, the basic architecture of the space-time graph neural network is initialized, the initial network layer configuration is performed, and an initialized space-time graph neural network is generated; Based on the initialized spatiotemporal graph neural network, define the network structure search space, set the number of network layers and activation function type, and generate the search space configuration; Based on the search space configuration, the space-time graph neural network structure search algorithm is used to automatically search for the best network structure, maximize the ability to predict future power demand fluctuations, and gradually optimize the network structure through repeated training and verification to generate an optimized space-time graph neural network; Based on the optimized spatiotemporal graph neural network, deep learning processing is performed on the spatiotemporal graph structure to extract the future trend of electricity demand changes and generate spatiotemporal fluctuation prediction results.

4. The method according to claim 3, characterized in that Based on the search space configuration, the spatiotemporal graph neural network structure search algorithm is used to automatically search for the optimal network structure, maximize the ability to predict future power demand fluctuations, and gradually optimize the network structure through repeated training and verification to generate an optimized spatiotemporal graph neural network, including: Initializing a network structure based on the search space configuration; Using forward propagation to input the spatiotemporal graph structure, obtaining the predicted label of each sample, and recording the network weights and bias items to generate a fitness score of the network structure; The fitness score of the network structure is calculated through the following steps: Among them, A(W, X) is the fitness score of the network structure; W is the network weight matrix; X is the input spatiotemporal graph structure; y i is the true label of the i-th sample; is the predicted label of the i-th sample; w j is the weight of the jth layer; K is the number of bias terms; b k is the kth bias term; λ is the regularization parameter; i is the index of the training sample, from 1 to N; N is the number of training samples; j is the index of the network layer, from 1 to L; L is the number of network layers; k is the index of the bias term, from 1 to K; K is the number of bias terms; Based on the fitness score of the network structure, a nonlinear transformation optimization mechanism is introduced, and an optimization objective function is constructed by combining the network prediction error and the weight size to generate an optimization objective function value; The optimization objective function value is calculated using the following formula: Among them, O(W, X, t) is the optimization objective function value; W is the network weight matrix; X is the input spatiotemporal graph structure; A(W, X) is the fitness score of the network structure; y i is the true label of the i-th sample; is the predicted label of the i-th sample; w j is the weight of the jth layer; K is the number of bias terms; b k is the kth bias term; α is the chaotic search strength coefficient; β is the error sensitivity coefficient; γ is the weight regularization strength coefficient; t is the current iteration number; i is the index of the training sample, from 1 to N; N is the number of training samples; j is the index of the network layer, from 1 to L; L is the number of network layers; k is the index of the bias term, from 1 to K; K is the number of bias terms; Based on the optimization objective function value, through multiple rounds of iterative optimization, the network weight matrix and bias terms are gradually adjusted, and the network structure is updated to minimize the optimization objective function value and generate an optimized spatiotemporal graph neural network.

5. The method according to claim 2, characterized in that: Based on the spatiotemporal fluctuation prediction results, the Monte Carlo simulation technology is used to perform multiple random samplings to evaluate the uncertainty range of the prediction results and generate uncertainty assessments, including: Based on the spatiotemporal fluctuation prediction results, identifying probability distribution of prediction model output variables and generating a prediction value probability distribution model; Based on the predicted value probability distribution model, the number of Monte Carlo simulation iterations is set, random sampling is performed to ensure the randomness and representativeness of the input parameters, and random samples are generated; Based on the random samples, the prediction model is iterated multiple times, all iteration results are collected, and a prediction result set is generated; Based on the set of prediction results, statistical key features are identified, the uncertainty range of the set of prediction results is analyzed, and an uncertainty assessment is generated.

6. The method according to claim 1, characterized in that Based on the prediction and evaluation results, the adaptive chaos search algorithm is used in combination with dynamic programming technology to optimize the power consumption strategy inside the shelter, balance the energy efficiency ratio and comfort, and generate an optimized power consumption strategy solution, including: Based on the forecast evaluation results, analyze the time distribution characteristics, evaluate the impact of uncertainty, determine the key influencing factors, and generate basic input data; Based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the electricity consumption strategy inside the shelter, and a multi-stage multi-objective optimization process is performed to generate a preliminary electricity consumption strategy set; Based on the preliminary power consumption strategy set, combined with dynamic programming technology, the premise of ensuring that power demand is met, balancing energy efficiency maximization and comfort optimization, and generating an optimized power consumption strategy set; Based on the optimized electricity consumption strategy set, a multi-criteria decision analysis method is used for screening, electricity consumption flexibility is comprehensively considered, and an optimized electricity consumption strategy solution is generated.

7. The method according to claim 6, characterized in that Based on the basic input data, an adaptive chaos search algorithm is used to explore the optimization space of the electricity consumption strategy inside the shelter, perform multi-stage multi-objective optimization processing, and generate a preliminary electricity consumption strategy set, including: Based on the basic input data, an adaptive model of the optimization problem is constructed, the optimization objectives and constraints are defined, and an adaptive optimization model is generated; Based on the adaptive optimization model and in combination with the operating parameter range of each electrical equipment in the shelter, a preset number of individual samples are randomly selected to generate an initial population individual set; Based on the initial population individual set, an adaptive chaos search algorithm is used to dynamically adjust the search strategy, simulate the dynamic behavior of the chaotic system, enrich the optimal solution space, and generate a new population individual set; Based on the new population individual set, through multi-stage iterative processing, the fitness of candidate strategies is traversed and evaluated, and combined with the Pareto optimization technology, the search scope is gradually narrowed to generate a preliminary electricity consumption strategy set.

8. The method according to claim 7, characterized in that Based on the initial population individual set, an adaptive chaotic search algorithm is used to dynamically adjust the search strategy, simulate the dynamic behavior of the chaotic system, enrich the optimal solution space, and generate a new population individual set, including: Based on the initial population individual set, evaluating the fitness value of each individual; By analyzing the distribution of individuals in the population, determining the center point of the current iteration, obtaining the average fitness value of the current population, and setting parameters related to population diversity, a search strategy is generated; The search strategy is calculated using the following formula: Among them, S(t+1) is the search strategy of the t+1th iteration; s(t) is the search strategy of the tth iteration; C(t) is the center point of the tth iteration; ηη is the learning rate; σ is the chaotic search intensity coefficient; t is the current iteration number; T is the total iteration number; ω is the nonlinear weight coefficient; k is the nonlinear adjustment coefficient; F i (t) is the fitness value of the i-th individual at the t-th iteration; F avg (t) is the average fitness value of the tth iteration; Based on the search strategy, combined with the current individual position, nonlinear adjustment items and search strategy influencing items are introduced to enhance the dynamic adjustment capability, ensure the robustness and effectiveness of the optimization process, and generate the population fitness value; The population fitness value is calculated using the following formula: Among them, F i (t+1) is the population fitness value of the i-th individual at the t+1th iteration; F i (t) is the population fitness value of the i-th individual at the t-th iteration; F best (t) is the optimal population fitness value of the tth iteration; α is the step size factor; β is the nonlinear adjustment coefficient; X ij (t) is the position of the i-th individual in the j-th dimension; X best，j (t) is the position of the optimal solution in the jth dimension; t is the current iteration number; S(t+1) is the search strategy for the t+1th iteration; θ is the influence coefficient of the search strategy; j is the index of the individual dimension, from 1 to D; D is the number of individual dimensions; Based on the population fitness value, combined with the Pareto optimization technology, the search range is gradually narrowed, and the population diversity is enriched through crossover and mutation operations to converge the optimization process and generate a new set of population individuals.

9. The method according to claim 1, characterized in that: The optimized power consumption strategy scheme, monitoring the deviation between actual power consumption and expected effect, recording environmental parameters during the implementation of the scheme, performing feedback and iterative optimization, and generating a power demand response model include: Based on the optimized power consumption strategy, adjust the power consumption strategy inside the shelter, monitor the actual power consumption, record the environmental parameters during the implementation, and generate actual operation parameter records; Based on the actual operating parameter records, the actual power consumption is compared with the expected effect, multi-dimensional deviation results are calculated, and a deviation analysis report is generated; Based on the deviation analysis report, feedback and iterative optimization are performed, the electricity consumption strategy inside the shelter is evaluated and optimized multiple times, the actual operating parameters are gradually adjusted, and the optimization iterative results are generated; Based on the optimization iteration results, effective strategies in the optimization iteration process are integrated to improve the accuracy of the prediction model and the effectiveness of the response measures, and generate a power demand response model.

10. A shelter power demand response system based on artificial intelligence algorithm, characterized in that: include: The collection module is used to collect the real-time power consumption data of each electrical device in the shelter and generate the original data set by combining it with the environmental parameter data; A processing module, used to pre-process the original data set based on an artificial intelligence algorithm, perform feature selection and dimensionality reduction processing, and generate a comprehensive power demand data set; A prediction module, which is used to predict future power demand fluctuations based on the comprehensive power demand data set by using a spatiotemporal graph neural network structure search algorithm, and to use a Monte Carlo simulation technique to evaluate the uncertainty of the prediction results and generate a prediction evaluation result; An optimization module is used to optimize the power consumption strategy inside the shelter based on the prediction and evaluation results, using an adaptive chaos search algorithm combined with dynamic programming technology, balancing energy efficiency and comfort, and generating an optimized power consumption strategy plan; The monitoring module is used to monitor the deviation between actual power consumption and expected effect based on the optimized power consumption strategy plan, record environmental parameters during the implementation of the plan, perform feedback and iterative optimization, and generate a power demand response model.

Images