Community energy consumption prediction method, device and equipment based on Bayesian calibration

Abstract

The present invention provides a community energy consumption prediction method, device and equipment based on Bayesian calibration, which relates to the technical field of energy consumption prediction. The method includes inputting historical energy consumption data and community energy impact data into a trained energy consumption prediction model, generating first prediction data, and determining difference data; extracting a time series histogram, determining the distribution characteristics of the impact data, and determining the prior distribution of the energy consumption prediction model; based on the Bayesian theorem, determining the likelihood value according to the difference data, and generating a posterior distribution; inputting the posterior distribution and future community energy impact data into the trained energy consumption prediction model, generating second prediction data and prediction confidence; extracting the first similarity and the target time interval; extracting historical similar energy consumption data and predicted similar energy consumption data respectively, and determining the second similarity; generating a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity. The present invention can improve the accuracy of prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of energy consumption prediction, and in particular to a method, device and equipment for predicting community energy consumption based on Bayesian calibration. Background Art

[0002] Community energy consumption forecasting is a key step in optimizing energy use, improving energy efficiency, and formulating effective energy conservation and emission reduction strategies. Community energy consumption forecasting involves data analysis and technical applications at multiple levels, aiming to accurately estimate the energy consumption of the community in the future. This not only helps to effectively allocate resources, but also promotes the use of clean energy and the construction of environmentally friendly communities.

[0003] In the existing technology, energy consumption forecasting at the community level mainly relies on statistical models and rule-based methods. Although such methods can provide forecast data, there are nonlinear and uncertain environmental factors in the forecast of community energy consumption, which makes it difficult for current technical methods to handle complex and changeable actual conditions, and the prediction accuracy is low. Summary of the invention

[0004] The problem solved by the present invention is how to improve the accuracy of energy consumption prediction.

[0005] To solve the above problems, in a first aspect, the present invention provides a community energy consumption prediction method based on Bayesian calibration, comprising:

[0006] Obtain historical energy consumption data, historical community energy impact data, and future community energy impact data;

[0007] Inputting the historical energy consumption data and the historical community energy impact data into a trained energy consumption prediction model to generate first prediction data, and determining difference data between the first prediction data and the historical energy consumption data;

[0008] Extracting time series histograms of the historical energy consumption data and the historical community energy impact data respectively, determining the impact data distribution characteristics according to the time series histograms, and determining the prior distribution of the energy consumption prediction model according to the impact data distribution characteristics and expert knowledge;

[0009] Based on the Bayesian theorem, a likelihood value is determined according to the difference data, and a posterior distribution is generated according to the prior distribution and the likelihood value;

[0010] Input the posterior distribution and the future community energy impact data in multiple different future time intervals into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals;

[0011] Extracting a first similarity between the impact data in each of the historical time intervals and the future community energy impact data in the corresponding future time interval, and extracting a target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval;

[0012] Respectively extracting historical similar energy consumption data corresponding to the target historical time interval from the historical energy consumption data, and predicted similar energy consumption data corresponding to the target future time interval from the second predicted data, and determining a second similarity between the historical similar energy consumption data and the predicted similar energy consumption data;

[0013] A final energy consumption prediction result is generated according to the second prediction data, the prediction confidence and the second similarity.

[0014] Optionally, generating a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity includes:

[0015] Extracting high-confidence prediction data whose prediction confidence is greater than a preset confidence from the second prediction data;

[0016] Associating the high-confidence prediction data with the second similarity, dividing the high-confidence prediction data according to the second similarity and a second preset similarity, and generating a determined prediction result and a pending prediction result;

[0017] Extracting the pending future time interval of the pending prediction result and the pending historical time interval corresponding to the pending future time interval, and extracting the future community energy impact data corresponding to the pending future time interval according to the pending future time interval, and extracting the historical energy consumption data corresponding to the pending historical time interval according to the pending historical time interval;

[0018] Inputting the historical energy consumption data corresponding to the undetermined historical time interval and the future community energy impact data corresponding to the undetermined future time interval into the trained energy consumption prediction model to generate a retest result;

[0019] The determined prediction result and the re-measurement result are integrated to generate the final energy consumption prediction result.

[0020] Optionally, the trained energy consumption prediction model includes: an input layer, a feature extraction module, a multimodal fusion module, an integrated learning module, an uncertainty quantification module and an output layer;

[0021] The input layer is used to obtain the historical energy consumption data and the historical community energy impact data, or to obtain the posterior distribution and the future community energy impact data;

[0022] The feature extraction module is used to extract the first long-term time feature and the first spatial feature of the historical energy consumption data and the historical community energy impact data, or to extract the second long-term time feature and the second spatial feature of the posterior distribution and the future community energy impact data, respectively;

[0023] The multimodal fusion module is used to fuse the corresponding first long-term time feature and the first spatial feature respectively to generate a first comprehensive feature, or to fuse the corresponding second long-term time feature and the second spatial feature respectively to generate a second comprehensive feature;

[0024] The integrated learning module is used to perform integrated learning on all the first comprehensive features to generate the first prediction data, or to perform integrated learning on all the second comprehensive features to generate the second prediction data;

[0025] The uncertainty quantification module is used to generate the corresponding prediction confidence according to the second comprehensive feature and the second prediction data;

[0026] The output layer is used to output the first prediction data, or to output the second prediction data and the prediction confidence.

[0027] Optionally, the feature extraction module includes a convolutional neural network, a long short-term memory network and a graph neural network;

[0028] The convolutional neural network is used to extract the first local time features of the historical energy consumption data and the historical community energy impact data, or to extract the second local time features of the posterior distribution and the future community energy impact data, respectively;

[0029] The long short-term memory network is used to generate the first long-term time feature according to the first local time feature, or is used to generate the second long-term time feature according to the second local time feature;

[0030] The graph neural network is used to respectively extract the first spatial features of the historical energy consumption data and the historical community energy impact data, or to respectively extract the second spatial features of the posterior distribution and the future community energy impact data;

[0031] The multimodal fusion module includes an attention unit and a multimodal feature fusion unit;

[0032] The attention unit is used to adopt a self-attention mechanism to convert the first long-term time feature and the first spatial feature into a first weighted time feature and a first weighted spatial feature, respectively, and adopt a cross-attention mechanism to convert the first weighted time feature and the first weighted spatial feature into a first cross-modal fusion feature, or to adopt the self-attention mechanism to convert the second long-term time feature and the second spatial feature into a second weighted time feature and a second weighted spatial feature, respectively, and adopt the cross-attention mechanism to convert the second weighted time feature and the second weighted spatial feature into a second cross-modal fusion feature;

[0033] The multimodal feature fusion unit is used to transform the first cross-modal fusion feature into the first comprehensive feature by using a multi-layer perceptron, or to transform the second cross-modal fusion feature into the second comprehensive feature.

[0034] Optionally, the integrated learning module includes a plurality of sub-prediction models, an integrated unit and a weighted average unit;

[0035] The sub-prediction model is used to generate a corresponding first sub-prediction result according to the first comprehensive feature, or is used to generate a corresponding second sub-prediction result according to the second comprehensive feature;

[0036] The integration unit is used to stack all of the first sub-prediction results to generate a first prediction result data set, or to stack all of the second sub-prediction results to generate a second prediction result data set;

[0037] The weighted averaging unit is used to generate the first prediction data according to the first prediction result data set and the weight data set, or is used to generate the second prediction data according to the second prediction result data set and the weight data set;

[0038] The uncertainty quantification module includes a Monte Carlo unit and a confidence interval unit;

[0039] The Monte Carlo unit is used to generate mean variance data corresponding to the second prediction data according to the Monte Carlo method and the second comprehensive feature;

[0040] The confidence interval unit is used to generate the prediction confidence according to the mean variance data.

[0041] Optionally, the determining the likelihood value based on the difference data based on the Bayesian theorem includes:

[0042] Based on the Bayesian theorem, according to the difference data, the likelihood value is determined using a Gaussian likelihood function formula, wherein the Gaussian likelihood function formula includes:

[0043] ;

[0044] in, is the likelihood value of the i-th first prediction data, are model parameters, is the i-th first prediction data, is the input feature vector of the i-th first prediction data, is the i-th historical energy consumption data, is the noise variance.

[0045] Optionally, determining the prior distribution of the energy consumption prediction model according to the influencing data distribution characteristics and expert knowledge includes:

[0046] The prior distribution is determined by using a weighted average method according to the influencing data distribution characteristics and the expert knowledge.

[0047] In a second aspect, the present invention provides a community energy consumption prediction device based on Bayesian calibration, which implements the community energy consumption prediction method based on Bayesian calibration as described in the first aspect. The community energy consumption prediction device based on Bayesian calibration includes:

[0048] An acquisition module, used to acquire historical energy consumption data, historical community energy impact data, and future community energy impact data;

[0049] a difference module, for inputting the historical energy consumption data and the historical community energy impact data into a trained energy consumption prediction model, generating first prediction data, and determining difference data between the first prediction data and the historical energy consumption data;

[0050] A distribution module, used to extract time series histograms of the historical energy consumption data and the historical community energy impact data, determine the distribution characteristics of the impact data according to the time series histograms, and determine the prior distribution of the energy consumption prediction model according to the distribution characteristics of the impact data and expert knowledge;

[0051] A calibration module, configured to determine a likelihood value based on the difference data based on the Bayesian theorem, and to generate a posterior distribution based on the prior distribution and the likelihood value;

[0052] A prediction module, used for inputting the posterior distribution and the future community energy impact data in multiple different future time intervals into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals;

[0053] An interval module, used to extract a first similarity between the impact data in each of the historical time intervals and the future community energy impact data in the corresponding future time interval, and extract a target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval;

[0054] a similarity module, configured to respectively extract the historical similar energy consumption data corresponding to the target historical time interval from the historical energy consumption data, and the predicted similar energy consumption data corresponding to the target future time interval from the second predicted data, and determine a second similarity between the historical similar energy consumption data and the predicted similar energy consumption data;

[0055] A result module is used to generate a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity.

[0056] In a third aspect, the present invention provides an electronic device, including a memory and a processor;

[0057] The memory is used to store computer programs;

[0058] The processor is used to implement the community energy consumption prediction method based on Bayesian calibration as described in the first aspect when executing the computer program.

[0059] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the community energy consumption prediction method based on Bayesian calibration as described in the first aspect is implemented.

[0060] The community energy consumption prediction method, device, electronic device and storage medium based on Bayesian calibration of the present invention have the following beneficial effects:

[0061] By inputting the acquired historical energy consumption data and community energy impact data into the trained energy consumption prediction model, generating the first prediction data, and determining the difference data between the first prediction data and the historical energy consumption data, the basic performance of the model can be evaluated, and the prediction accuracy of the current energy consumption prediction model can be determined, which is convenient for subsequent Bayesian calibration. Then, through the time series histogram of historical energy consumption data and historical community energy impact data, the distribution characteristics of the impact data are determined, and a reasonable prior distribution is set in combination with expert knowledge, which can improve the accuracy of parameter estimation in the existing Bayesian method and avoid the subjective problems caused by the existing Bayesian method relying entirely on expert experience. Then, based on the Bayesian theorem, the likelihood value is determined according to the difference data, and the data observed by the model, that is, the difference data, is taken into consideration. According to the prior distribution and the likelihood value, a more accurate parameter estimate, that is, the posterior distribution, is obtained through the Bayesian update process, so as to perform the first calibration to improve the adaptability and prediction accuracy of the model. The updated parameter distribution and future input data are then used for prediction, and the corresponding confidence level is given, so that the prediction result contains not only point estimates but also uncertainty information, which enhances the reliability of the prediction and facilitates the second calibration. The first similarity between the impact data in each historical time interval and the future community energy impact data in the corresponding future time interval is used, and the target time interval with the first similarity greater than the preset threshold is extracted. At the same time, the second similarity between the corresponding historical similar energy consumption data and the predicted similar energy consumption data in the target time interval is determined to identify the possible performance of energy consumption under similar conditions, providing reliable information for the second prediction calibration. Finally, the second calibration is performed based on the second prediction data, prediction confidence level and second similarity level to obtain an accurate final energy consumption prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a process flow of a community energy consumption prediction method based on Bayesian calibration provided in an embodiment of the present invention;

[0063] Figure 2 One of the structural schematic diagrams of the energy consumption prediction model provided by the embodiment of the present invention;

[0064] Figure 3 The second structural diagram of the energy consumption prediction model provided by the embodiment of the present invention;

[0065] Figure 4 A schematic diagram of the structure of a community energy consumption prediction device based on Bayesian calibration provided in an embodiment of the present invention;

[0066] Figure 5 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be interpreted as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the present invention.

[0068] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0069] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0070] It should be noted that the modifications of "one" and "plurality" mentioned in the present invention are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0071] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes, and are not used to limit the scope of these messages or information.

[0072] In view of the problems existing in the above-mentioned related technologies, this embodiment provides a community energy consumption prediction method, device, electronic device and storage medium based on Bayesian calibration.

[0073] like Figure 1 As shown, a community energy consumption prediction method based on Bayesian calibration provided by an embodiment of the present invention includes:

[0074] Get historical energy consumption data, historical community energy impact data, and future community energy impact data.

[0075] Specifically, historical energy consumption data refers to the energy consumption data of the community within a certain period of time in the past, historical community energy impact data refers to the data corresponding to the factors that have an impact on energy consumption within a certain period of time in the past, and future community energy impact data refers to the data corresponding to the factors that have an impact on energy consumption within the expected time in the future. Exemplarily, historical community energy impact data include historical meteorological data, historical community activity level data, historical renewable energy power generation data, and historical policy impact data; future community energy impact data include future meteorological data, future community activity level data, future renewable energy power generation data, and future policy impact data. Among them, meteorological data include temperature, humidity, wind speed, etc., which can be collected based on sensors, community activity level data can be calculated based on data such as population density and commercial activity intensity; renewable energy power generation data include solar energy and wind energy, which can be obtained through statistics, and policy impact data can be obtained based on statistics.

[0076] The historical energy consumption data and the historical community energy impact data are input into a trained energy consumption prediction model to generate first prediction data, and difference data between the first prediction data and the historical energy consumption data is determined.

[0077] Specifically, the energy consumption prediction model can adopt neural network models such as convolutional neural network models, random forest models, and optimization models. First, the historical energy consumption data and community energy impact data are input into the trained energy consumption prediction model to generate the first prediction data, and then the difference data between the first prediction data and the historical energy consumption data is determined, so as to make a preliminary evaluation of the energy consumption prediction model based on the difference data, so as to facilitate subsequent verification.

[0078] The time series histograms of the historical energy consumption data and the historical community energy impact data are extracted respectively, the distribution characteristics of the impact data are determined according to the time series histograms, and the prior distribution of the energy consumption prediction model is determined according to the distribution characteristics of the impact data and expert knowledge.

[0079] Specifically, in the prior art, the determination of the prior distribution in the Bayesian method mainly uses expert knowledge. Since expert knowledge is highly subjective and does not depend on objective conditions, the time series histograms of historical energy consumption data and historical community energy impact data are first extracted respectively. Through the analysis of the time series histograms, the distribution characteristics of the influencing data are determined, such as normal distribution, Gaussian distribution, etc., and the distribution characteristics of the influencing data and expert knowledge are combined to avoid the subjective problem caused by the existing Bayesian method relying entirely on expert experience, thereby improving the accuracy of the model, realizing the combination of subjective experience and objective facts, and determining the prior distribution of the energy consumption prediction model from two perspectives, thereby using the prior distribution to represent the uncertainty of the parameters. For example, normal distribution, uniform distribution or other suitable probability distributions can be used.

[0080] Based on Bayes' theorem, a likelihood value is determined according to the difference data, and a posterior distribution is generated according to the prior distribution and the likelihood value.

[0081] Specifically, a suitable likelihood function is first defined, and the likelihood value is determined based on the likelihood function and the difference data to measure the degree of match between the model prediction results and the actual observed data under a given set of parameter conditions. Then, the Bayesian formula of the Bayesian theorem is used to combine the prior distribution and the likelihood value obtained by the likelihood function to calculate the posterior distribution of the parameters. This step can be completed by numerical methods, such as Markov Chain Monte Carlo MCMC, or analytical methods.

[0082] The posterior distribution and the future community energy impact data in multiple different future time intervals are input into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals.

[0083] Specifically, the historical time interval corresponds to the future time interval, and there are multiple historical time intervals, each of which has corresponding historical energy consumption data and historical community energy impact data, and each future time interval has corresponding prediction data and future community energy impact data. The parameter values ​​sampled from the posterior distribution and the future community energy impact data are used for prediction, so that a series of prediction results can be obtained, reflecting the possibilities under different parameter combinations. By performing statistical analysis on these results, such as calculating the mean, variance, etc., the second prediction data and the corresponding prediction confidence can be obtained. Exemplarily, at this time, the historical energy consumption data can be used as an influencing factor and input into the trained energy consumption prediction model together with the future community energy impact data, because the historical energy consumption data may also affect the future energy consumption data. For example, the energy consumption data of the day before yesterday and yesterday will have a certain impact on tomorrow's energy consumption.

[0084] Extract the first similarity between the impact data in each historical time interval and the future community energy impact data in the corresponding future time interval, and extract the target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval.

[0085] Specifically, in some cases, the community energy impact data of a certain period in the past and a certain period in the future are similar or even the same, so in this case, the energy consumption caused by the same influencing factors should be similar, so the first similarity between the impact data in each historical time interval and the future community energy impact data in the corresponding future time interval is extracted, and the target time interval corresponding to the first similarity greater than the preset threshold is extracted to determine the time period where this situation exists, for example, Monday to Wednesday, Friday to Saturday, Sunday to next Tuesday, etc. Exemplarily, the prediction threshold is set according to the actual situation.

[0086] The historical similar energy consumption data corresponding to the target historical time interval and the predicted similar energy consumption data corresponding to the target future time interval in the second predicted data are respectively extracted to determine the second similarity between the historical similar energy consumption data and the predicted similar energy consumption data.

[0087] Specifically, energy consumption data corresponding to target time periods similar to historical community energy impact data and future community energy impact data are extracted, i.e., historical similar energy consumption data and predicted similar energy consumption data, so as to judge the similarity between the predicted results and the actual results under similar influencing factors of the energy consumption prediction model, i.e., the second similarity, and then judge the practicality and accuracy of the energy consumption prediction model.

[0088] A final energy consumption prediction result is generated according to the second prediction data, the prediction confidence and the second similarity.

[0089] Specifically, based on the model's prediction confidence for the second prediction data and the similarity between its predicted results and the actual situation under similar influencing factors, when both the prediction confidence and similarity are high, the final energy consumption prediction result is obtained; if both the similarities are low, the prediction is performed again.

[0090] In this embodiment, by inputting the acquired historical energy consumption data and community energy impact data into the trained energy consumption prediction model, generating the first prediction data, and determining the difference data between the first prediction data and the historical energy consumption data, the basic performance of the model can be evaluated, and the prediction accuracy of the current energy consumption prediction model can be determined, which is convenient for subsequent Bayesian calibration. Then, through the time series histogram of historical energy consumption data and historical community energy impact data, the distribution characteristics of the impact data are determined, and a reasonable prior distribution is set in combination with expert knowledge, which can improve the accuracy of parameter estimation in the existing Bayesian method and avoid the subjective problems caused by the existing Bayesian method relying entirely on expert experience. Then, based on the Bayesian theorem, the likelihood value is determined according to the difference data, and the data observed by the model, that is, the difference data, is taken into consideration, and according to the prior distribution and the likelihood value, a more accurate parameter estimate, that is, the posterior distribution, is obtained through the Bayesian update process, so as to perform the first calibration to improve the adaptability and prediction accuracy of the model. The updated parameter distribution and future input data are then used for prediction, and the corresponding confidence level is given at the same time, so that the prediction result contains not only point estimates but also uncertainty information, thereby enhancing the reliability of the prediction and facilitating the second calibration. The first similarity between the historical community energy impact data and the future community energy impact data is extracted. The first similarity between the impact data in each historical time interval and the future community energy impact data in the corresponding future time interval is used, and the target time interval whose first similarity is greater than a preset threshold is extracted. At the same time, the second similarity between the corresponding historical similar energy consumption data and the predicted similar energy consumption data in the target time interval is determined to identify the possible performance of energy consumption under similar conditions, thereby providing reliable information for the second prediction calibration. Finally, the second calibration is performed based on the second prediction data, prediction confidence level and second similarity level to obtain an accurate final energy consumption prediction result.

[0091] Optionally, generating a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity includes:

[0092] Extracting high-confidence prediction data whose prediction confidence is greater than a preset confidence from the second prediction data;

[0093] Associating the high-confidence prediction data with the second similarity, dividing the high-confidence prediction data according to the second similarity and a second preset similarity, and generating a determined prediction result and a pending prediction result;

[0094] Extracting the pending future time interval of the pending prediction result and the pending historical time interval corresponding to the pending future time interval, and extracting the future community energy impact data corresponding to the pending future time interval according to the pending future time interval, and extracting the historical energy consumption data corresponding to the pending historical time interval according to the pending historical time interval;

[0095] Inputting the historical energy consumption data corresponding to the undetermined historical time interval and the future community energy impact data corresponding to the undetermined future time interval into the trained energy consumption prediction model to generate a retest result;

[0096] The determined prediction result and the re-measurement result are integrated to generate the final energy consumption prediction result.

[0097] Specifically, the prediction data with the highest confidence in the second prediction data, that is, the high-confidence prediction data, is selected, and then the high-confidence prediction data is associated with the second similarity, and the high-confidence prediction data is divided according to the second similarity and the second preset similarity to generate a determined prediction result and a pending prediction result. It should be understood that not all high-confidence prediction data have a corresponding second similarity. At this time, only the high-confidence prediction data within the time interval is divided, and the high-confidence prediction data without the corresponding second similarity is directly used as the determined prediction result. For the pending prediction result, it is necessary to determine the time range corresponding to the pending prediction result, that is, the pending future time interval, and obtain the pending historical time interval corresponding to the pending future time interval, and input the future community energy impact data corresponding to the pending future time interval and the historical energy consumption data corresponding to the pending historical time interval into the trained energy consumption prediction model, and re-predict the pending future time interval to improve accuracy. It should be noted that when retesting, the historical energy consumption data is input into the energy consumption prediction model as prior knowledge to correct the prediction data according to the historical energy consumption data to obtain a more accurate result. Exemplarily, among all high-confidence prediction data, the high-confidence prediction data whose second similarity is greater than the second preset similarity can be used as the determined prediction result, and the high-confidence prediction data whose second similarity is less than or equal to the second preset similarity can be used as the pending prediction result.

[0098] Optionally, the trained energy consumption prediction model includes: an input layer, a feature extraction module, a multimodal fusion module, an integrated learning module, an uncertainty quantification module and an output layer;

[0099] The input layer is used to obtain the historical energy consumption data and the historical community energy impact data, or to obtain the posterior distribution and the future community energy impact data;

[0100] The feature extraction module is used to extract the first long-term time feature and the first spatial feature of the historical energy consumption data and the historical community energy impact data, or to extract the second long-term time feature and the second spatial feature of the posterior distribution and the future community energy impact data, respectively;

[0101] The multimodal fusion module is used to fuse the corresponding first long-term time feature and the first spatial feature respectively to generate a first comprehensive feature, or to fuse the corresponding second long-term time feature and the second spatial feature respectively to generate a second comprehensive feature;

[0102] The integrated learning module is used to perform integrated learning on all the first comprehensive features to generate the first prediction data, or to perform integrated learning on all the second comprehensive features to generate the second prediction data;

[0103] The uncertainty quantification module is used to generate the corresponding prediction confidence according to the second comprehensive feature and the second prediction data;

[0104] The output layer is used to output the first prediction data, or to output the second prediction data and the prediction confidence.

[0105] Specifically, Figure 2 As shown, the trained energy consumption prediction model includes an input layer, a feature extraction module, a multimodal fusion module, an integrated learning module, an uncertainty quantification module and an output layer, and the input layer, the feature extraction module, the multimodal fusion module, the integrated learning module, the uncertainty quantification module and the output layer are connected in sequence. The input layer is used to obtain historical energy consumption data and historical community energy impact data, or to obtain posterior distribution and future community energy impact data; the feature extraction module is used to extract the first long-term time features and first spatial features of historical energy consumption data and historical community energy impact data, or to extract the second long-term time features and second spatial features of posterior distribution and future community energy impact data, respectively; the multimodal fusion module is used to fuse the corresponding first long-term time features and first spatial features, respectively, to generate the first comprehensive feature, or to fuse the corresponding second long-term time features and second spatial features, respectively, to generate the second comprehensive feature; the integrated learning module is used to perform integrated learning on all first comprehensive features to generate the first prediction data, or to perform integrated learning on all second comprehensive features to generate the second prediction data; the uncertainty quantification module is used to generate the corresponding prediction confidence according to the second comprehensive feature and the second prediction data; the output layer is used to output the first prediction data, or to output the second prediction data and the prediction confidence.

[0106] Optionally, the feature extraction module includes a convolutional neural network, a long short-term memory network and a graph neural network;

[0107] The convolutional neural network is used to extract the first local time features of the historical energy consumption data and the historical community energy impact data, or to extract the second local time features of the posterior distribution and the future community energy impact data, respectively;

[0108] The long short-term memory network is used to generate the first long-term time feature according to the first local time feature, or is used to generate the second long-term time feature according to the second local time feature;

[0109] The graph neural network is used to respectively extract the first spatial features of the historical energy consumption data and the historical community energy impact data, or to respectively extract the second spatial features of the posterior distribution and the future community energy impact data;

[0110] The multimodal fusion module includes an attention unit and a multimodal feature fusion unit;

[0111] The attention unit is used to adopt a self-attention mechanism to convert the first long-term time feature and the first spatial feature into a first weighted time feature and a first weighted spatial feature, respectively, and adopt a cross-attention mechanism to convert the first weighted time feature and the first weighted spatial feature into a first cross-modal fusion feature, or to adopt the self-attention mechanism to convert the second long-term time feature and the second spatial feature into a second weighted time feature and a second weighted spatial feature, respectively, and adopt the cross-attention mechanism to convert the second weighted time feature and the second weighted spatial feature into a second cross-modal fusion feature;

[0112] The multimodal feature fusion unit is used to transform the first cross-modal fusion feature into the first comprehensive feature by using a multi-layer perceptron, or to transform the second cross-modal fusion feature into the second comprehensive feature.

[0113] Specifically, Figure 3As shown, the feature extraction module includes a convolutional neural network, a long short-term memory network and a graph neural network; the convolutional neural network is used to extract the first local time features of the historical energy consumption data and the historical community energy impact data, or to extract the second local time features of the posterior distribution and the future community energy impact data, respectively; the long short-term memory network is used to generate the first long-term time feature according to the first local time feature, or to generate the second long-term time feature according to the second local time feature; the graph neural network is used to extract the first spatial features of the historical energy consumption data and the historical community energy impact data, or to extract the second spatial features of the posterior distribution and the future community energy impact data, respectively; the multimodal fusion module includes an attention unit and a multimodal feature fusion unit; The attention unit is used to adopt a self-attention mechanism to convert the first long-term time feature and the first spatial feature into a first weighted time feature and a first weighted spatial feature, respectively, and adopt a cross-attention mechanism to convert the first weighted time feature and the first weighted spatial feature into a first cross-modal fusion feature, or to adopt a self-attention mechanism to convert the second long-term time feature and the second spatial feature into a second weighted time feature and a second weighted spatial feature, respectively, and adopt a cross-attention mechanism to convert the second weighted time feature and the second weighted spatial feature into a second cross-modal fusion feature; the multimodal feature fusion unit is used to adopt a multi-layer perceptron to convert the first cross-modal fusion feature into a first comprehensive feature, or to convert the second cross-modal fusion feature into a second comprehensive feature.

[0114] Optionally, the integrated learning module includes a plurality of sub-prediction models, an integrated unit and a weighted average unit;

[0115] The sub-prediction model is used to generate a corresponding first sub-prediction result according to the first comprehensive feature, or is used to generate a corresponding second sub-prediction result according to the second comprehensive feature;

[0116] The integration unit is used to stack all of the first sub-prediction results to generate a first prediction result data set, or to stack all of the second sub-prediction results to generate a second prediction result data set;

[0117] The weighted averaging unit is used to generate the first prediction data according to the first prediction result data set and the weight data set, or is used to generate the second prediction data according to the second prediction result data set and the weight data set;

[0118] The uncertainty quantification module includes a Monte Carlo unit and a confidence interval unit;

[0119] The Monte Carlo unit is used to generate mean variance data corresponding to the second prediction data according to the Monte Carlo method and the second comprehensive feature;

[0120] The confidence interval unit is used to generate the prediction confidence according to the mean variance data.

[0121] Specifically, Figure 3 As shown, the integrated learning module includes multiple sub-prediction models, an integration unit and a weighted average unit; the sub-prediction models include LSTM models, GRU models, CNN models and GNN models, etc., and the specific model can be set according to the actual situation. The sub-prediction model is used to generate the corresponding first sub-prediction result according to the first comprehensive feature, or to generate the corresponding second sub-prediction result according to the second comprehensive feature; the integration unit is used to stack all the first sub-prediction results to generate a first prediction result data set, or to stack all the second sub-prediction results to generate a second prediction result data set; the weighted average unit is used to generate the first prediction data according to the first prediction result data set and the weight data set, or to generate the second prediction data according to the second prediction result data set and the weight data set; the uncertainty quantification module includes a Monte Carlo unit and a confidence interval unit; the Monte Carlo unit is used to generate mean variance data corresponding to the second prediction data according to the Monte Carlo method and the second comprehensive feature; the confidence interval unit is used to generate prediction confidence according to the mean variance data.

[0122] Optionally, the determining the likelihood value based on the difference data based on the Bayesian theorem includes:

[0123] Based on the Bayesian theorem, according to the difference data, the likelihood value is determined using a Gaussian likelihood function formula, wherein the Gaussian likelihood function formula includes:

[0124] ;

[0125] in, is the likelihood value of the i-th first prediction data, are model parameters, is the i-th first prediction data, is the input feature vector of the i-th first prediction data, is the i-th historical energy consumption data, is the noise variance.

[0126] Specifically, after experimental calculation, the energy consumption data generally belongs to a normal distribution, so the Gaussian likelihood function formula is used to determine the likelihood value.

[0127] Optionally, determining the prior distribution of the energy consumption prediction model according to the influencing data distribution characteristics and expert knowledge includes:

[0128] The prior distribution is determined by using a weighted average method according to the influencing data distribution characteristics and the expert knowledge.

[0129] Specifically, the weighted average method is used to integrate the influencing data distribution characteristics and expert knowledge to balance the advantages of influencing data distribution characteristics and expert knowledge, thereby obtaining a more reasonable and accurate prior distribution.

[0130] like Figure 4 As shown, an embodiment of the present invention provides a community energy consumption prediction device 400 based on Bayesian calibration, which implements the community energy consumption prediction method based on Bayesian calibration as described above. The community energy consumption prediction device 400 based on Bayesian calibration includes:

[0131] An acquisition module 410 is used to acquire historical energy consumption data, historical community energy impact data, and future community energy impact data;

[0132] A difference module 420, for inputting the historical energy consumption data and the historical community energy impact data into a trained energy consumption prediction model, generating first prediction data, and determining difference data between the first prediction data and the historical energy consumption data;

[0133] A distribution module 430 is used to extract time series histograms of the historical energy consumption data and the historical community energy impact data, determine the impact data distribution characteristics according to the time series histograms, and determine the prior distribution of the energy consumption prediction model according to the impact data distribution characteristics and expert knowledge;

[0134] A calibration module 440, configured to determine a likelihood value based on the difference data based on Bayes' theorem, and to generate a posterior distribution based on the prior distribution and the likelihood value;

[0135] A prediction module 450 is used to input the posterior distribution and the future community energy impact data in multiple different future time intervals into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, and the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals;

[0136] The interval module 460 is used to extract the first similarity between the impact data in each of the historical time intervals and the future community energy impact data in the corresponding future time interval, and extract the target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval;

[0137] A similarity module 470 is used to respectively extract the historical similar energy consumption data corresponding to the target historical time interval from the historical energy consumption data, and the predicted similar energy consumption data corresponding to the target future time interval from the second predicted data, and determine a second similarity between the historical similar energy consumption data and the predicted similar energy consumption data;

[0138] The result module 480 is used to generate a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity.

[0139] like Figure 5 As shown, an electronic device 500 provided in an embodiment of the present invention includes a memory 510 and a processor 520; the memory 510 is used to store a computer program; the processor 520 is used to implement the community energy consumption prediction method based on Bayesian calibration as described above when executing the computer program.

[0140] In other words, an electronic device 500 includes a memory 510 and a processor 520 coupled to the memory 510; the memory 510 is configured to store a computer program; and the processor 520 is configured to perform the following operations when executing the computer program:

[0141] Obtain historical energy consumption data, historical community energy impact data, and future community energy impact data;

[0142] Inputting the historical energy consumption data and the historical community energy impact data into a trained energy consumption prediction model to generate first prediction data, and determining difference data between the first prediction data and the historical energy consumption data;

[0143] Extracting time series histograms of the historical energy consumption data and the historical community energy impact data respectively, determining the impact data distribution characteristics according to the time series histograms, and determining the prior distribution of the energy consumption prediction model according to the impact data distribution characteristics and expert knowledge;

[0144] Based on the Bayesian theorem, a likelihood value is determined according to the difference data, and a posterior distribution is generated according to the prior distribution and the likelihood value;

[0145] Input the posterior distribution and the future community energy impact data in multiple different future time intervals into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals;

[0146] Extracting a first similarity between the impact data in each of the historical time intervals and the future community energy impact data in the corresponding future time interval, and extracting a target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval;

[0147] Respectively extracting historical similar energy consumption data corresponding to the target historical time interval from the historical energy consumption data, and predicted similar energy consumption data corresponding to the target future time interval from the second predicted data, and determining a second similarity between the historical similar energy consumption data and the predicted similar energy consumption data;

[0148] A final energy consumption prediction result is generated according to the second prediction data, the prediction confidence and the second similarity.

[0149] In other words, a non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor performs the following operations:

[0150] Obtain historical energy consumption data, historical community energy impact data, and future community energy impact data;

[0151] Inputting the historical energy consumption data and the historical community energy impact data into a trained energy consumption prediction model to generate first prediction data, and determining difference data between the first prediction data and the historical energy consumption data;

[0152] Extracting time series histograms of the historical energy consumption data and the historical community energy impact data respectively, determining the impact data distribution characteristics according to the time series histograms, and determining the prior distribution of the energy consumption prediction model according to the impact data distribution characteristics and expert knowledge;

[0153] Based on the Bayesian theorem, a likelihood value is determined according to the difference data, and a posterior distribution is generated according to the prior distribution and the likelihood value;

[0154] Input the posterior distribution and the future community energy impact data in multiple different future time intervals into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals;

[0155] Extracting a first similarity between the impact data in each of the historical time intervals and the future community energy impact data in the corresponding future time interval, and extracting a target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval;

[0156] Respectively extracting historical similar energy consumption data corresponding to the target historical time interval from the historical energy consumption data, and predicted similar energy consumption data corresponding to the target future time interval from the second predicted data, and determining a second similarity between the historical similar energy consumption data and the predicted similar energy consumption data;

[0157] A final energy consumption prediction result is generated according to the second prediction data, the prediction confidence and the second similarity.

[0158] An electronic device 500 that can be used as a server or client of the present invention will now be described, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device 500 is intended to represent various forms of digital electronic computer equipment, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 500 can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples, and are not intended to limit the implementation of the present invention described and / or required herein.

[0159] The electronic device 500 includes a computing unit, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) or a computer program loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for the operation of the device can also be stored. The computing unit, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0160] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc. In the present application, the unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment of the present invention. In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0161] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A community energy consumption prediction method based on Bayesian calibration, characterized in that: include: Obtain historical energy consumption data, historical community energy impact data, and future community energy impact data; Inputting the historical energy consumption data and the historical community energy impact data into a trained energy consumption prediction model to generate first prediction data, and determining difference data between the first prediction data and the historical energy consumption data; Extracting time series histograms of the historical energy consumption data and the historical community energy impact data respectively, determining the impact data distribution characteristics according to the time series histograms, and determining the prior distribution of the energy consumption prediction model according to the impact data distribution characteristics and expert knowledge; Based on the Bayesian theorem, a likelihood value is determined according to the difference data, and a posterior distribution is generated according to the prior distribution and the likelihood value; Input the posterior distribution and the future community energy impact data in multiple different future time intervals into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals; Extracting a first similarity between the impact data in each of the historical time intervals and the future community energy impact data in the corresponding future time interval, and extracting a target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval; Respectively extracting historical similar energy consumption data corresponding to the target historical time interval from the historical energy consumption data, and predicted similar energy consumption data corresponding to the target future time interval from the second predicted data, and determining a second similarity between the historical similar energy consumption data and the predicted similar energy consumption data; Generate a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity; Generating a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity includes: Extracting high-confidence prediction data whose prediction confidence is greater than a preset confidence from the second prediction data; Associating the high-confidence prediction data with the second similarity, dividing the high-confidence prediction data according to the second similarity and a second preset similarity, and generating a determined prediction result and a pending prediction result; Extracting the pending future time interval of the pending prediction result and the pending historical time interval corresponding to the pending future time interval, and extracting the future community energy impact data corresponding to the pending future time interval according to the pending future time interval, and extracting the historical energy consumption data corresponding to the pending historical time interval according to the pending historical time interval; Inputting the historical energy consumption data corresponding to the to-be-determined historical time interval and the future community energy impact data corresponding to the to-be-determined future time interval into the trained energy consumption prediction model to generate a retest result; The determined prediction result and the re-measurement result are integrated to generate the final energy consumption prediction result.

2. The community energy consumption prediction method based on Bayesian calibration according to claim 1 is characterized in that: The trained energy consumption prediction model includes: an input layer, a feature extraction module, a multimodal fusion module, an integrated learning module, an uncertainty quantification module and an output layer; The input layer is used to obtain the historical energy consumption data and the historical community energy impact data, or to obtain the posterior distribution and the future community energy impact data; The feature extraction module is used to extract the first long-term time feature and the first spatial feature of the historical energy consumption data and the historical community energy impact data, or to extract the second long-term time feature and the second spatial feature of the posterior distribution and the future community energy impact data, respectively; The multimodal fusion module is used to fuse the corresponding first long-term time feature and the first spatial feature respectively to generate a first comprehensive feature, or to fuse the corresponding second long-term time feature and the second spatial feature respectively to generate a second comprehensive feature; The integrated learning module is used to perform integrated learning on all the first comprehensive features to generate the first prediction data, or to perform integrated learning on all the second comprehensive features to generate the second prediction data; The uncertainty quantification module is used to generate the corresponding prediction confidence according to the second comprehensive feature and the second prediction data; The output layer is used to output the first prediction data, or to output the second prediction data and the prediction confidence.

3. The community energy consumption prediction method based on Bayesian calibration according to claim 2 is characterized in that: The feature extraction module includes a convolutional neural network, a long short-term memory network and a graph neural network; The convolutional neural network is used to extract the first local time features of the historical energy consumption data and the historical community energy impact data, or to extract the second local time features of the posterior distribution and the future community energy impact data, respectively; The long short-term memory network is used to generate the first long-term time feature according to the first local time feature, or is used to generate the second long-term time feature according to the second local time feature; The graph neural network is used to respectively extract first spatial features of the historical energy consumption data and the historical community energy impact data, or to respectively extract second spatial features of the posterior distribution and the future community energy impact data; The multimodal fusion module includes an attention unit and a multimodal feature fusion unit; The attention unit is used to adopt a self-attention mechanism to convert the first long-term time feature and the first spatial feature into a first weighted time feature and a first weighted spatial feature, respectively, and adopt a cross-attention mechanism to convert the first weighted time feature and the first weighted spatial feature into a first cross-modal fusion feature, or to adopt the self-attention mechanism to convert the second long-term time feature and the second spatial feature into a second weighted time feature and a second weighted spatial feature, respectively, and adopt the cross-attention mechanism to convert the second weighted time feature and the second weighted spatial feature into a second cross-modal fusion feature; The multimodal feature fusion unit is used to transform the first cross-modal fusion feature into the first comprehensive feature by using a multi-layer perceptron, or to transform the second cross-modal fusion feature into the second comprehensive feature.

4. The community energy consumption prediction method based on Bayesian calibration according to claim 3 is characterized in that: The integrated learning module includes multiple sub-prediction models, an integrated unit and a weighted average unit; The sub-prediction model is used to generate a corresponding first sub-prediction result according to the first comprehensive feature, or is used to generate a corresponding second sub-prediction result according to the second comprehensive feature; The integration unit is used to stack all of the first sub-prediction results to generate a first prediction result data set, or to stack all of the second sub-prediction results to generate a second prediction result data set; The weighted averaging unit is used to generate the first prediction data according to the first prediction result data set and the weight data set, or is used to generate the second prediction data according to the second prediction result data set and the weight data set; The uncertainty quantification module includes a Monte Carlo unit and a confidence interval unit; The Monte Carlo unit is used to generate mean variance data corresponding to the second prediction data according to the Monte Carlo method and the second comprehensive feature; The confidence interval unit is used to generate the prediction confidence according to the mean variance data.

5. The community energy consumption prediction method based on Bayesian calibration according to claim 1 is characterized in that: The determining of the likelihood value based on the difference data based on the Bayesian theorem includes: Based on the Bayesian theorem, according to the difference data, the likelihood value is determined using a Gaussian likelihood function formula, wherein the Gaussian likelihood function formula includes: ； in, is the likelihood value of the i-th first prediction data, are model parameters, is the i-th first prediction data, is the input feature vector of the i-th first prediction data, is the i-th historical energy consumption data, is the noise variance.

6. The community energy consumption prediction method based on Bayesian calibration according to claim 1 is characterized in that: Determining the prior distribution of the energy consumption prediction model according to the influencing data distribution characteristics and expert knowledge includes: The prior distribution is determined by using a weighted average method according to the influencing data distribution characteristics and the expert knowledge.

7. A community energy consumption prediction device based on Bayesian calibration, characterized in that: To implement the community energy consumption prediction method based on Bayesian calibration as described in any one of claims 1 to 6, the community energy consumption prediction device based on Bayesian calibration includes: An acquisition module, used to acquire historical energy consumption data, historical community energy impact data, and future community energy impact data; a difference module, for inputting the historical energy consumption data and the historical community energy impact data into a trained energy consumption prediction model, generating first prediction data, and determining difference data between the first prediction data and the historical energy consumption data; A distribution module, used to extract time series histograms of the historical energy consumption data and the historical community energy impact data, determine the distribution characteristics of the impact data according to the time series histograms, and determine the prior distribution of the energy consumption prediction model according to the distribution characteristics of the impact data and expert knowledge; A calibration module, configured to determine a likelihood value based on the difference data based on the Bayesian theorem, and to generate a posterior distribution based on the prior distribution and the likelihood value; A prediction module, used for inputting the posterior distribution and the future community energy impact data in multiple different future time intervals into the trained energy consumption prediction model to generate second prediction data and corresponding prediction confidence, wherein the second prediction data includes prediction data in multiple future time intervals, the historical energy consumption data includes energy consumption data in multiple different historical time intervals, the historical community energy impact data includes impact data in multiple historical time intervals, and the historical time intervals correspond one-to-one to the future time intervals; An interval module, used to extract a first similarity between the impact data in each of the historical time intervals and the future community energy impact data in the corresponding future time interval, and extract a target time interval corresponding to the first similarity greater than a preset threshold, wherein the target time interval includes a target historical time interval and a target future time interval; a similarity module, configured to respectively extract the historical similar energy consumption data corresponding to the target historical time interval from the historical energy consumption data, and the predicted similar energy consumption data corresponding to the target future time interval from the second predicted data, and determine a second similarity between the historical similar energy consumption data and the predicted similar energy consumption data; A result module, used for generating a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity; Generating a final energy consumption prediction result according to the second prediction data, the prediction confidence and the second similarity includes: Extracting high-confidence prediction data whose prediction confidence is greater than a preset confidence from the second prediction data; Associating the high-confidence prediction data with the second similarity, dividing the high-confidence prediction data according to the second similarity and a second preset similarity, and generating a determined prediction result and a pending prediction result; Extracting the pending future time interval of the pending prediction result and the pending historical time interval corresponding to the pending future time interval, and extracting the future community energy impact data corresponding to the pending future time interval according to the pending future time interval, and extracting the historical energy consumption data corresponding to the pending historical time interval according to the pending historical time interval; Inputting the historical energy consumption data corresponding to the undetermined historical time interval and the future community energy impact data corresponding to the undetermined future time interval into the trained energy consumption prediction model to generate a retest result; The determined prediction result and the re-measurement result are integrated to generate the final energy consumption prediction result.

8. An electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is used to implement the community energy consumption prediction method based on Bayesian calibration as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the community energy consumption prediction method based on Bayesian calibration as described in any one of claims 1 to 6 is implemented.

Images