Urban energy mutual aid comprehensive management system

Abstract

The invention discloses an urban energy mutual aid comprehensive management system, and relates to the technical field of urban heat supply management. According to the urban energy mutual aid comprehensive management system, an optimal model is determined by combining historical heat supply consumption data, key features and optimized features obtained after the key features are optimized; the heat supply demand is predicted by using the optimal model according to real-time monitoring data and historical data, so that the defects of a single prediction model are avoided, the prediction result deviation is reduced, and the prediction accuracy of the heat supply demand is improved; therefore, the energy supply plan can be further matched with the actual demand, so that the heat supply shortage is effectively avoided at the peak of the heat supply demand, and the normal heat supply of a city is ensured; and in the low ebb period, heat energy waste can be prevented, and unnecessary cost increase is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of urban heating management, and in particular to an urban energy mutual assistance integrated management system. Background Art

[0002] In the current field of urban heating management, with the rapid economic development and accelerated urbanization, urban heating consumption shows a trend of rapid growth and increasingly complex structure; the traditional urban energy mutual assistance integrated management system has many shortcomings and it is difficult to meet the needs of efficient utilization and stable supply of modern urban heating; most existing systems currently lack accuracy in heat demand forecasting; they often fail to fully consider the various complex factors that affect heating demand, such as weather conditions, equipment operating status, and personnel activities, and only rely on simple historical data statistics or a single forecasting model, resulting in large deviations in forecast results; this makes it difficult to match heating supply plans with actual demand, and shortages may occur during peak heat supply periods, affecting the normal operation of the city; and energy waste may occur during low periods, increasing unnecessary costs. Summary of the invention

[0003] 1. Technical issues to be resolved In view of the deficiencies in the prior art, the present invention provides an integrated management system for urban energy mutual assistance, which solves the problem that the existing systems lack accuracy in predicting heating demand. They often fail to fully consider the various complex factors that affect heating demand, such as weather conditions, equipment operating status, and personnel activities, and only rely on simple historical data statistics or a single prediction model, resulting in large deviations in the prediction results. This makes it difficult to match the heating supply plan with actual demand, and shortages may occur during peak heating supply periods, affecting the normal operation of the city. During low periods, it may cause energy waste and increase unnecessary costs.

[0004] (II) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: an urban energy mutual aid integrated management system, comprising: Data collection and monitoring module, used to collect heat consumption data in urban commercial areas or industrial parks in real time, and transmit the heat consumption data to the central database; Intelligent scheduling and optimization module, which is used to predict changes in heating demand in advance and optimize heating distribution strategies based on historical data and real-time monitoring of heating consumption data; The energy storage system management module is used to control the energy storage system to store excess heat energy when the heating demand is low, and release it during peak hours to balance the load of the heating system; The heating mutual assistance control module is used to establish a mutual assistance mechanism between different energy forms. When the heating supply is insufficient, natural gas boilers are used for heating or combined heating technology is used for supplementation to achieve energy complementarity and optimal configuration; the heating mutual assistance control module can coordinate the supply of different energy forms to ensure that other energy forms are used for supplementation when the heating supply is insufficient.

[0005] Preferably, the data acquisition and monitoring module collects the heat consumption data in the urban commercial area or industrial park in real time, and transmits the heat consumption data to the central database. The process includes: according to the actual energy consumption of the urban commercial area or industrial park, selecting heating sensors, water flow sensors, gas flow sensors, and data collectors, and deploying them on energy consumption equipment accordingly. The energy consumption equipment includes heat meters, water meters, gas meters, and heating systems. The heating sensors are deployed on the heat meters and heating systems. The data collector collects the original data of heat consumption in real time. The original data includes heat, water, and gas consumption data; the original data is deduplicated and denoised to obtain heat consumption data, and transmitted to the central database.

[0006] Preferably, the process in which the intelligent scheduling and optimization module predicts changes in heating demand in advance based on historical data and real-time monitored heating consumption data and optimizes the heating distribution strategy includes: first, collecting historical heating consumption data and data on factors affecting heating consumption, and then extracting key features that affect heating demand based on historical data and real-time monitored heating consumption data using a feature selection algorithm; then determining the optimal model by combining historical heating consumption data, key features, and optimized features after the key features are optimized; using the optimal model, predicting heating demand based on real-time monitoring data and historical data to obtain prediction results; based on the prediction results, combining heating supply conditions and heating distribution strategies, using intelligent optimization algorithms to schedule and optimize heating in real time, with the goal of ensuring that the heating system is not in short supply during peak hours and avoiding waste during trough hours, and obtaining a heating distribution strategy, wherein the intelligent optimization algorithms include genetic algorithms and particle swarm algorithms.

[0007] Preferably, the process of extracting key features affecting heating demand using feature selection algorithm based on historical data and real-time monitored heating consumption data includes: obtaining heating consumption data from a central database, and collecting influencing factor data through real-time monitoring, the influencing factor data including weather conditions, equipment operating status, and personnel activities; cleaning the collected heating consumption data and influencing factor data to remove duplicate, missing or abnormal values; and preliminarily screening out preliminary features related to heating demand based on domain knowledge and experience, the preliminary features including time, weather conditions, equipment status, and personnel activities; and calculating the correlation between preliminary features and target variables, i.e., heating demand, based on statistical methods. The importance of features is evaluated by correlation, importance score and contribution index, and preliminary features are evaluated to obtain preliminary feature importance evaluation results; according to the preliminary feature importance evaluation results, features with absolute value of correlation coefficient greater than 0.6, importance score greater than 70 points and contribution index greater than 30% are selected as key features, recorded as key features, and key features will be used in subsequent heating demand forecasting and optimization scheduling models. K-fold cross-validation is used to cross-validate the extracted key features to obtain cross-validation results. According to the cross-validation results, the extracted key features are optimized by adding new relevant features, deleting features with contribution index less than 20% or comprehensive score less than 0.4, and adjusting the feature combination method, so as to obtain optimized features. The optimized optimized features are used to build a real prediction model for heating demand, and the best model is determined by comparing the prediction performance of different models: accuracy, recall rate, and F1 score; the selected best model is deployed in the intelligent scheduling and optimization module for real-time prediction and optimization of heating demand.

[0008] Preferably, based on domain knowledge and experience, the process of preliminarily screening out preliminary features related to heating demand includes: first, from energy suppliers, smart heat meters, gas meters, and water meters, wherein the smart heat meters are used to record heating data; collecting historical consumption data of each heating type, and then displaying each heating type based on line graphs, bar graphs, and pie charts, comparing historical consumption data of different time periods, analyzing the heating usage patterns of cycles, seasons, days and nights, weekdays and holidays, identifying the peaks and troughs of heating demand and the complementarity between various thermal energies, wherein heating types include centralized heating and regional heating. At the same time, comparing the heating consumption structure of different regions, different industries or different user groups, analyzing the proportion of each heating type in total heating consumption, clarifying the consumption structure of each energy source, and using correlation analysis and regression analysis methods to explore the seasonality, diurnal changes and regional differences of each heating type, obtaining the correlation factors of each heating type, and based on the seasonality, diurnal changes and regional differences of each heating type, forming the correlation characteristics of each heating type.

[0009] Through time series analysis, the consumption data of each heating type in different time periods: seasons, months, weeks, and hours are analyzed to identify the cyclical, seasonal, and trend characteristics of heating consumption, find out the peak and valley values ​​of each heating type in different time periods, and analyze the causes of the peak and valley values ​​by comparing the consumption peaks and valley values ​​in different time periods to understand the fluctuation patterns and changing trends of heating consumption. The time period: one day, one week, one month, or one year is divided into several sub-time periods, and the heating consumption in each sub-time period is analyzed to obtain the changing patterns of heating consumption in different time periods. PowerBI is used to display the heating data in the form of charts to clarify the changing patterns and trends of heating consumption in different time periods, and based on the correlation characteristics, changing patterns, and trends of each heating type, the heating consumption patterns corresponding to each heating type are formed.

[0010] Identify external factors that affect heating demand, including weather conditions, economic activities and policy orientations. Weather conditions include temperature, humidity, wind speed and rainfall. Economic activities include industrial production, commercial activities and residents' lives. Policy orientations include energy policies and environmental protection policies. Collect academic papers, research reports and industry white papers in the fields of heating demand forecasting, energy management and energy optimization and scheduling to obtain the latest research results and practical experience. Communicate with experts, scholars, engineers and practitioners in the energy field to acquire professional knowledge and practical experience.

[0011] In combination with the heating consumption pattern and the external factors affecting the heating demand, all the features that may be related to the heating demand are listed to form a potential feature set; using the collected professional knowledge and practical experience, the potential feature set is screened to exclude the features that have little to do with the heating demand or are difficult to quantify, and the screened features are organized into a preliminary feature list; by comparing historical data, it is preliminarily verified whether the features in the preliminary feature list have a significant correlation with the heating demand, and the verification results are obtained. Based on the verification results, the preliminary feature list is adjusted, and the features that are not significantly correlated with the heating demand are deleted or new related features are added to obtain a preliminary feature set, i.e., preliminary features.

[0012] Preferably, the process of evaluating the preliminary features and obtaining the evaluation results of the importance of the preliminary features includes: for the calculation of correlation, first, according to the preset correlation threshold, the Pearson correlation coefficient is selected for evaluation of the linear relationship of the preliminary features to obtain the correlation coefficient, the Spearman rank correlation coefficient is selected for evaluation of the non-linear relationship of the preliminary features to obtain the correlation coefficient, and for the correlation evaluation between the binary variables and the continuous variables of the preliminary features, the point-by-point correlation coefficient is selected for evaluation to obtain the correlation coefficient. During the evaluation, the preliminary features whose absolute values ​​of the correlation coefficients are greater than the correlation threshold are preliminarily considered to be important features.

[0013] For categorical features, the chi-square test is used to determine whether there is a significant correlation between them and the heating demand, so as to determine their importance and obtain the importance score of the categorical features; for continuous features, variance analysis is used to evaluate the differences in heating demand at different value levels, and then determine their importance and obtain the importance score of the continuous features; the importance score based on the categorical feature and the importance score of the continuous feature together constitute the importance score of the preliminary feature.

[0014] With heating demand as the dependent variable and preliminary characteristics as the independent variables, a regression model is constructed. The regression model includes a linear regression model and a multiple linear regression model. The contribution of the characteristics is evaluated by the absolute value of the coefficient of the regression model and the significance level of the reference t-test and F-test, so as to obtain the contribution index. Among them, the absolute value of the regression coefficient reflects the influence of the characteristic on the heating demand. The larger the coefficient, the higher the contribution.

[0015] Set the weights of the correlation coefficient, importance score, and contribution index respectively, calculate the comprehensive score of each feature, adjust the weight ratio of each indicator according to actual conditions and experience, and sort the preliminary features according to the comprehensive score. The higher the score, the higher the importance of the feature. Then, according to the set importance threshold, determine which features are important and which are relatively unimportant, so as to obtain the preliminary feature importance evaluation results.

[0016] Preferably, K-fold cross-validation is used to cross-validate the extracted key features, and the process of obtaining the cross-validation results includes: first setting the K value, and then dividing the data set containing the heat consumption data and the corresponding key feature data according to the K value, each time using one subset as a test set, and combining the remaining K-1 subsets as training sets, to obtain K different combinations of training sets and test sets.

[0017] Using a training set containing historical heating consumption data and key features: weather-related features and equipment status features, the first neural network model is trained through multiple iterations to continuously adjust its weights to fit the data; after the training is completed, an initial prediction model for heating demand is obtained, and the key feature data in the test set is input into the trained initial prediction model for heating demand to obtain the predicted heating demand value; wherein, when the training set and the test set are used for the first operation, the corresponding mean square error, mean absolute error, and determination coefficient are recorded, and the evaluation index results are recorded in turn in each subsequent round; after completing K rounds of training and testing cycles, the average value of K mean square errors is calculated to reflect the average accuracy of the model in predicting heating demand based on key features, and the standard deviation is calculated to understand the stability of the model's prediction performance on different data subsets.

[0018] If the average mean square error is less than or equal to the set threshold of 0.5 and the standard deviation is less than or equal to the threshold of 0.2, it means that the key feature is accurate and stable in predicting heating demand, and the key feature can be judged to be statistically significant, that is, the p value is less than 0.05; conversely, if the average mean square error is greater than the set threshold of 0.5 and the standard deviation is greater than the threshold of 0.2, it means that the key feature may need further optimization or re-screening.

[0019] Preferably, the optimized optimization features are used to construct a true prediction model for heating demand, and the process of determining the best model by comparing the prediction performance of different models: accuracy, recall rate, and F1 score includes: based on the machine learning algorithm, a linear regression model is selected, and the optimized optimization features are used as independent variables, and the heating demand is used as the dependent variable. The model is constructed according to the modeling rules of linear regression, and the coefficients of the model are determined; for the deep learning algorithm, the long short-term memory network in the second neural network model is selected to construct a multilayer perceptron model, the number of network layers and the number of neurons in each layer are determined, and the optimized feature data is normalized and input into the neural network for training.

[0020] The data set containing the optimized features and the corresponding actual values ​​of heating demand is collected and organized, and divided into training set and test set according to the set ratio: 7:3 or 8:2.

[0021] For the selected model, the training set data is used for training. When training the linear regression model and the second neural network model, the small batch gradient descent algorithm is used to update the weight parameters of the network. After iterative training, the model gradually converges to obtain the trained linear regression model and the second neural network model.

[0022] The optimized feature data of the test set are respectively input into the trained linear regression model, the second neural network model and the initial heating demand prediction model to obtain the corresponding heating demand prediction value.

[0023] During the winter heating period, the predicted heating demand is divided into three categories: high, medium and low. The accuracy calculation formula is based on: Accuracy = (number of samples predicted correctly / total number of samples) × 100%; calculate the accuracy in the prediction performance indicator.

[0024] Regarding the focus in heat demand prediction: the heat demand higher than the set safety threshold is regarded as a positive example, based on the recall rate calculation formula: recall rate = (the number of samples that are actually positive examples and predicted correctly / the number of actual positive samples); calculate the recall rate in the prediction performance indicator.

[0025] At the same time, according to the F1 score calculation formula: F1=2×(accuracy×recall) / (accuracy+recall), the F1 score in the prediction performance indicator is calculated, and the accuracy, recall, and F1 scores calculated by each model on the test set are organized into a table for comparison.

[0026] The process of determining the best model is as follows: if you focus on the accuracy of the prediction, then choose a model with high accuracy; if you have better performance in accurately identifying specific heating demand situations, then choose a model with high recall; and the model with a higher F1 score has a more balanced overall performance, and it is selected as the best model under comprehensive consideration.

[0027] Preferably, the process of obtaining a heat distribution strategy includes: first, based on the forecast results output by the heat demand forecasting model, clarify the quantity of heat demand, demand trends, and possible heat peaks and troughs in different time periods in the future; at the same time, collect supply data of the heating system in real time, the supply data includes heat reserve, production or transmission capacity, system stability, and analyze the seasonal and cyclical fluctuations of energy supply in combination with historical supply data; to meet the heat demand, ensure that the supply of various energy sources can cover the predicted demand, take into account the maximization of energy utilization efficiency to reduce energy waste, and consider cost control, and optimize the allocation strategy based on the procurement cost, storage cost, and transmission cost of different energy sources to balance the heat supply and demand and reduce heat energy waste as the optimization goal.

[0028] By simulating the biological evolution process, the heat distribution plan is encoded into chromosomes, new plans are generated through selection, crossover and mutation, and better plans are screened according to the fitness function set in combination with the optimization goal, that is, the initial heat distribution strategy obtained based on the genetic algorithm; each heat distribution plan is regarded as the position of a particle in a multidimensional space. The particle adjusts its flight direction and speed through the speed update formula according to its own historical optimal position and the historical optimal position of the group, and continuously searches for a better solution of the particle swarm algorithm to find the position that makes the fitness function optimal, that is, the initial heat distribution strategy obtained based on the particle swarm algorithm.

[0029] According to the actual situation of the energy system, the parameters of the optimization algorithm are set. The process is: in the genetic algorithm, the population size is set, that is, the number of initial heating distribution schemes, the crossover probability is set, and the mutation probability is set; in the particle swarm algorithm, the number of particles, inertia weight, and learning factor are set; based on the optimization goal, optimization algorithm, and the parameter setting of the optimization algorithm, the optimization model is constructed.

[0030] The established optimization model is iteratively solved using the selected optimization algorithm. In each iteration, a new initial heat distribution strategy is generated based on the current prediction results, energy supply conditions, and existing heat distribution strategies. The new initial heat distribution strategy is evaluated through the fitness function. The iteration is continued until the preset maximum number of iterations is met and the stopping condition that the fitness function value changes very little after multiple rounds of iterations is met. Then the heat distribution strategy is obtained.

[0031] Preferably, the process of the heat mutual assistance control module establishing a mutual assistance mechanism between different energy forms is: based on the heat consumption data collected by the data acquisition and monitoring module, the heat consumption data not only includes the real-time flow, power, pressure, and temperature, but also involves the operating status information of each energy facility; analyzing the usage patterns of different energy sources in the past year and quarter in terms of seasons, days and nights, weekdays and holidays, identifying the peak and trough periods of heat demand and the complementary characteristics between each energy source, and obtaining energy data monitoring and analysis results.

[0032] For heating and natural gas, study gas boiler heating technology and natural gas heat energy conversion technology, determine the feasibility of using natural gas boiler heating to supplement heating gaps or using excess heat energy for heat storage under current system conditions, and analyze the applicability of combined heating, heat recovery and heat energy conversion technologies for heating and auxiliary heat energy. Combined with conversion efficiency, equipment cost, operation and maintenance difficulty, and local heating market prices and policy orientations, evaluate the economic feasibility of heat energy conversion and supplementary solutions. At the same time, consider the procurement cost, sales price, subsidy policy and carbon trading cost of different energy sources; draw conclusions on the feasibility of conversion between different energy sources; According to the results of energy data monitoring and analysis and the conclusions of feasibility assessment, a preliminary energy mutual assistance strategy is formulated. The energy mutual assistance strategy covers the timing of energy conversion, conversion ratio, and energy storage equipment scheduling. The preliminary mutual assistance strategy is then optimized using an intelligent optimization algorithm. The objective function is established with the optimization goals of maximizing energy supply stability and minimizing the total cost of energy utilization. During the optimization process, the real-time constraints of the energy system are combined: the maximum / minimum output limit of energy production equipment, the capacity limit of energy storage equipment, and the power constraint of the energy transmission network. Through multiple iterative solutions, the mutual assistance strategy parameters are continuously adjusted to obtain the mutual assistance strategy.

[0033] (III) Beneficial effects The present invention provides a comprehensive management system for urban energy mutual assistance, which has the following beneficial effects: (I) The city's energy mutual assistance integrated management system determines the best model by combining historical heating consumption data, key features, and optimized features after the key features are optimized; using the best model, the heating demand is predicted based on real-time monitoring data and historical data, avoiding the shortcomings of a single prediction model, reducing the deviation of the prediction results, and improving the accuracy of the prediction of heating demand; this enables the energy supply plan to be further matched with actual demand, thereby effectively avoiding heating shortages during peak heating demand periods and ensuring normal heating in the city; and preventing heat energy waste during low periods and reducing unnecessary cost increases.

[0034] (II) The city's energy mutual assistance integrated management system, based on the energy data of the data collection and monitoring module, deeply analyzes the usage patterns and complementary characteristics of different energy sources in different time periods, and comprehensively evaluates the technical and economic feasibility of energy conversion; accordingly, it formulates and optimizes energy mutual assistance strategies, and can use natural gas boilers for heating or adopt combined heating technology to supplement when heating is insufficient, or reasonably convert and store energy when there is excess energy; it not only enhances the stability of the heating system and reduces the risks caused by fluctuations in the supply of a single energy source, but also reduces the overall thermal energy utilization cost by optimizing the energy combination, thereby improving the overall economy and risk resistance of the city's energy system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the overall framework of the present invention; Figure 2 It is a schematic diagram of the framework of the intelligent scheduling and optimization module of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] See also Figure 1 and Figure 2 The present invention provides a technical solution: a comprehensive management system for urban energy mutual assistance, comprising: Data collection and monitoring module, used to collect heat consumption data in urban commercial areas or industrial parks in real time, and transmit the heat consumption data to the central database; Intelligent scheduling and optimization module, which is used to predict changes in heating demand in advance and optimize heating distribution strategies based on historical data and real-time monitoring of heating consumption data; The energy storage system management module is used to control the energy storage system to store excess heat energy when the heating demand is low, and release it during peak hours to balance the load of the heating system; The heating mutual assistance control module is used to establish a mutual assistance mechanism between different energy forms. When the heating supply is insufficient, it can be supplemented by using natural gas boilers or combined heating technology to achieve energy complementarity and optimal configuration. The heating mutual assistance control module can coordinate the supply of different energy forms to ensure that when the heating supply is insufficient, it can be supplemented by other energy forms.

[0038] The data acquisition and monitoring module collects the heat consumption data in the urban commercial area or industrial park in real time, and transmits the heat consumption data to the central database. The process includes: according to the actual energy consumption of the urban commercial area or industrial park, select heating sensors, water flow sensors, gas flow sensors, and data collectors, and deploy them on energy consumption equipment accordingly. Energy consumption equipment includes heat meters, water meters, gas meters, and heating systems. The heating sensors are deployed on heat meters and heating systems, water flow sensors are deployed on water meters, and gas meters are deployed on gas meters. By deploying corresponding sensors on these major energy consumption equipment, it is convenient to cover the main energy consumption points. The data collector collects the original data of heat consumption in real time. The original data includes heat, water, and gas consumption data; the original data is deduplicated and denoised to obtain heat consumption data, and transmitted to the central database.

[0039] The intelligent scheduling and optimization module predicts the change of heating demand in advance and optimizes the heating distribution strategy based on the historical data and the real-time monitored heating consumption data. The process includes: firstly, collecting the historical heating consumption data and the data of the factors affecting the heating consumption, and then extracting the key features affecting the heating demand by using the feature selection algorithm based on the historical data and the real-time monitored heating consumption data; then, determining the best model by combining the historical heating consumption data, the key features and the optimized features after the key features are optimized; using the best model, predicting the heating demand according to the real-time monitoring data and the historical data, and obtaining the prediction results; according to the prediction results, combining the heating supply situation and the heating distribution strategy, using the intelligent optimization algorithm to schedule and optimize the heating in real time, and obtaining the heating distribution strategy, wherein the intelligent optimization algorithm includes the genetic algorithm and the particle swarm algorithm.

[0040] Based on historical data and real-time monitored heating consumption data, the process of extracting key features that affect heating demand using feature selection algorithms includes: obtaining heating consumption data from a central database, and collecting influencing factor data through real-time monitoring. The influencing factor data include weather conditions, equipment operating status, and personnel activities. The collected heating consumption data and influencing factor data are cleaned to remove duplicate, missing, or outlier values. Based on domain knowledge and experience, preliminary features related to heating demand are preliminarily screened out. The preliminary features include time, weather conditions, equipment status, and personnel activities. Time is season, month, week, and hour. Weather conditions include temperature, humidity, wind speed, and rainfall. Equipment status includes equipment on / off status and operating efficiency. Personnel activities include the number of people and activity intensity. Based on statistical methods, the importance of features is evaluated by calculating the correlation between preliminary features and target variables, namely, heating demand, importance scores, and contribution indicators. The preliminary features are evaluated to obtain preliminary feature importance evaluation results.

[0041] According to the evaluation results of the preliminary feature importance, the features with an absolute value of the correlation coefficient greater than 0.6, that is, the absolute value of the Pearson correlation coefficient greater than 0.6, the importance score greater than 70 points (percentage), and the contribution index greater than 30% (weight ratio) are selected as key features and recorded as key features. The key features will be used in the subsequent heating demand prediction and optimization scheduling model. The extracted key features are cross-validated using K-fold cross-validation to obtain cross-validation results. According to the cross-validation results, the extracted key features are optimized by adding new relevant features, deleting features with a contribution index less than 20% (weight ratio) or a comprehensive score less than 0.4, and adjusting the feature combination method, so as to obtain optimized features, that is, features adjusted by K-fold cross-validation. The optimized optimized features are used to construct a real prediction model for heating demand, and the best model is determined by comparing the prediction performance of different models: accuracy, recall rate, and F1 score; the selected best model is deployed in the intelligent scheduling and optimization module for real-time prediction and optimization of heating demand.

[0042] Based on domain knowledge and experience, the process of preliminarily screening out preliminary features related to heating demand includes: first, collecting historical consumption data of each heating type from energy suppliers, smart heat meters, gas meters, and water meters, and then displaying each heating type based on line charts, bar charts, and pie charts, comparing historical consumption data of different time periods, analyzing the heating usage patterns of cycles, seasons, days and nights, weekdays, and holidays, identifying the peaks and troughs of heating demand and the complementarity between various thermal energies. Among them, heating types include centralized heating and regional heating. At the same time, comparing the heating consumption structure of different regions, different industries, or different user groups, analyzing the proportion of each heating type in total heating consumption, clarifying the consumption structure of each energy source, and using correlation analysis and regression analysis methods to explore the seasonality, diurnal changes, and regional differences of each heating type, and obtain the correlation factors of each heating type, and based on the seasonality, diurnal changes, and regional differences of each heating type, form the correlation characteristics of each heating type; Through time series analysis, the consumption data of each heating type in different time periods: seasons, months, weeks, and hours are analyzed to identify the cyclical, seasonal, and trend characteristics of heating consumption, find out the peak and valley values ​​of each heating type in different time periods, and analyze the causes of the peak and valley values ​​by comparing the consumption peaks and valley values ​​in different time periods to understand the fluctuation patterns and changing trends of heating consumption. The time period: one day, one week, one month, or one year is divided into several sub-time periods, and the heating consumption in each sub-time period is analyzed to obtain the changing patterns of heating consumption in different time periods. PowerBI is used to display the heating data in the form of charts to clarify the changing patterns and trends of heating consumption in different time periods, and based on the correlation characteristics, changing patterns, and trends of each heating type, the heating consumption patterns corresponding to each heating type are formed.

[0043] Identify external factors that affect heating demand, including weather conditions, economic activities and policy orientations. Weather conditions include temperature, humidity, wind speed and rainfall. Economic activities include industrial production, commercial activities and residents' lives. Policy orientations include energy policies and environmental protection policies. Collect academic papers, research reports and industry white papers in the fields of heating demand forecasting, energy management and energy optimization and scheduling to obtain the latest research results and practical experience. Communicate with experts, scholars, engineers and practitioners in the energy field to acquire professional knowledge and practical experience.

[0044] In combination with the heating consumption pattern and the external factors affecting the heating demand, all the features that may be related to the heating demand are listed to form a potential feature set; using the collected professional knowledge and practical experience, the potential feature set is screened to exclude the features that have little to do with the heating demand or are difficult to quantify, and the screened features are organized into a preliminary feature list; by comparing historical data, it is preliminarily verified whether the features in the preliminary feature list have a significant correlation with the heating demand, and the verification results are obtained. Based on the verification results, the preliminary feature list is adjusted, and the features that are not significantly correlated with the heating demand are deleted or new related features are added to obtain a preliminary feature set, i.e., preliminary features.

[0045] The process of evaluating the preliminary features and obtaining the evaluation results of the importance of the preliminary features includes: for the calculation of the correlation, first according to the preset correlation threshold, the Pearson correlation coefficient is selected for the linear relationship of the preliminary features to be evaluated to obtain the correlation coefficient, the Spearman rank correlation coefficient is selected for the non-linear relationship of the preliminary features to be evaluated to obtain the correlation coefficient, for the correlation evaluation between the binary variables and the continuous variables of the preliminary features, the point-by-point correlation coefficient is selected for evaluation to obtain the correlation coefficient, during the evaluation period, the preliminary features whose absolute values ​​of the correlation coefficients are greater than the correlation threshold are preliminarily considered to be important features; the correlation threshold is set to 0.5, if the absolute value of the correlation coefficient between the preliminary features and the heating demand is greater than 0.5, then the preliminary features are preliminarily considered to be important features.

[0046] For categorical features, such as the feature of the equipment on / off status, the chi-square test is used to determine whether there is a significant correlation between it and the heating demand, so as to determine its importance and obtain the importance score of the categorical feature; for continuous features such as temperature, variance analysis is used to evaluate the difference in heating demand at different value levels, and then determine its importance and obtain the importance score of the continuous feature; the importance score based on the categorical feature and the importance score of the continuous feature together constitute the importance score of the preliminary feature.

[0047] With heating demand as the dependent variable and preliminary characteristics as the independent variables, a regression model is constructed. The regression model includes a linear regression model and a multiple linear regression model. The contribution of the feature is evaluated by the absolute value of the coefficient of the regression model and the significance level of the t-test and F-test, so as to obtain the contribution index. Among them, the absolute value of the regression coefficient reflects the influence of the feature on the heating demand. The larger the coefficient, the higher the contribution. At the same time, referring to the statistics: the significance level of the t-test and the F-test, the significant features have a higher contribution, that is, the contribution index based on the coefficient weight of the regression model is greater than 35%.

[0048] Set the weights of the correlation coefficient, importance score, and contribution index respectively, and calculate the comprehensive score of each feature. The weight ratio of each indicator can be adjusted according to actual conditions and experience. For example, set the correlation coefficient to 40%, the importance score to 30%, and the contribution index to 30%, and calculate the comprehensive score of each feature; sort the preliminary features according to the comprehensive score, the higher the score, the higher the importance of the feature, and then, according to the set importance threshold, determine which features are important and which are relatively unimportant, so as to obtain the preliminary feature importance assessment results; for example, set the features with a comprehensive score greater than 0.6 as important features, and those with a comprehensive score less than 0.6 as relatively unimportant features.

[0049] K-fold cross validation is used to cross-validate the extracted key features. The process of obtaining the cross-validation results includes: first setting the K value, and then dividing the data set containing the heating consumption data and the corresponding key feature data according to the K value, each time taking one of the subsets as the test set, and combining the remaining K-1 subsets as the training set, to obtain K different combinations of training sets and test sets. If the data set is of moderate size, K=5 can be selected to divide the entire data set roughly evenly into 5 subsets; using the training set containing historical heating consumption data, key features: weather-related features, equipment status features, through The first neural network model is trained iteratively for multiple times to continuously adjust its weights to fit the data. After the training is completed, the initial prediction model for heating demand is obtained. The performance of the trained model is tested using the corresponding test set, and the corresponding evaluation indicators are calculated: mean square error, mean absolute error, and determination coefficient. The key feature data in the test set are input into the trained initial prediction model for heating demand to obtain the predicted heating demand value, which is then compared with the actual heating demand value in the test set to calculate the mean square error, mean absolute error, and determination coefficient to measure the prediction accuracy of the model on the test set. In each round of training and testing, In the process, the calculated evaluation index values ​​are recorded, including: when the training set and test set are used for the first operation, the corresponding mean square error, mean absolute error, and determination coefficient are recorded. The evaluation index results are recorded in each subsequent round, and the calculation results reflect the accuracy and stability of the heating demand forecast; after completing K rounds of training and testing cycles, the average value of K mean square errors is calculated to reflect the average accuracy of the model in predicting heating demand based on key features, and the standard deviation is calculated to understand the stability of the model's prediction performance on different data subsets; if the average mean square error is less than or equal to the set The threshold is set to 0.5 and the standard deviation is less than or equal to the threshold of 0.2, indicating that the key features are accurate and stable in predicting heating demand. The accuracy is that the mean square error is less than 0.5 (unit: GJ) and the mean absolute error is less than 0.3 (unit: GJ). It can be judged that the key features are statistically significant, that is, the p value is less than 0.05; conversely, if the average mean square error is greater than the set threshold of 0.5 and the standard deviation is greater than the threshold of 0.2, it means that the key features may need further optimization or re-screening, so as to give a comprehensive evaluation result on the stability and accuracy of the key features in predicting heating demand.

[0050] The optimized optimization features are used to build a true prediction model for heating demand, and the prediction performance of different models is compared: accuracy, recall rate, and F1 score to determine the best model. The process includes: based on the machine learning algorithm, a linear regression model is selected, and the optimized optimization features are used as independent variables, and the heating demand is used as the dependent variable. The model is constructed according to the modeling rules of linear regression, and the coefficients of the model are determined; for the deep learning algorithm, the long short-term memory network in the second neural network model is selected to build a multi-layer perceptron model, and the number of network layers and the number of neurons in each layer are determined. After the optimized feature data is normalized, it is input into the neural network for training.

[0051] Collect and organize data sets containing optimized features and corresponding actual values ​​of heating demand, and divide them into training set and test set according to the set ratio of 8:2; use 80% of the data as training set for model training to let the model learn the relationship between features and heating demand, and use the remaining 20% ​​of the data as test set for subsequent evaluation of the prediction performance of the model; for the selected model, use the training set data for training, and use the small batch gradient descent algorithm to update the weight parameters of the network when training the linear regression model and the second neural network model. After iterative training, the model gradually converges to obtain the trained linear regression model and the second neural network model; input the optimized feature data of the test set into the trained linear regression model, the second neural network model and the initial prediction model of heating demand, respectively, to obtain the corresponding heating demand prediction value; for example, input the equipment status and weather conditions on a certain day in the test set into different models, and output the prediction results of the heating demand on that day.

[0052] During the winter heating period, the predicted heating demand is divided into three categories: high, medium and low. The accuracy calculation formula is based on: Accuracy = (number of correctly predicted samples / total number of samples) × 100%; calculate the accuracy in the prediction performance indicator; for example, there are 100 samples in the test set, and the model accurately predicts the heating demand category of 80 samples, then the accuracy is 80%.

[0053] The recall rate mainly measures the proportion of positive examples that the model can correctly identify to the actual positive examples. For the heat demand forecast, the heat demand higher than the set safety threshold is regarded as a positive example. The recall rate calculation formula is: recall rate = (the number of samples that are actually positive and predicted correctly / the number of actual positive samples); calculate the recall rate in the prediction performance indicator. For example, there are 50 samples with actual heat demand higher than the safety threshold, and the model accurately predicts 40 of them, then the recall rate is 80%.

[0054] At the same time, according to the F1 score calculation formula: F1=2×(accuracy×recall) / (accuracy+recall); calculate the F1 score in the prediction performance indicator; by calculating the F1 score, it can more comprehensively reflect the performance of the model and avoid the one-sidedness caused by relying solely on accuracy or recall; organize the accuracy, recall, and F1 scores calculated by each model on the test set into a table for comparison.

[0055] The process of determining the best model is as follows: if you focus on the accuracy of the prediction, then select a model with high accuracy; if you want better performance in accurately identifying specific heating demand situations, then select a model with high recall; and the model with a higher F1 score has a more balanced overall performance, and it is selected as the best model under comprehensive consideration; for example, through comparison, it is found that the F1 score of model Y is the highest among all compared models, then model Y can be determined as the best model for subsequent application scenarios of real-time prediction of heating demand and optimized scheduling.

[0056] The process of obtaining a heat distribution strategy includes: first, based on the forecast results output by the heat demand forecasting model, clarify the quantity of heat demand, demand trends, and possible heat peaks and troughs in different time periods in the future; at the same time, collect the supply data of the heating system in real time, the supply data includes heat reserve, production or transmission capacity, system stability, and combine with historical supply data to analyze the seasonal and cyclical fluctuations in energy supply; in order to meet the heat demand, ensure that the supply of various energy sources can cover the predicted demand, take into account the maximization of energy utilization efficiency to reduce energy waste, and consider cost control, and optimize the allocation strategy based on the procurement cost, storage cost, and transmission cost of different energy sources to balance the heat supply and demand and reduce the waste of heat energy as the optimization goal.

[0057] By simulating the biological evolution process, the heat distribution plan is encoded into chromosomes, new plans are generated through selection, crossover and mutation, and better plans are screened according to the fitness function set in combination with the optimization goal, that is, the initial heat distribution strategy obtained based on the genetic algorithm, where the fitness function is constructed with the objective function of maximizing energy utilization efficiency and minimizing cost; each heat distribution plan is regarded as the position of a particle in a multidimensional space, and the particle adjusts its flight direction and speed through the speed update formula according to its own historical optimal position and the historical optimal position of the group, and continuously searches for a better solution of the particle swarm algorithm to find the position that makes the fitness function optimal, that is, the initial heat distribution strategy obtained based on the particle swarm algorithm.

[0058] According to the actual situation of the energy system, the parameters of the optimization algorithm are set. The process is: in the genetic algorithm, the population size is set, that is, the number of initial heating distribution schemes, the crossover probability is set, and the mutation probability is set; in the particle swarm algorithm, the number of particles, inertia weight, and learning factor are set; based on the optimization goal, optimization algorithm, and the parameter setting of the optimization algorithm, the optimization model is constructed.

[0059] The established optimization model is iteratively solved using the selected optimization algorithm. In each iteration, a new initial heat distribution strategy is generated based on the current prediction results, energy supply conditions, and existing heat distribution strategies. The new initial heat distribution strategy is evaluated through the fitness function. The iteration is continued until the preset maximum number of iterations is met and the stopping condition that the fitness function value changes very little after multiple rounds of iterations is met. Then the heat distribution strategy is obtained.

[0060] Using computer simulation technology, the generated heat distribution strategy is substituted into the virtual energy system model to simulate the energy supply and demand situation in the future and monitor whether there will be energy shortages, energy waste or excessive costs. If any problems are found, the heat distribution strategy will be adjusted in time. The adjustment methods include fine-tuning the parameters of the optimization algorithm, regenerating the distribution plan, or modifying some heat distribution ratios in a targeted manner according to actual problems. The operation of the actual energy system is continuously monitored and compared with the predicted results. The energy supply data is updated in real time. When the absolute error between the actual demand and the predicted demand is greater than 15% or the error is greater than 10% for three consecutive time periods, the optimization process is triggered again to ensure that the heat distribution strategy always adapts to the actual demand.

[0061] The process of the heating mutual assistance control module to establish a mutual assistance mechanism between different energy forms is as follows: based on the heating consumption data collected by the data acquisition and monitoring module, the heating consumption data not only includes the real-time flow, power, pressure, and temperature, but also involves the operating status information of each energy facility; analyzing the usage patterns of different heating types in the past year and quarter in terms of seasons, days and nights, weekdays and holidays, identifying the peak and trough periods of heating demand and the complementary characteristics between various energy sources, and obtaining the energy data monitoring and analysis results; for example, it is found that the demand for heating in commercial areas is strong during the day in winter, while the demand for hot water in residential areas increases at night, the consumption of natural gas is large during the winter heating season, and the heating demand is relatively stable in some periods, which can be used to supplement the demand patterns of other energy conversion.

[0062] For heating and natural gas, we study gas boiler heating technology and natural gas heat energy conversion technology, determine the feasibility of using natural gas boiler heating to supplement the heating gap or using excess heat energy for heat storage under the current system conditions, and for electricity and heating, analyze the applicability of waste heat technology for combined heat and power, electric heat pumps, and organic Rankine cycle power generation. Combined with conversion efficiency, equipment cost, operation and maintenance difficulty, and local heating market prices and policy orientations, we evaluate the economic feasibility of heat energy conversion and supplementary solutions. At the same time, we consider the procurement cost, sales price, subsidy policy, and carbon trading cost of different energy sources; and draw a feasibility evaluation of conversion between different energy sources. Evaluation conclusions; According to the energy data monitoring and analysis results and the feasibility assessment conclusions, a preliminary energy mutual assistance strategy is formulated. The energy mutual assistance strategy covers the timing of energy conversion, conversion ratio, and energy storage equipment scheduling; Then the intelligent optimization algorithm is used to optimize the preliminary mutual assistance strategy, with maximizing energy supply stability and minimizing the total cost of energy utilization as the optimization goals, and an objective function is established. In the optimization process, the real-time constraints of the energy system are combined: the maximum / minimum output limit of energy production equipment, the capacity limit of energy storage equipment, and the power constraint of the energy transmission network; Through multiple iterative solutions, the mutual assistance strategy parameters are continuously adjusted to obtain the mutual assistance strategy.

[0063] It should be further explained that, in the specific implementation process, by collecting historical data on heat consumption and influencing factors, using feature selection algorithms to extract key features, and determining the best model through K-fold cross-validation and model comparison, accurate prediction of heat demand can be achieved; at the same time, based on the prediction results and combined with the heat supply situation, an intelligent optimization algorithm is used to optimize heat distribution, effectively improve energy utilization efficiency, reduce energy waste, ensure a dynamic balance between heat supply and demand, meet the heat demand of urban commercial areas or industrial parks at different times, and ensure the stable operation of production and life.

[0064] Based on the energy data of the data collection and monitoring module, we deeply analyze the usage patterns and complementary characteristics of different energy sources in different time periods, and comprehensively evaluate the technical and economic feasibility of energy conversion. Based on this, we formulate and optimize energy mutual assistance strategies. When the heating supply is insufficient, we can use natural gas boilers for heating or adopt combined heating technology to supplement it, or reasonably convert and store energy when there is a surplus of energy. This not only enhances the stability of the heating system and reduces the risks caused by fluctuations in the supply of a single energy source, but also reduces the overall thermal energy utilization cost by optimizing the energy combination, thereby improving the overall economy and risk resistance of the urban energy system.

[0065] The system forms a complete data-driven management closed loop from heat consumption data collection, key feature extraction, model construction to the formulation and implementation of heat distribution and mutual assistance strategies; during operation, the actual heating system operation data is continuously monitored, and strategies and model parameters are adjusted in a timely manner based on feedback, such as simulation verification and dynamic adjustment of heat distribution strategies, as well as updating and optimization of key features and prediction models; this continuous optimization mechanism enables the system to adapt to the dynamic changes in urban heat demand and supply, always maintain an efficient operation state, provide intelligent and dynamic solutions for urban heat management, and contribute to the sustainable development of urban energy.

[0066] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0067] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An integrated urban energy management system, characterized in that: include: Data collection and monitoring module, used to collect heat consumption data in urban commercial areas or industrial parks in real time, and transmit the heat consumption data to the central database; The intelligent scheduling and optimization module is used to collect historical and real-time heating data and influencing factors, extract key features through feature selection and determine the best prediction model, and complete the real-time prediction and scheduling optimization of heating demand based on the prediction results and the heating supply and allocation strategy, that is, optimize the heating allocation strategy and obtain the heating allocation strategy; The energy storage system management module is used to control the energy storage system to store excess heat energy during low heat demand periods and release it during peak hours; The heating mutual assistance control module is used to establish a mutual assistance mechanism between different energy forms. When the heating is insufficient, it can be supplemented by using natural gas boilers or combined heating technology.

2. The urban energy mutual aid integrated management system according to claim 1 is characterized by: The data acquisition and monitoring module collects the heat consumption data in the urban commercial area or industrial park in real time, and transmits the heat consumption data to the central database. The process includes: according to the actual energy consumption of the urban commercial area or industrial park, selecting heating sensors, water flow sensors, gas flow sensors, and data collectors, and deploying them on energy consumption equipment accordingly. The data collector collects the original data of heat consumption in real time, deduplicates and denoises the original data, obtains the heat consumption data, and transmits it to the central database.

3. The urban energy mutual aid integrated management system according to claim 2 is characterized by: The process of the intelligent scheduling and optimization module predicting the change of heating demand in advance and optimizing the heating distribution strategy based on historical data and real-time monitored heating consumption data includes: collecting historical heating consumption data and data on factors affecting heating consumption, and extracting key features that affect heating demand based on historical data and real-time monitored heating consumption data using a feature selection algorithm; determining the best model by combining historical heating consumption data, key features, and optimized features after the key features are optimized; using the best model, predicting the heating demand based on real-time monitoring data and historical data to obtain a prediction result; based on the prediction result, combining the heating supply situation and the heating distribution strategy, using an intelligent optimization algorithm to schedule and optimize the heating in real time to obtain a heating distribution strategy, wherein the intelligent optimization algorithm includes a genetic algorithm and a particle swarm algorithm.

4. The city energy mutual aid integrated management system according to claim 3 is characterized by: Based on historical data and real-time monitored heating consumption data, the process of extracting key features that affect heating demand using feature selection algorithms includes: obtaining heating consumption data from a central database, and collecting influencing factor data through real-time monitoring, cleaning the collected heating consumption data and influencing factor data, removing duplicate, missing or outlier values, and preliminarily screening out preliminary features related to heating demand based on domain knowledge and experience. The preliminary features include time, weather conditions, equipment status, and personnel activities; based on statistical methods, the importance of features is evaluated by calculating the correlation, importance score, and contribution index between the preliminary features and heating demand, and the preliminary features are evaluated to obtain preliminary feature importance evaluation results; based on the preliminary feature importance evaluation results, features with an absolute value of a correlation coefficient greater than 0.6, an importance score greater than 70 points, and a contribution index greater than 30% are selected as key features and recorded as key features.

5. The city energy mutual aid integrated management system according to claim 4 is characterized by: K-fold cross-validation is used to cross-validate the extracted key features to obtain cross-validation results. According to the cross-validation results, the extracted key features are optimized by adding new relevant features, deleting features with contribution indicators less than 20% or comprehensive scores less than 0.4, and adjusting the feature combination method to obtain optimized features. The optimized features are used to build a real prediction model for heating demand, and the best model is determined by comparing the accuracy, recall rate, and F1 score of different models. The selected best model is deployed in the intelligent scheduling and optimization module for real-time prediction and optimization of heating demand.

6. The city energy mutual aid integrated management system according to claim 5 is characterized by: K-fold cross validation is used to cross-validate the extracted key features, and the process of obtaining the cross-validation results includes: setting the K value, dividing the data set containing the heat consumption data and the corresponding key feature data according to the K value, taking one subset as the test set each time, and combining the remaining K-1 subsets as the training set, to obtain K different combinations of training sets and test sets, using the training set containing historical heat consumption data, key features: weather-related features, and equipment status features, to train the first neural network model through multiple iterations, so that it continuously adjusts the weights to fit the data; after the training is completed, the initial prediction model for heat demand is obtained; the key feature data in the test set is input into the trained initial prediction model for heat demand to obtain the predicted heat demand value; wherein, when the training set and the test set are used for the first operation, the corresponding mean square error, mean absolute error, and determination coefficient are recorded, and the evaluation index results are recorded in turn in each subsequent round; after completing K rounds of training and testing cycles, the average value of the K-times mean square error and the calculation standard deviation are calculated; If the average mean square error is less than or equal to the set threshold of 0.5 and the standard deviation is less than or equal to the threshold of 0.2, the key feature is judged to be statistically significant, that is, the p value is less than 0.05; conversely, if the average mean square error is greater than the set threshold of 0.5 and the standard deviation is greater than the threshold of 0.2, the key feature is re-optimized or screened.

7. The city energy mutual aid integrated management system according to claim 6 is characterized by: The process of evaluating the preliminary features and obtaining the evaluation results of the importance of the preliminary features includes: for the calculation of correlation, according to the preset correlation threshold, the Pearson correlation coefficient is selected for the linear relationship of the preliminary features to evaluate and obtain the correlation coefficient; for the nonlinear relationship of the preliminary features, the Spearman rank correlation coefficient is selected for evaluation and obtain the correlation coefficient; for the correlation evaluation between the binary variables and the continuous variables of the preliminary features, the point-by-point correlation coefficient is selected for evaluation and obtain the correlation coefficient; for the categorical features, the chi-square test is used to determine whether there is a significant correlation between them and the heating demand, determine their importance, and obtain the importance score of the categorical features; for the continuous features, the variance test is used to evaluate the correlation between the binary variables and the continuous variables, ... The difference in heating demand under different value levels is evaluated by analysis, its importance is determined, and the importance score of continuous features is obtained; the importance score of the categorical features and the importance score of the continuous features are jointly constructed into the importance score of the preliminary features; the heating demand is taken as the dependent variable and the preliminary features are taken as the independent variables, and a regression model is constructed. The contribution of the features is evaluated by the absolute value of the coefficient of the regression model and the significance level of the reference t-test and F-test, and the contribution index is obtained; the weights of the correlation coefficient, importance score, and contribution index are set respectively, and the comprehensive score of each feature is calculated. The preliminary features are sorted according to the comprehensive score, and the evaluation results of the importance of the preliminary features are obtained according to the set importance threshold.

8. The city energy mutual aid integrated management system according to claim 7 is characterized by: The process of using the optimized optimization features to build a true prediction model for heating demand and determining the best model by comparing the prediction performance of different models includes: based on the machine learning algorithm, selecting a linear regression model, using the optimized optimization features as independent variables and the heating demand as the dependent variable, building the model according to the modeling rules of linear regression, and determining the coefficients of the model; for the deep learning algorithm, selecting the long short-term memory network in the second neural network model, building a multi-layer perceptron model, determining the number of network layers and the number of neurons in each layer, normalizing the optimization feature data, and inputting it into the neural network for training; collecting and arranging data sets containing the optimization features and the corresponding actual values ​​of the heating demand, and dividing them into training sets and test sets according to the set ratio; for the selected model, using the training set data for training, and using a small batch ladder when training the linear regression model and the second neural network model. The degree descent algorithm is used to update the weight parameters of the network. After iterative training, the model converges to obtain the trained linear regression model and the second neural network model; the optimized feature data of the test set are respectively input into the linear regression model, the second neural network model and the initial prediction model of the heating demand to obtain the corresponding heating demand prediction value; during the winter heating period, the predicted heating demand is divided into three categories: high, medium and low. Based on the accuracy calculation formula, the accuracy of the prediction performance index is calculated; for the heating demand prediction, the heating demand higher than the set safety threshold is regarded as a positive example, and the recall rate in the prediction performance index is calculated based on the recall rate calculation formula; at the same time, according to the F1 score calculation formula, the F1 score in the prediction performance index is calculated; the accuracy, recall rate and F1 score calculated by each model on the test set are sorted into a table for comparison to determine the best model.

9. The city energy mutual aid integrated management system according to claim 8, characterized in that: The process of obtaining the heat distribution strategy includes: based on the forecast results output by the heat demand forecasting model, clarify the heat demand quantity, demand trend and possible heat supply peaks and valleys in different time periods in the future; at the same time, collect the supply data of the heating system in real time, and analyze the seasonal and cyclical fluctuations of energy supply in combination with historical supply data; in order to meet the heat demand, ensure that the supply of various energy sources can cover the predicted demand, take into account the maximization of energy utilization efficiency to reduce energy waste, and consider cost control, and optimize the allocation strategy based on the procurement cost, storage cost and transportation cost of different energy sources to balance the heat supply and demand and reduce the heat energy waste as the optimization goal; By simulating the biological evolution process, the heat distribution scheme is encoded into chromosomes, new schemes are generated through selection, crossover and mutation, and better schemes are screened according to the fitness function set in combination with the optimization goal, that is, the initial heat distribution strategy obtained based on the genetic algorithm; each heat distribution scheme is regarded as the position of a particle in a multidimensional space, and the particle adjusts its flight direction and speed through the speed update formula according to its own historical optimal position and the historical optimal position of the group, and continuously searches for a better solution of the particle swarm algorithm to find the position that makes the fitness function optimal, that is, the initial heat distribution strategy obtained based on the particle swarm algorithm; According to the actual situation of the energy system, the parameters of the optimization algorithm are set. The process is as follows: in the genetic algorithm, the population size is set, that is, the number of initial heat distribution schemes, the crossover probability is set, and the mutation probability is set; in the particle swarm algorithm, the number of particles, inertia weight, and learning factor are set; based on the optimization goal, optimization algorithm, and the parameter setting of the optimization algorithm, the optimization model is constructed; The established optimization model is iteratively solved using the selected optimization algorithm. In each iteration, a new initial heat distribution strategy is generated based on the current prediction results, energy supply conditions, and existing heat distribution strategies. The new initial heat distribution strategy is evaluated through the fitness function. The iteration is continued until the preset maximum number of iterations is met and the stopping condition that the fitness function value changes very little after multiple rounds of iterations is met. Then the heat distribution strategy is obtained.

10. The city energy mutual aid integrated management system according to claim 9, characterized in that: The process of the heating mutual assistance control module establishing a mutual assistance mechanism between different energy forms is: based on the heating consumption data collected by the data acquisition and monitoring module, the usage patterns of different energy sources in the past year and quarter in terms of seasons, days and nights, weekdays and holidays are analyzed, the peak and trough periods of heating demand and the complementary characteristics between various energy sources are identified, and the energy data monitoring and analysis results are obtained.

Images