Smart energy management system and method based on big data analysis

Abstract

The invention relates to a smart energy management system and method based on big data analysis, and the system comprises an energy demand prediction module which is used for inputting the historical energy demand quantity into an energy demand prediction model, so as to determine the energy demand prediction quantity of a to-be-managed building site; the energy risk prediction module is used for calling the energy risk assessment model to determine an energy risk value of the to-be-managed building site according to the historical energy risk site and the energy risk frequency; the energy distribution control module is used for constructing an energy distribution control model corresponding to the target area according to the energy demand predicted quantity and the energy risk value corresponding to each to-be-managed building site, and solving the energy distribution control model by adopting a multi-target optimization algorithm so as to determine the energy distribution quantity of each to-be-managed building site; and the energy management module is used for inputting the energy distribution amount to the corresponding to-be-managed building site. According to the invention, the energy utilization efficiency is improved, and the safety redundancy capability of the energy management system is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy management, in particular to a smart energy management system and method based on big data analysis. BACKGROUND

[0002] Currently, energy management in scenarios such as industrial parks, commercial complexes, and urban functional areas is facing multiple challenges such as inaccurate demand prediction, lack of risk assessment, and rigid distribution logic. Traditional management models have been unable to adapt to the actual needs of complex scenarios.

[0003] In China, carbon emissions during the operation stage of buildings account for a large proportion of the country's total emissions. Industrial parks, commercial complexes, and urban functional areas are major energy-consuming entities, and traditional energy systems generally have prominent problems such as high carbon dependence, low energy efficiency, and extensive management. The main technical defects are as follows: Firstly, outdoor temperature, personnel density, equipment operating state, and other multi-dimensional factors can cause fluctuations in energy demand. Traditional energy demand prediction models are prone to dimension disaster when dealing with multi-factor coupled long-period data. Moreover, these models rely on manual operation and experience-based judgment, with low intelligence and insufficient prediction accuracy. This directly leads to energy distribution based on experience values, often resulting in insufficient supply during high demand and energy waste during low demand. This not only fails to adapt to the dynamic changes in building cooling load, but also cannot meet the deep decarbonization demand.

[0004] Secondly, buildings in industrial parks, commercial complexes, and urban functional areas differ significantly in terms of geographical location and functional attributes. For example, the risk tolerance of buildings such as hospitals and core production plants is much lower than that of ordinary commercial buildings. They often face energy risks such as pipe leakage, power grid fluctuations, and equipment failures leading to energy interruptions. Traditional energy management systems often use a "one-size-fits-all" equalization energy supply strategy, without developing differentiated protection mechanisms for different buildings based on their risk levels. This results in unclear priorities for energy supply protection for high-risk buildings. For example, a certain building once had a serious energy consumption problem and poor stability of high-zone air-cooled units due to system segmentation. The risk of equipment aging and failure increased, and energy interruptions were likely to occur during extreme weather. Such energy interruptions could cause safety accidents or significant economic losses in high-risk buildings. At the same time, traditional systems rely on manual meter reading to calculate energy consumption, and energy consumption "comes and goes without a trace". This makes it impossible to discover potential risks in a timely manner, further exacerbating energy supply safety hazards.

[0005] Thirdly, the total supply of the energy system has a clear upper limit, while the traditional energy management system is mostly based on historical consumption proportion or fixed quota, lacking dynamic optimization mechanism. On the one hand, the existing system lacks full-stack intelligent management and control capability, and cannot respond to real-time demand fluctuations, which may cause a large deviation between the total energy distribution amount and the total energy demand, such as the failure of the building automation system (BAS system) and the water pump frequency conversion system of a building, which cannot realize fine regulation and control and energy consumption monitoring, resulting in low energy utilization efficiency. On the other hand, the system does not take the risk level into account in the distribution weight, which makes it difficult to prioritize key buildings in energy shortage, resulting in the inability of energy resources to tilt towards core buildings with high demand and high risk. In addition, there are problems of technical fragmentation and lack of collaboration in the current building energy field. Renewable energy is difficult to efficiently integrate into the building energy system due to output fluctuations, further reducing energy utilization efficiency and weakening the system's safety redundancy.

[0006] In summary, for the traditional energy management system in the prior art, there are problems such as rigid energy supply strategy, lack of risk protection, and insufficient dynamic response, which result in unclear energy supply priority for high-risk buildings, coexistence of energy waste and supply shortage, and insufficient system safety redundancy. The present applicant makes corresponding exploration to solve the above problems. SUMMARY

[0007] The present application aims to solve the above problems and provide a smart energy management system and method based on big data analysis.

[0008] To achieve the various purposes of the present application, the present application adopts the following technical solutions: A smart energy management system based on big data analysis is proposed to achieve one of the purposes of the present application, comprising: A data set acquisition module is used to acquire a first sample data set and a second sample data set corresponding to a target area, wherein the target area includes a plurality of to-be-managed building sites, the first sample data set represents time series data constructed by a plurality of first sample data points, each first sample data point represents historical energy demand in a preset time step and main influencing factors within a historical time period; the second sample data set includes a plurality of second sample data points, each second sample data point contains each historical energy risk site where an energy risk event occurs and the energy risk frequency of the energy risk event in the historical energy risk site; An energy demand prediction module is used to input the historical energy demand of the to-be-managed building site into an energy demand prediction model trained by the first sample data set to determine the energy demand prediction of the to-be-managed building site in a plurality of future time steps; an energy risk prediction module configured to invoke a preset energy risk assessment model to determine an energy risk value of the building site to be managed according to the historical energy risk sites and the energy risk frequency; an energy distribution control module configured to construct an energy distribution control model corresponding to the target region according to the energy demand prediction and the energy risk value of each building site to be managed in the target region, and solve the energy distribution control model by using a preset multi-objective optimization algorithm to determine an energy distribution amount of each building site to be managed in the target region; an energy management module configured to deliver the energy distribution amount to the corresponding building site to be managed to perform energy management of each building site to be managed in the target region.

[0009] Optionally, the energy risk prediction module comprises a risk model construction unit, wherein the risk model construction unit is configured to obtain the building site to be managed, each historical energy risk site where an energy risk event occurs, and an energy risk frequency of the energy risk event in the historical energy risk site; and construct an energy risk assessment model according to the coordinates of the building site to be managed, the coordinates of the historical energy risk site, and the energy risk frequency based on a preset Gaussian kernel density estimation function.

[0010] Optionally, the energy distribution control module comprises a control model construction unit, wherein the control model construction unit is configured to obtain the energy demand prediction and the energy risk value of each building site to be managed in the target region; calculate a first product and sum of a first weight coefficient, the energy risk value of each building site to be managed, and an energy gap amount of each building site to be managed to construct a risk-oriented term; calculate a second product of a second weight coefficient and a square of a relative deviation between a total energy distribution amount and a total energy demand prediction to construct a supply-demand balance term; and construct a target function of the energy distribution control model according to a first sum between the risk-oriented term and the supply-demand balance term.

[0011] Optionally, the first weight coefficient is a weight coefficient of the risk-oriented term, and is used to prioritize the energy demand of a high-risk building site in energy distribution, and the energy gap amount is constructed by a maximum function according to a first difference between the energy demand prediction and the energy distribution amount of each building site to be managed. The second weight coefficient represents an importance of the supply-demand balance term in the energy distribution control model, and the relative deviation represents a second difference between the total energy distribution amount and the total energy demand prediction and a first ratio between the total energy demand prediction.

[0012] Optionally, the energy allocation control module further includes an energy allocation solution unit, wherein the energy allocation solution unit is used to initialize each particle in the particle swarm according to a preset multi-objective optimization algorithm, determine the global optimal solution in the particle swarm according to the boundary conditions and target output conditions of the energy allocation control model, determine the optimal energy allocation combination corresponding to the energy allocation control model according to the global optimal solution, and determine the energy allocation corresponding to each managed building location in the target area according to the optimal energy allocation combination.

[0013] Optionally, the target output condition is to minimize the objective function of the energy allocation control model, and the boundary conditions are that the energy allocation of each building location to be managed is not zero, the energy allocation of each building location to be managed is less than or equal to its corresponding maximum energy allocation, and the total energy allocation is less than or equal to the total energy supply of the energy system. The particle population comprises multiple particles, each representing a set of energy allocation combinations constructed from the energy allocations of each managed building location in the target area.

[0014] Optionally, the dataset acquisition module includes a data preprocessing unit, wherein the data preprocessing unit is used to call a preset K-nearest neighbor interpolation method to identify K sample data points in the first sample dataset that are spatially similar to the sample data points containing missing data points; extract variable values ​​from the K sample data points that belong to the same variable type as the missing data points; calculate the average value of the variable values ​​of the same type in the K sample data points; use the average value as the missing value corresponding to the missing data point to fill the missing data points in the first sample dataset; repeat the process until all missing data points in the first sample dataset are filled to determine a first sample dataset without missing data points.

[0015] Optionally, the main influencing factors include indoor and outdoor temperature, population density, equipment operating status, rainfall, wind speed, humidity, or holiday information.

[0016] Optionally, the target area includes industrial parks, commercial complexes, or urban functional areas; the energy risk events include power outages, heating interruptions, or abnormal energy quality events; the basic network architecture of the energy demand forecasting model includes a long short-term memory network model; the multi-objective optimization algorithm includes a particle swarm optimization algorithm or a hybrid genetic particle swarm optimization algorithm; and the energy risk assessment model is constructed based on a Gaussian kernel density estimation function.

[0017] A smart energy management method based on big data analytics, proposed to suit another purpose of this application, is applied to a smart energy management system based on big data analytics as described in any of the above claims, the method comprising: Obtain a first sample dataset and a second sample dataset corresponding to the target area. The target area includes multiple building locations to be managed. The first sample dataset represents time series data constructed from multiple first sample data points. Each first sample data point represents the historical energy demand and main influencing factors within a preset time step in a historical time period. The second sample dataset includes multiple second sample data points. Each second sample data point contains various historical energy risk locations where energy risk events have occurred and the energy risk frequency of the energy risk events in those historical energy risk locations. The historical energy demand of the building site to be managed is input into the energy demand prediction model trained using the first sample dataset to determine the predicted energy demand of the building site to be managed in the future at multiple time steps. The preset energy risk assessment model is invoked to determine the energy risk value of the building site to be managed based on the historical energy risk locations and the energy risk frequency; Based on the predicted energy demand and energy risk value of each building location to be managed in the target area, an energy allocation control model corresponding to the target area is constructed. The energy allocation control model is solved by a preset multi-objective optimization algorithm to determine the energy allocation corresponding to each building location to be managed in the target area. The energy allocation is delivered to its corresponding managed building location to perform energy management for each managed building location in the target area.

[0018] Compared to existing technologies, this application addresses the problems of rigid energy supply strategies, lack of risk protection, and insufficient dynamic response in traditional energy management systems, which lead to unclear energy supply priorities for high-risk buildings, coexistence of energy waste and supply shortages, and insufficient system safety redundancy. The benefits include, but are not limited to, the following: Firstly, this application utilizes a first sample dataset constructed from historical energy demand and key influencing factors such as outdoor temperature, population density, equipment operating status, rainfall, wind speed, humidity, and holiday information. This dataset provides a multi-dimensional training foundation for energy demand forecasting models, effectively addressing the curse of dimensionality problem in traditional models when dealing with multi-factor coupled data. The Long Short-Term Memory (LSTM) network trained on this first sample dataset possesses powerful temporal feature extraction capabilities. It can automatically learn the nonlinear correlation between key influencing factors such as outdoor temperature, population density, equipment operating status, rainfall, wind speed, humidity, and holiday information and energy demand, significantly improving the accuracy of future multi-time-step demand forecasts. Accurate energy demand forecasts replace traditional empirical values ​​as the allocation benchmark, preventing supply shortages during high demand periods and energy waste during low demand periods, thus laying the foundation for efficient energy utilization.

[0019] Secondly, addressing the issue of traditional energy management systems' one-size-fits-all approach leading to a lack of protection for high-risk buildings, this application integrates historical energy risk locations and frequencies using a second sample dataset. Based on Gaussian kernel density estimation, it spatializes and quantifies the energy risk of buildings under management, accurately calculating the energy risk value of each building and clearly identifying the high-risk attributes of critical buildings, thus overcoming the ambiguity of risk priority in traditional management. During energy allocation, high-risk values ​​are directly converted into protection weights, ensuring that critical buildings receive priority energy supply in any scenario, thereby mitigating safety incidents and economic losses caused by energy outages through a mechanism.

[0020] Thirdly, an objective function is constructed based on the predicted energy demand and energy risk value of each building location to be managed. This is combined with a multi-objective optimization algorithm to achieve dynamic allocation, solving the problem that traditional energy management systems often rely on historical consumption ratios or fixed quotas and lack dynamic optimization mechanisms. The multi-objective optimization algorithm takes minimizing the objective function of the energy allocation control model as its output condition. On the one hand, it can respond to energy demand fluctuations in real time, avoiding excessive deviation between total energy allocation and total energy demand. On the other hand, by deeply integrating energy risk values ​​into the optimization process through weight design, it ensures that resources are tilted towards high-demand, high-risk core buildings during energy shortages. This optimization mechanism achieves a balance between safety priority and optimal efficiency under the constraint of the total energy supply ceiling, improving energy utilization efficiency and strengthening the safety redundancy capability of the energy management system.

[0021] Furthermore, this application effectively addresses the shortcomings of traditional energy management systems, which often employ a one-size-fits-all equalization energy supply strategy. This leads to unclear priority in energy supply for high-risk buildings, potential safety accidents caused by energy outages, and an inability to respond to real-time demand fluctuations or prioritize critical buildings during energy shortages. It significantly improves the safety, rationality, and efficiency of energy management in industrial parks, commercial complexes, and urban functional areas, providing a solid foundation for the realization of refined energy management in the building sector. Attached Figure Description

[0022] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a structural diagram of the smart energy management system based on big data analysis in the embodiments of this application; Figure 2 This is a flowchart illustrating the smart energy management method based on big data analysis in the embodiments of this application. Detailed Implementation

[0023] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0024] Unless otherwise expressly stated, the various embodiments disclosed in this application can be combined in a cross-cutting manner to flexibly construct new embodiments, as long as such combination does not depart from the inventive spirit of this application and can meet the needs of the prior art or solve a certain deficiency in the prior art. Those skilled in the art should be aware of such modifications.

[0025] Please see Figure 1 In one embodiment of the smart energy management system based on big data analytics of this application, the system includes: The dataset acquisition module 1100 is used to acquire a first sample dataset and a second sample dataset corresponding to a target area. The target area includes multiple building locations to be managed. The first sample dataset represents time-series data constructed from multiple first sample data points, each representing historical energy demand and key influencing factors within a preset time step over a historical period. The second sample dataset includes multiple second sample data points, each containing historical energy risk locations where energy risk events have occurred, and the energy risk frequency of the energy risk events at those locations. The key influencing factors include indoor and outdoor temperature, population density, equipment operating status, rainfall, wind speed, humidity, and holiday information, etc. The target area includes industrial parks, commercial complexes, or urban functional areas.

[0026] Specifically, the target area represents a geographical or functional area that requires unified energy scheduling and management, including urban industrial parks or urban commercial complexes, etc. The buildings to be managed include hospitals, production plants, and high-rise buildings, etc. This application takes the smart energy management system of an industrial park in a certain city as an example, which does not constitute any limitation on this application.

[0027] The buildings to be managed represent independent buildings or functional units within the target area that require separate energy allocation and monitoring of energy demand and risks. These include production plants, office buildings, or industrial park wastewater treatment plants, etc. The target area is broken down into the smallest quantifiable and manageable units to ensure that energy allocation is accurate to individual buildings and avoid a one-size-fits-all, extensive allocation approach.

[0028] The first sample dataset represents a historical time-series data set used to train the energy demand forecasting model. It contains time-series data constructed from multiple first sample data points. Each first sample data point represents the historical energy demand and key influencing factors within a preset time step over a historical period. The historical time period can be two years, three years, or five years, etc., and the preset time step represents the time interval between the first sample data points, which can be 30 minutes or one hour, etc. The historical energy demand represents the actual energy consumed by the managed building location at a certain historical time step, such as electricity, heat, gas, etc. The key influencing factors represent the critical factors affecting building energy demand and are input features of the energy demand forecasting model. Key influencing factors strongly correlated with energy demand need to be selected to improve forecasting accuracy. The second sample dataset is a historical risk data set used to assess the energy risk of each managed building location. The second sample dataset can be collected from fault records, maintenance logs, accident reports, etc., of the park's energy management system, such as Building 1 in 2023. In June of this year, there were three power outages due to unstable voltage; in March 2024, the heating pipe of Office Building No. 3 ruptured once; and so on. Energy risk values ​​for each building are calculated using historical risk data. For example, the higher the risk frequency, the greater the energy risk value, providing a risk-priority decision-making basis for subsequent energy allocation. For instance, the high-risk Factory Building No. 1 needs priority energy allocation to avoid energy shortages leading to malfunctions. The energy risk events represent events that cause interruptions or abnormalities in building energy use due to insufficient energy supply, equipment failure, or external emergencies. These include power outages, heating interruptions, and abnormal energy quality events. Power outages refer to power outages caused by grid failures or line maintenance; heating interruptions refer to power outages caused by heating pipe ruptures or insufficient gas pressure; and abnormal energy quality events refer to situations affecting the normal use of the target building, such as unstable voltage or substandard heating temperatures. The energy risk frequency represents the unit time frequency of energy risk events occurring at a managed building location within a historical time period, and is a core indicator for quantifying the building's energy risk level. For example, if Factory Building No. 1 experienced nine power outages in March, the risk frequency is three times per month.

[0029] In some embodiments, the dataset acquisition module includes a data preprocessing unit, wherein the data preprocessing unit is used to invoke a preset K-nearest neighbor interpolation method to identify K sample data points in the first sample dataset that are spatially similar to the sample data points containing missing data points; extract variable values ​​from the K sample data points that belong to the same variable type as the missing data points; calculate the average value of the variable values ​​of the same type among the K sample data points; use the average value as the missing value corresponding to the missing data point to fill the missing data points in the first sample dataset; repeat the process until all missing data points in the first sample dataset are filled to determine a first sample dataset without missing data points. Specifically, based on the spatial similarity screening logic of the K-nearest neighbor interpolation method, it ensures that the K sample data points used for filling have a strong correlation with the missing data points, and the average value of the variable values ​​of the same type can fit the data distribution pattern. The filling accuracy is higher than that of random filling, single mean filling, and other methods, reducing data distortion. By iteratively filling all missing data points, a first sample dataset without missing data is formed, avoiding insufficient training sample size or abnormal input features of the energy demand prediction model due to data missing data, and ensuring that the model can capture complete temporal dependencies and correlation patterns of influencing factors. By selecting spatially similar samples and focusing on the values ​​of variables of the same type to calculate the mean, the interference of different variables and irrelevant samples on the imputation results is avoided, reducing the error caused by outlier imputation. This provides high-quality data support for subsequent model training and indirectly improves the accuracy of energy demand forecasts. The temporal continuity of energy data is strongly correlated with influencing factors. This preprocessing method is well-suited to the characteristics of the data, avoiding temporal breaks caused by missing values, ensuring data continuity and consistency, and providing a reliable data foundation for subsequent energy allocation optimization decisions.

[0030] In some embodiments, the target area includes an industrial park, a commercial complex, or an urban functional area; the energy risk events include power outages, heating interruptions, or abnormal energy quality events; the basic network architecture of the energy demand forecasting model includes a long short-term memory network model; the multi-objective optimization algorithm includes a particle swarm optimization algorithm or a hybrid genetic particle swarm optimization algorithm; and the energy risk assessment model is constructed based on a Gaussian kernel density estimation function.

[0031] The energy demand forecasting module 1200 is used to input the historical energy demand of the building site to be managed into the energy demand forecasting model trained using the first sample dataset, so as to determine the predicted energy demand of the building site in the future at multiple time steps; wherein, the target area includes industrial parks, commercial complexes or urban functional areas; the energy risk events include power outage events, heating interruption events or energy quality anomaly events; the basic network architecture of the energy demand forecasting model includes long short-term memory network models, etc.

[0032] After obtaining the first sample dataset and the second sample dataset corresponding to the target area, the historical energy demand of the building site to be managed is input into the energy demand prediction model trained using the first sample dataset in the energy demand prediction module to determine the predicted energy demand of the building site to be managed in multiple future time steps; wherein, the target area includes industrial parks, commercial complexes or urban functional areas; the energy risk events include power outage events, heating interruption events or energy quality anomaly events; the basic network architecture of the energy demand prediction model includes long short-term memory network models, etc.

[0033] Specifically, energy demand exhibits strong temporal correlations, such as daily cycles, seasonal cycles, and holiday patterns. Traditional recurrent neural networks (RNNs) are prone to gradient vanishing and cannot learn long-distance dependencies. However, long short-term memory (LSTM) models, through gating mechanisms including forget gates, input gates, and output gates, can effectively memorize key historical information, such as the heating demand pattern of the previous winter, thus adapting to the long-term characteristics of energy data. Therefore, the LSTM model is used as the energy demand prediction model in this application.

[0034] The historical energy demand of the building site to be managed must be completely matched with the input format when training the first sample dataset. That is, it should be organized into a continuous time series according to the preset time step, such as the historical energy demand of the past 72 hours and the time series data of the corresponding temperature, holiday information and other major influencing factors. The sequence length and feature dimensions should be consistent with those during model training. The input data should be preprocessed, such as missing value imputation and outlier correction, to avoid abnormal model predictions due to data defects and to ensure that the input sequence can reflect the real time series characteristics of building energy demand, such as fluctuations in weekdays, holidays and seasonal changes.

[0035] The Long Short-Term Memory (LSTM) network model trained on the first sample dataset is invoked. This model has learned to predict the energy demand of the managed building site at multiple future time steps based on the historical energy demand of the managed building site.

[0036] In the forget gate of the Long Short-Term Memory (LSTM) network model, key information from historical data, such as long-term seasonal trends and peak demand at fixed time periods, is filtered out, while irrelevant noise, such as short-term fluctuations caused by temporary equipment start-ups and shutdowns, is discarded. In the input gate of the LSTM network model, the current input historical energy demand is integrated with influencing factors such as temperature and holidays to update the model's memory state and strengthen effective features, such as the correlation between high temperature and cooling demand, and the correspondence between holidays and low load. In the output gate of the LSTM network model, based on the updated memory state, the energy demand forecast for multiple future time steps is mapped through a fully connected layer, realizing a sequence-to-sequence prediction transformation.

[0037] The energy demand forecasting model outputs the predicted energy demand for the building site to be managed at multiple time steps in the future, such as the hourly electricity consumption in the next 24 hours and the daily heating demand in the next 7 days. This predicted energy demand can be directly used as the core input of the subsequent energy allocation control model, providing an accurate demand benchmark for risk-prioritized and supply-demand balanced energy allocation optimization.

[0038] The energy risk prediction module 1300 is used to call a preset energy risk assessment model to determine the energy risk value of the building location to be managed based on the historical energy risk locations and the energy risk frequency; wherein, the energy risk assessment model is constructed based on the Gaussian kernel density estimation function.

[0039] The historical energy demand of the building site to be managed is input into the energy demand prediction model trained using the first sample dataset to determine the predicted energy demand of the building site to be managed in the future multiple time steps. Then, in the energy risk prediction module, a preset energy risk assessment model is called to determine the energy risk value of the building site to be managed based on the historical energy risk location and the energy risk frequency. The energy risk assessment model is constructed based on the Gaussian kernel density estimation function.

[0040] Based on the Gaussian Kernel Density Estimation (KDE) function, and combining the spatial relationship between the managed building location and historical energy risk locations, as well as the frequency of historical energy risks, the energy risk value of each managed building location is calculated. According to the fusion model of spatial correlation and historical risk intensity, it adapts to the clustering effect of energy risks in geographic space. For example, the aging of the pipeline network in a certain area may lead to multiple buildings experiencing consecutive risk events. Compared with the calculation method based solely on its own historical risk frequency, it is more generalizable and accurate.

[0041] In some embodiments, the energy risk prediction module includes a risk model construction unit, wherein the risk model construction unit is used to obtain the location of the building to be managed, each historical energy risk location where an energy risk event has occurred, and the energy risk frequency of the energy risk event in the historical energy risk location; and to construct an energy risk assessment model based on a preset Gaussian kernel density estimation function according to the coordinates of the building to be managed, the coordinates of the historical energy risk locations, and the energy risk frequency.

[0042] In a specific embodiment, the following steps are taken: First, the location of the building to be managed, all historical energy risk locations where energy risk events have occurred, and the energy risk frequency of the energy risk events at those historical energy risk locations are obtained. Second, based on a preset Gaussian kernel density estimation function, the square of the Euclidean distance between the coordinates of the building to be managed and the coordinates of each historical energy risk location is calculated. Third, the ratio between the square of the Euclidean distance between the coordinates of the building to be managed and the coordinates of each historical energy risk location and twice the square of the bandwidth parameter is calculated. Fourth, the natural exponential function value of this ratio is calculated to quantify the degree of spatial attenuation of the historical energy risk locations' influence on the building to be managed. Fifth, the product between the natural exponential function value corresponding to each historical energy risk location and the energy risk frequency corresponding to that historical energy risk location is calculated to determine the risk contribution value of a single historical energy risk location to the building to be managed. Finally, the risk contribution values ​​corresponding to all historical energy risk locations are summed to construct an energy risk assessment model. The bandwidth parameter, which characterizes the control factor for the spatial impact range of historical energy risk events in Gaussian kernel density estimation, can be determined as needed by those skilled in the art based on actual application scenarios. No limitations are imposed here. For example, if the energy risk is a localized event (such as equipment failure inside a building or leakage in a small-scale heating network), the impact radius is typically between 100 and 300 meters, and the bandwidth parameter is between 100 and 300 meters to ensure that the risk impact is focused on nearby buildings and avoids interference from unrelated areas. If the energy risk is a regional event (such as fluctuations in the regional power grid or failures in the main pipeline network), the impact radius can reach 500 to 1000 meters, and the bandwidth parameter is between 500 and 1000 meters to cover the entire range of risk diffusion.

[0043] As can be seen from the above embodiments, based on the constructed energy risk assessment model, the energy risk value of the building to be managed can be determined according to the coordinates of the historical energy risk locations and their corresponding energy risk frequencies. Spatial correlation is constructed based on the square of the Euclidean distance between the coordinates of the building to be managed and the coordinates of each historical energy risk location. Combined with the natural exponential decay characteristic of the Gaussian kernel function, the spatial impact of historical energy risk locations on the target building is accurately quantified. The closer the distance, the stronger the impact. The energy risk frequency is superimposed to avoid the bias of single-dimensional assessment and conform to the spatial propagation law of energy risk.

[0044] By summing and integrating the contribution values ​​of all historical energy risk points, this approach covers both the strong impact of high-frequency risk points and the cumulative effect of low-frequency risk points. It also addresses the pain point of new buildings lacking their own risk data, allowing for the inference of their risk level through surrounding historical risks. The output energy risk value can be directly used as a priority for energy allocation, providing data support for optimizing energy supply to high-risk buildings and helping to improve the safety and rationality of energy allocation.

[0045] The energy allocation control module 1400 is used to construct an energy allocation control model corresponding to the target area based on the energy demand forecast and energy risk value corresponding to each building location to be managed in the target area, and to solve the energy allocation control model using a preset multi-objective optimization algorithm to determine the energy allocation corresponding to each building location to be managed in the target area. After determining the energy risk value of the building location to be managed based on the historical energy risk location and the energy risk frequency by calling the preset energy risk assessment model, the energy allocation control module 1400 constructs an energy allocation control model corresponding to the target area based on the predicted energy demand and energy risk value of each building location to be managed in the target area. The preset multi-objective optimization algorithm is used to solve the energy allocation control model to determine the energy allocation amount corresponding to each building location to be managed in the target area.

[0046] In some embodiments, the energy allocation control module includes a control model construction unit, wherein the control model construction unit is used to obtain the energy demand forecast and energy risk value corresponding to each managed building location in the target area; calculate and sum the first product between a first weighting coefficient, the energy risk value corresponding to each managed building location, and the energy deficit corresponding to each managed building location to construct a risk-oriented term; calculate and sum the square of the relative deviation between the total energy allocation and the total energy demand forecast, and the second product between the second weighting coefficients to construct a supply-demand balance term; and construct the objective function of the energy allocation control model based on the first sum between the risk-oriented term and the supply-demand balance term. The first weighting coefficient is the weighting coefficient of the risk-oriented term, used to prioritize the energy demand of high-risk building locations in energy allocation; the energy deficit is constructed by a maximum value function based on the first difference between the energy demand forecast and the energy allocation corresponding to each managed building location; the second weighting coefficient characterizes the importance of the supply-demand balance term in the energy allocation control model; and the relative deviation characterizes the first ratio between the second difference between the total energy allocation and the total energy demand forecast and the total energy demand forecast. The first weighting coefficient and the second weighting coefficient can be determined by those skilled in the art as needed according to the actual application scenario, and no limitation is made here.

[0047] Specifically, the total energy allocation represents the sum of energy allocations for all managed building locations within the target area; the total energy demand forecast represents the sum of predicted energy demand for all managed building locations within the target area; based on the maximum value function, an energy gap is constructed according to the difference between the predicted energy demand and the energy allocation for each managed building location; that is, when the predicted energy demand is greater than the energy allocation, the difference between the predicted energy demand and the energy allocation for each managed building location is taken as the value of the energy gap; when the predicted energy demand is less than or equal to the energy allocation, the value of the energy gap is 0.

[0048] First, we obtain the predicted energy demand and energy risk values ​​for each building location to be managed within the target area. This provides a dual-dimensional quantitative basis for constructing the objective function of the energy allocation control model, using both energy demand benchmarks and risk priorities. The predicted energy demand clearly defines the energy consumption scale of each building, preventing allocation schemes from deviating from actual energy needs. The energy risk values ​​differentiate the safety assurance levels of buildings, addressing the fairness issue of a one-size-fits-all approach in traditional allocation. The combination of these two factors ensures that subsequent optimization aligns with energy consumption patterns while focusing on safety priorities, injecting engineering practical value into the objective function and ensuring that the optimization direction does not deviate from the core objectives of on-demand energy allocation and risk control.

[0049] Furthermore, the first product of the determined first weighting coefficient, the energy risk value corresponding to each managed building location, and the energy deficit corresponding to each managed building location is calculated and summed to construct a risk-oriented term. This risk-oriented term quantifies the energy loss from high-risk buildings, driving a multi-objective optimization algorithm to prioritize the energy needs of high-risk, high-deficit buildings, thus preventing safety accidents or significant losses due to energy shortages. The energy deficit only takes effect when the allocated amount is insufficient to meet demand, avoiding ineffective penalties for scenarios of energy surplus and aligning with actual energy consumption logic. Specifically, when the predicted energy demand is greater than the allocated energy, the difference between the predicted energy demand and the allocated energy for each managed building location is taken as the energy deficit value; when the predicted energy demand is less than or equal to the allocated energy, the energy deficit value is 0. Furthermore, the square of the relative deviation between the total energy allocation and the total energy demand forecast is calculated, and the second product between the second weighting coefficient is used to construct a supply and demand balance term. The supply and demand balance term is used to quantify the degree of deviation between the total energy allocation and the total demand forecast, to penalize energy surplus or shortage, and to avoid global resource waste or shortage caused by local priority.

[0050] Furthermore, the objective function of the energy allocation control model is constructed based on the first sum between the risk-oriented term and the supply-demand balance term. The risk-oriented term and the supply-demand balance term together constitute the dual core of the objective function of the energy allocation control model. A dynamic trade-off between safety and efficiency is achieved through weighted summation. Focusing solely on the risk-oriented term can easily lead to oversupply or severe energy shortages in low-risk buildings; focusing solely on the supply-demand balance term can easily neglect the safety needs of high-risk buildings, potentially causing safety accidents. When combined, the multi-objective optimization algorithm must find the optimal solution between prioritizing high-risk buildings and controlling the overall supply-demand deviation, ensuring that the final allocation scheme is both safe and reasonable.

[0051] In some embodiments, the energy allocation control module further includes an energy allocation solution unit, wherein the energy allocation solution unit is used to initialize each particle in the particle swarm according to a preset multi-objective optimization algorithm, determine the global optimal solution in the particle swarm according to the boundary conditions and target output conditions of the energy allocation control model, determine the optimal energy allocation combination corresponding to the energy allocation control model according to the global optimal solution, and determine the energy allocation corresponding to each managed building location in the target area according to the optimal energy allocation combination. The multi-objective optimization algorithm includes a particle swarm optimization algorithm or a genetic particle swarm optimization algorithm; the energy risk assessment model is constructed based on a Gaussian kernel density estimation function; the target output condition is to minimize the objective function of the energy allocation control model; the boundary conditions are that the energy allocation of each managed building location is not zero, the energy allocation of each managed building location is less than or equal to its corresponding maximum energy allocation, and the total energy allocation is less than or equal to the total energy supply of the energy system; the particle swarm includes multiple particles, each particle representing a set of energy allocation combinations, the energy allocation combinations being constructed from the energy allocation of each managed building location in the target area.

[0052] As can be seen from the above embodiments, the energy allocation for each managed building location is not zero to ensure the energy consumption of the building foundation. The energy allocation for each managed building location is less than or equal to its corresponding maximum energy allocation to match the pipeline transmission capacity. The total energy allocation is less than or equal to the total energy supply of the energy system to avoid energy system overload. This constraint restricts the particle iteration process, preventing the optimized energy allocation combination from being theoretically feasible but practically unenforceable.

[0053] Particle swarm optimization (PSO) aims to minimize the objective function of the energy allocation control model. While prioritizing energy supply to high-risk buildings such as hospitals and core factories, it controls overall allocation deviation through supply-demand balance constraints. For example, in industrial park scheduling, it avoids production line shutdowns due to energy shortages while preventing steam waste caused by excessive energy supply, significantly improving energy supply security while greatly reducing energy waste. Particle swarm optimization (PSO) algorithm rapidly searches for the global optimum through population iteration. Compared to energy allocation based on experience, it can output the optimal energy allocation combination within minutes. When faced with fluctuations in energy demand (such as peak electricity consumption in commercial areas in the evening), it can quickly recalculate and provide real-time decision-making basis for dispatchers, reducing the risk of response lag.

[0054] Each particle corresponds to a set of energy allocation combinations, and the global optimal solution is directly transformed into the energy allocation of each building within the target area, such as 300kWh for workshop A and 80kWh for office building B. This transforms energy dispatch from fuzzy estimation to precise metering, meeting the refined energy management needs of enterprises.

[0055] The energy management module 1500 is used to deliver the energy allocation to its corresponding managed building location in order to perform energy management for each managed building location in the target area.

[0056] In the energy distribution control module 1400, an energy distribution control model corresponding to the target area is constructed based on the predicted energy demand and energy risk value of each building location to be managed in the target area. The energy distribution control model is solved using a preset multi-objective optimization algorithm to determine the energy distribution amount corresponding to each building location to be managed in the target area. In the energy management module 1500, the energy distribution amount is delivered to its corresponding building location to be managed in order to perform energy management of each building location to be managed in the target area.

[0057] Specifically, the energy allocation amount of each building location to be managed in the target area, obtained by solving the energy allocation control module 1400, is input to the energy management module. The energy management module then executes specific energy delivery instructions. Based on the allocation amount, the energy management module controls the energy supply to each building location to be managed, such as adjusting the power supply circuit power and the flow rate of the heating pipes, to ensure that the actual energy supply is consistent with the energy allocation amount. This enables the buildings in the target area to receive energy on demand and according to risk priority, ensuring the safe energy use of high-risk buildings while avoiding energy waste or insufficient supply.

[0058] In some embodiments, the smart energy management system based on big data analysis further includes a carbon management module, wherein the carbon management module includes a carbon data acquisition unit, a carbon accounting analysis unit, a carbon emission reduction optimization unit, and a carbon monitoring visualization unit; the carbon data acquisition unit uses multi-dimensional sensors to collect energy consumption data corresponding to various equipment systems in the building site to be managed, wherein the equipment systems include air conditioning systems, lighting systems, and power distribution systems, etc.; the multi-dimensional sensors include a first sensor, a second sensor, and a third sensor, wherein the first sensor is used to collect power consumption of the chilled source unit of the building air conditioning system, chilled water supply and return water temperature difference and flow data, etc.; the second sensor is used to collect power consumption of LED light sources of the lighting system, and start-stop duration data of underground garage sensor lighting, etc.; the third sensor is used to collect total power consumption of the power distribution system and renewable energy power generation data, wherein renewable energy includes photovoltaic, hydrogen energy, etc.; the carbon accounting analysis unit calculates the carbon emissions of each equipment system based on the energy consumption data corresponding to each equipment system in the building site to be managed, combined with the carbon emission factors of different energy types. The system compares the total carbon emissions of the building with the preset carbon reduction target value to generate a carbon emission deviation analysis report. The energy types include grid electricity and renewable energy. The carbon reduction optimization unit, based on a preset AI scheduling algorithm and combined with the deviation analysis report output by the carbon accounting analysis unit, as well as building cooling load change data and lighting system usage data, dynamically adjusts the operating parameters of the air conditioning system's chiller, water pump, and cooling tower, and the light source start-stop strategy of the lighting system to optimize carbon emissions in real time. The building cooling load change data includes summer outdoor temperature and humidity and the number of chiller units operating during the transition season. The lighting system usage data includes the density of people in public areas and the frequency of vehicle entry and exit in underground parking garages. The carbon monitoring visualization unit is connected to the smart energy management platform and displays the carbon emission percentage by building subsystem, floor, or preset time period in the carbon emission visualization interface. It also displays the carbon emission reduction trend after the implementation of the carbon reduction optimization unit and supports automatic alarm triggering when carbon emissions exceed the standard. The preset time period includes daily, weekly, or monthly.

[0059] As can be seen from the above embodiments, in view of the problems of rigid energy supply strategies, lack of risk protection, and insufficient dynamic response in the traditional energy management system of the prior art, which leads to unclear energy supply priorities for high-risk buildings, coexistence of energy waste and supply shortage, and insufficient system safety redundancy, this application has, but is not limited to, the following beneficial effects: Firstly, this application utilizes a first sample dataset constructed from historical energy demand and key influencing factors such as outdoor temperature, population density, equipment operating status, rainfall, wind speed, humidity, and holiday information. This dataset provides a multi-dimensional training foundation for energy demand forecasting models, effectively addressing the curse of dimensionality problem in traditional models when dealing with multi-factor coupled data. The Long Short-Term Memory (LSTM) network trained on this first sample dataset possesses powerful temporal feature extraction capabilities. It can automatically learn the nonlinear correlation between key influencing factors such as outdoor temperature, population density, equipment operating status, rainfall, wind speed, humidity, and holiday information and energy demand, significantly improving the accuracy of future multi-time-step demand forecasts. Accurate energy demand forecasts replace traditional empirical values ​​as the allocation benchmark, preventing supply shortages during high demand periods and energy waste during low demand periods, thus laying the foundation for efficient energy utilization.

[0060] Secondly, addressing the issue of traditional energy management systems' one-size-fits-all approach leading to a lack of protection for high-risk buildings, this application integrates historical energy risk locations and frequencies using a second sample dataset. Based on Gaussian kernel density estimation, it spatializes and quantifies the energy risk of buildings under management, accurately calculating the energy risk value of each building and clearly identifying the high-risk attributes of critical buildings, thus overcoming the ambiguity of risk priority in traditional management. During energy allocation, high-risk values ​​are directly converted into protection weights, ensuring that critical buildings receive priority energy supply in any scenario, thereby mitigating safety incidents and economic losses caused by energy outages through a mechanism.

[0061] Thirdly, an objective function is constructed based on the predicted energy demand and energy risk value of each building location to be managed. This is combined with a multi-objective optimization algorithm to achieve dynamic allocation, solving the problem that traditional energy management systems often rely on historical consumption ratios or fixed quotas and lack dynamic optimization mechanisms. The multi-objective optimization algorithm takes minimizing the objective function of the energy allocation control model as its output condition. On the one hand, it can respond to energy demand fluctuations in real time, avoiding excessive deviation between total energy allocation and total energy demand. On the other hand, by deeply integrating energy risk values ​​into the optimization process through weight design, it ensures that resources are tilted towards high-demand, high-risk core buildings during energy shortages. This optimization mechanism achieves a balance between safety priority and optimal efficiency under the constraint of the total energy supply ceiling, improving energy utilization efficiency and strengthening the safety redundancy capability of the energy management system.

[0062] Furthermore, this application effectively addresses the shortcomings of traditional energy management systems, which often employ a one-size-fits-all equalization energy supply strategy. This leads to unclear priority in energy supply for high-risk buildings, potential safety accidents caused by energy outages, and an inability to respond to real-time demand fluctuations or prioritize critical buildings during energy shortages. It significantly improves the safety, rationality, and efficiency of energy management in industrial parks, commercial complexes, and urban functional areas, providing a solid foundation for the realization of refined energy management in the building sector.

[0063] Please see Figure 2 A smart energy management method based on big data analysis, proposed to suit another purpose of this application, is applied to a smart energy management system based on big data analysis as described in any of the above claims, the method comprising: Step S10: Obtain a first sample dataset and a second sample dataset corresponding to the target area. The target area includes multiple building locations to be managed. The first sample dataset represents time series data constructed from multiple first sample data points. Each first sample data point represents the historical energy demand and main influencing factors within a preset time step in a historical time period. The second sample dataset includes multiple second sample data points. Each second sample data point contains various historical energy risk locations where energy risk events have occurred and the energy risk frequency of the energy risk events in those historical energy risk locations. Step S20: Input the historical energy demand of the building site to be managed into the energy demand prediction model trained using the first sample dataset to determine the predicted energy demand of the building site in the future at multiple time steps. Step S30: Invoke the preset energy risk assessment model to determine the energy risk value of the building location to be managed based on the historical energy risk locations and the energy risk frequency; Step S40: Based on the predicted energy demand and energy risk value of each building location to be managed in the target area, construct an energy allocation control model corresponding to the target area, and use a preset multi-objective optimization algorithm to solve the energy allocation control model to determine the energy allocation corresponding to each building location to be managed in the target area. Step S50: The energy allocation is delivered to the corresponding building location to be managed, so as to perform energy management for each building location to be managed in the target area.

[0064] The above description is only a partial embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0066] In summary, this application effectively addresses the shortcomings of traditional energy management systems, which often employ a one-size-fits-all equalization energy supply strategy. These shortcomings include unclear priority for energy supply to high-risk buildings, potential safety accidents caused by energy outages, inability to respond to real-time demand fluctuations, and difficulty in prioritizing critical buildings during energy shortages. This significantly improves the safety, rationality, and efficiency of energy management in industrial parks, commercial complexes, and urban functional areas, providing a solid foundation for the realization of refined energy management in the building sector.

Claims

1. A smart energy management system based on big data analytics, characterized in that, include: The dataset acquisition module is used to acquire a first sample dataset and a second sample dataset corresponding to a target area. The target area includes multiple building locations to be managed. The first sample dataset represents time series data constructed from multiple first sample data points. Each first sample data point represents the historical energy demand and main influencing factors within a preset time step in a historical time period. The second sample dataset includes multiple second sample data points. Each second sample data point contains various historical energy risk locations where energy risk events have occurred and the energy risk frequency of the energy risk events in those historical energy risk locations. An energy demand forecasting module is used to input the historical energy demand of the building site to be managed into an energy demand forecasting model trained using the first sample dataset, so as to determine the predicted energy demand of the building site to be managed in the future multiple time steps. The energy risk prediction module is used to call a preset energy risk assessment model to determine the energy risk value of the building site to be managed based on the historical energy risk locations and the energy risk frequency. An energy allocation control module is used to construct an energy allocation control model corresponding to the target area based on the energy demand forecast and energy risk value of each building location to be managed in the target area, and to solve the energy allocation control model using a preset multi-objective optimization algorithm to determine the energy allocation amount corresponding to each building location to be managed in the target area. An energy management module is used to deliver the energy allocation to its corresponding managed building location in order to perform energy management for each managed building location in the target area.

2. The smart energy management system based on big data analysis according to claim 1, characterized in that, The energy risk prediction module includes a risk model construction unit, which is used to obtain the location of the building to be managed, the historical energy risk locations where energy risk events have occurred, and the energy risk frequency of the energy risk events in the historical energy risk locations; and to construct an energy risk assessment model based on a preset Gaussian kernel density estimation function according to the coordinates of the building to be managed, the coordinates of the historical energy risk locations, and the energy risk frequency.

3. The smart energy management system based on big data analysis according to claim 1, characterized in that, The energy allocation control module includes a control model construction unit, which is used to obtain the predicted energy demand and energy risk value corresponding to each managed building location in the target area; calculate and sum the first product between a first weighting coefficient, the energy risk value corresponding to each managed building location, and the energy deficit corresponding to each managed building location to construct a risk-oriented term; calculate and sum the square of the relative deviation between the total energy allocation and the total energy demand prediction, and the second product between the second weighting coefficients to construct a supply-demand balance term; and construct the objective function of the energy allocation control model based on the first sum between the risk-oriented term and the supply-demand balance term.

4. The smart energy management system based on big data analysis according to claim 3, characterized in that, The first weighting coefficient is the weighting coefficient of the risk-oriented item, which is used to prioritize the energy demand of high-risk building sites in energy allocation. The energy gap is constructed by the maximum value function based on the first difference between the predicted energy demand and the energy allocation for each building site to be managed. The second weighting coefficient represents the importance of the supply and demand balance term in the energy allocation control model, and the relative deviation represents the ratio between the second difference between the total energy allocation and the total energy demand forecast and the first ratio between the total energy demand forecast.

5. The smart energy management system based on big data analysis according to claim 1, characterized in that, The energy allocation control module further includes an energy allocation solution unit, wherein the energy allocation solution unit is used to initialize each particle in the particle swarm according to a preset multi-objective optimization algorithm, determine the global optimal solution in the particle swarm according to the boundary conditions and target output conditions of the energy allocation control model, determine the optimal energy allocation combination corresponding to the energy allocation control model according to the global optimal solution, and determine the energy allocation corresponding to each building location to be managed in the target area according to the optimal energy allocation combination.

6. The smart energy management system based on big data analysis according to claim 5, characterized in that, The target output condition is to minimize the objective function of the energy allocation control model. The boundary conditions are that the energy allocation of each managed building location is not zero, the energy allocation of each managed building location is less than or equal to its corresponding maximum energy allocation, and the total energy allocation is less than or equal to the total energy supply of the energy system. The particle population comprises multiple particles, each representing a set of energy allocation combinations constructed from the energy allocations of each managed building location in the target area.

7. The smart energy management system based on big data analysis according to claim 1, characterized in that, The dataset acquisition module includes a data preprocessing unit, wherein the data preprocessing unit is used to call a preset K-nearest neighbor interpolation method to identify K sample data points in the first sample dataset that are spatially similar to the sample data points containing missing data points; extract variable values ​​from the K sample data points that belong to the same variable type as the missing data points; calculate the average value of the variable values ​​of the same type in the K sample data points; use the average value as the missing value corresponding to the missing data point to fill the missing data points in the first sample dataset; repeat the process until all missing data points in the first sample dataset are filled to determine a first sample dataset without missing data points.

8. The smart energy management system based on big data analysis according to claim 1, characterized in that, The main influencing factors include indoor and outdoor temperature, population density, equipment operating status, rainfall, wind speed, humidity, or holiday information.

9. The smart energy management system based on big data analysis according to any one of claims 1 to 8, characterized in that, The target area includes industrial parks, commercial complexes, or urban functional areas; the energy risk events include power outages, heating interruptions, or abnormal energy quality events; the basic network architecture of the energy demand forecasting model includes a long short-term memory network model; the multi-objective optimization algorithm includes particle swarm optimization or a hybrid genetic particle swarm optimization algorithm; and the energy risk assessment model is constructed based on a Gaussian kernel density estimation function.

10. A smart energy management method based on big data analytics, applied to the smart energy management system based on big data analytics as described in any one of claims 1 to 9, characterized in that, include: Obtain a first sample dataset and a second sample dataset corresponding to the target area. The target area includes multiple building locations to be managed. The first sample dataset represents time series data constructed from multiple first sample data points. Each first sample data point represents the historical energy demand and main influencing factors within a preset time step in a historical time period. The second sample dataset includes multiple second sample data points. Each second sample data point contains various historical energy risk locations where energy risk events have occurred and the energy risk frequency of the energy risk events in those historical energy risk locations. The historical energy demand of the building site to be managed is input into the energy demand prediction model trained using the first sample dataset to determine the predicted energy demand of the building site to be managed in the future at multiple time steps. The preset energy risk assessment model is invoked to determine the energy risk value of the building site to be managed based on the historical energy risk locations and the energy risk frequency; Based on the predicted energy demand and energy risk value of each building location to be managed in the target area, an energy allocation control model corresponding to the target area is constructed. The energy allocation control model is solved by a preset multi-objective optimization algorithm to determine the energy allocation corresponding to each building location to be managed in the target area. The energy allocation is delivered to its corresponding managed building location to perform energy management for each managed building location in the target area.

Images