CN118428774A

CN118428774A - Intelligent analysis method and system for energy consumption based on time sequence and load characteristics

Info

Publication number: CN118428774A
Application number: CN202410874562.2A
Authority: CN
Inventors: 许可; 何智勇; 朱琪; 许新; 刘菁钰; 程莹; 项扬; 谢晓雯; 李天扬; 张皓明
Original assignee: Three Gorges Group Zhejiang Energy Investment Co ltd
Current assignee: Three Gorges Group Zhejiang Energy Investment Co ltd
Priority date: 2024-07-02
Filing date: 2024-07-02
Publication date: 2024-08-02

Abstract

The application provides an intelligent analysis method and system for energy consumption based on time sequence and load characteristics, and particularly relates to the technical field of energy consumption analysis. The method comprises the steps of obtaining at least one target external factor data affecting equipment energy consumption in a period to be predicted, wherein the target external factor is determined according to the association relation between the equipment energy consumption and the external factor. And determining a prediction feature vector corresponding to the period to be predicted according to the at least one target external factor data and the equipment use mode of the key equipment. And inputting the predicted feature vector into a factor analysis model to obtain the equipment predicted load of the period to be predicted and the energy demand corresponding to the equipment predicted load, wherein the factor analysis model is determined according to the first historical energy consumption data and the historical equipment load features. The method can integrate real-time data flow and timely respond to external changes, and adopts external influence factors and equipment use modes to jointly predict the future energy demand, so that the prediction accuracy is improved.

Description

Intelligent analysis method and system for energy consumption based on time sequence and load characteristics

Technical Field

The application relates to the technical field of energy consumption analysis, in particular to an intelligent energy consumption analysis method and system based on time sequences and load characteristics.

Background

In modern industrial and commercial operations, energy management is an important area, especially in the context of ever increasing energy costs and increasingly stringent environmental regulations. The energy management means planning, controlling and optimizing the energy in a scientific and reasonable way so as to improve the energy utilization efficiency, reduce the energy consumption and reduce the environmental impact.

The energy management can optimize the energy use process, adopts high-efficiency equipment and technology, can obviously improve energy utilization efficiency, reduces energy waste, thereby reduces energy cost, can reduce energy consumption through energy management, and through diversified energy structure, improves energy self-supporting rate and reserve capacity, can strengthen the stability and the security of energy supply, reduces the dependence to external energy.

Conventional energy management systems typically rely on basic monitoring and control techniques that provide limited data analysis functionality, and are difficult to handle with complex data relationships or varying energy requirements in real time. In addition, the conventional method has obvious defects in data integration, real-time data processing and dynamic energy optimization, cannot process a large amount of data streams in real time, is difficult to quickly respond to sudden changes of external environments, and lacks effective tools to analyze and utilize complex relationships between equipment load characteristics and various external factors, which limits the improvement of energy efficiency and the further reduction of cost.

Disclosure of Invention

The application provides an intelligent analysis method and system for energy consumption based on time sequence and load characteristics, which are used for solving the problem that the prior art is difficult to process complex data relationship or real-time changing energy demand.

In a first aspect, the present application provides an intelligent analysis method for energy consumption based on time series and load characteristics, comprising:

acquiring at least one target external factor data affecting the energy consumption of the equipment in a period to be predicted, wherein the target external factor is determined according to the association relationship between the energy consumption of the equipment and the external factor;

determining a prediction feature vector corresponding to a period to be predicted according to the at least one target external factor data and the equipment use mode of the key equipment;

And inputting the prediction feature vector into a factor analysis model to obtain the equipment prediction load of the period to be predicted and the energy demand corresponding to the equipment prediction load, wherein the factor analysis model is determined according to the first historical energy consumption data and the historical equipment load features.

Optionally, before the obtaining at least one target external factor data affecting the energy consumption of the device in the period to be predicted, the method further includes:

acquiring first historical energy consumption data of the key equipment, load characteristic data corresponding to the first historical energy consumption data and a plurality of historical external factor data;

Inputting the first historical energy consumption data into a time cycle model to obtain an energy consumption mode output by the time cycle model;

training a general analysis model according to the energy consumption mode, the load characteristic data and a plurality of historical external factor data to obtain the factor analysis model after training, wherein the factor analysis model is used for indicating the association relation between the equipment energy consumption and the external factors.

Optionally, before the first historical energy consumption data is input into the time cycle model to obtain the energy consumption mode output by the time cycle model, the method further includes:

Screening the first historical energy consumption data to obtain second historical energy consumption data;

analyzing and processing the second historical energy consumption data to obtain seasonal characteristics and nonlinear characteristics of the second historical energy consumption data;

determining a neuron number range, a hierarchical number range and a time window of an initial time cycle model according to the seasonal characteristic and the nonlinear characteristic, wherein the time window is a time length covering a complete seasonal period;

Determining a plurality of network configurations according to the neuron number range and the hierarchy number range, wherein one neuron number and one hierarchy number form one network configuration;

Training and verifying the network configurations to obtain a target network configuration with the minimum verification error;

And training and optimizing the initial time circulation model according to the time window and the target network configuration, and determining a time circulation model.

Optionally, the screening the first historical energy consumption data to obtain second historical energy consumption data includes:

calculating the first historical energy consumption data, and determining standard scores of each first historical energy consumption data;

Judging whether the absolute value of the standard score is larger than a preset value or not;

If the absolute value of the standard score is larger than a preset value, determining that the first historical energy consumption data corresponding to the standard score is an abnormal point, and eliminating the abnormal point;

and if the absolute value of the standard score is not greater than a preset value, determining that the first historical energy consumption data corresponding to the standard score is the second historical energy consumption data.

Optionally, the training the general analysis model according to the energy consumption mode, the load characteristic data and the plurality of historical external factor data to obtain the trained factor analysis model includes:

Initializing the general analysis model to obtain a configured general analysis model;

normalizing the load characteristic data and the plurality of historical external factor data to obtain characteristic data;

And training the configured general analysis model according to the energy consumption mode and the characteristic data to obtain a trained factor analysis model.

Optionally, the determining, according to the at least one target external factor data and the device usage mode of the key device, a prediction feature vector corresponding to the period to be predicted includes:

Preprocessing the at least one target external factor data to obtain processed target external factor data;

And determining a prediction feature vector corresponding to the period to be predicted according to the processed target external factor data and the equipment use mode.

Optionally, the method further comprises:

A feedback collection system is adopted to monitor and acquire actual energy consumption data in real time;

determining an energy deviation value according to the energy demand corresponding to the actual energy consumption data and the equipment prediction load;

and updating the model parameters of the factor analysis model according to the energy deviation value, and adjusting an energy distribution strategy and a consumption strategy.

In a second aspect, the present application provides an intelligent analysis device for energy consumption based on time series and load characteristics, comprising:

The device comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring at least one target external factor data influencing the energy consumption of the device in a period to be predicted, and the target external factor is determined according to the association relation between the energy consumption of the device and the external factor;

The determining module is used for determining a prediction feature vector corresponding to a period to be predicted according to the at least one target external factor data and the equipment use mode of the key equipment;

The input module is used for inputting the prediction feature vector into a factor analysis model to obtain equipment prediction load of a period to be predicted and energy demand corresponding to the equipment prediction load, and the factor analysis model is obtained by determining according to first historical energy consumption data and historical equipment load features.

Optionally, the apparatus further includes: a training module;

The acquisition module is further used for acquiring first historical energy consumption data of the key equipment, load characteristic data corresponding to the first historical energy consumption data and a plurality of historical external factor data;

The input module is further configured to input the first historical energy consumption data into a time cycle model, so as to obtain an energy consumption mode output by the time cycle model;

The training module is used for training the general analysis model according to the energy consumption mode, the load characteristic data and the plurality of historical external factor data to obtain the factor analysis model after training, and the factor analysis model is used for indicating the association relation between the equipment energy consumption and the external factors.

Optionally, the apparatus further includes: a processing module;

the processing module is used for screening the first historical energy consumption data to obtain second historical energy consumption data;

The processing module is further used for analyzing and processing the second historical energy consumption data to obtain seasonal characteristics and nonlinear characteristics of the second historical energy consumption data;

the determining module is further configured to determine a neuron number range, a hierarchical number range and a time window of the initial time cycle model according to the seasonal characteristic and the nonlinear characteristic, where the time window is a time length covering a complete seasonal period;

the determining module is further configured to determine a plurality of network configurations according to the neuron number range and the hierarchy number range, where one neuron number and one hierarchy number form one network configuration;

the processing module is further used for training and verifying the plurality of network configurations to obtain a target network configuration with the minimum verification error;

The determining module is further configured to perform training optimization on the initial time circulation model according to the time window and the target network configuration, and determine a time circulation model.

Optionally, the apparatus further includes: a judging module;

the processing module is also used for carrying out calculation processing on the first historical energy consumption data and determining standard scores of each first historical energy consumption data;

The judging module is used for judging whether the absolute value of the standard score is larger than a preset value or not;

The determining module is further configured to determine that the first historical energy consumption data corresponding to the standard score is an outlier if the absolute value of the standard score is greater than a preset value, and reject the outlier;

and the determining module is further configured to determine that the first historical energy consumption data corresponding to the standard score is the second historical energy consumption data if the absolute value of the standard score is not greater than a preset value.

Optionally, the processing module is further configured to initialize the general analysis model to obtain a configured general analysis model;

The processing module is further used for carrying out normalization processing on the load characteristic data and the plurality of historical external factor data to obtain characteristic data;

And the processing module is also used for training the general analysis model with the configured energy consumption mode and the characteristic data to obtain a factor analysis model with the configured energy consumption mode and the characteristic data.

Optionally, the processing module is further configured to pre-process the at least one target external factor data to obtain processed target external factor data;

The determining module is further configured to determine a prediction feature vector corresponding to the period to be predicted according to the processed target external factor data and the device usage mode.

Optionally, the acquiring module is further configured to monitor and acquire actual energy consumption data in real time by adopting a feedback collecting system;

The determining module is also used for determining an energy deviation value according to the energy demand corresponding to the actual energy consumption data and the equipment prediction load;

The processing module is further used for updating the model parameters of the factor analysis model according to the energy deviation value and adjusting an energy distribution strategy and a consumption strategy.

In a third aspect, the present application provides an intelligent energy consumption analysis device based on time series and load characteristics, the device comprising:

a memory;

A processor;

wherein the memory stores computer-executable instructions;

the processor executes the computer-executable instructions stored in the memory to implement the intelligent analysis method for energy consumption based on time series and load characteristics as described in the first aspect and the various possible implementation manners of the first aspect.

In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a computer program for execution by a processor to implement the method for intelligent analysis of energy consumption based on time series and load characteristics as described in the first aspect and various possible implementations of the first aspect.

According to the intelligent analysis method and system for energy consumption based on time sequence and load characteristics, at least one target external factor data influencing equipment energy consumption in a period to be predicted is obtained. And determining a prediction feature vector corresponding to the period to be predicted according to the at least one target external factor data and the equipment use mode of the key equipment. And inputting the prediction feature vector into the factor analysis model to obtain the equipment prediction load of the period to be predicted and the energy demand corresponding to the equipment prediction load. The method can integrate real-time data flow and timely respond to external changes, and adopts external influence factors and equipment use modes to jointly predict the future energy demand, so that the prediction accuracy is improved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.

FIG. 1 is a schematic diagram of a framework of an intelligent analysis system for energy consumption according to an embodiment of the present application;

Fig. 2 is a schematic flow chart of an intelligent analysis method for energy consumption based on time sequence and load characteristics according to an embodiment of the present application;

fig. 3 is a second schematic flow chart of an intelligent analysis method for energy consumption based on time sequence and load characteristics according to an embodiment of the present application;

fig. 4 is a flowchart of an intelligent analysis method for energy consumption based on time sequence and load characteristics according to an embodiment of the present application;

Fig. 5 is a flow chart diagram of an intelligent analysis method for energy consumption based on time sequence and load characteristics according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of an intelligent analysis device for energy consumption based on time sequence and load characteristics provided by the application;

fig. 7 is a schematic structural diagram of an intelligent analysis device for energy consumption based on time sequence and load characteristics.

Specific embodiments of the present application have been shown by way of the above drawings and will be described in more detail below. The drawings and the written description are not intended to limit the scope of the inventive concepts in any way, but rather to illustrate the inventive concepts to those skilled in the art by reference to the specific embodiments.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the application, as detailed in the accompanying claims, rather than all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented, for example, in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, article, or apparatus.

In embodiments of the application, words such as "exemplary" or "such as" are used to mean examples, illustrations, or descriptions. Any embodiment or design described herein as "exemplary" or "for example" should not be construed as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "such as" is intended to present related concepts in a concrete fashion.

Conventional energy management systems typically rely on basic monitoring and control techniques that provide limited data analysis functionality, and are difficult to handle with complex data relationships or varying energy requirements in real time. Furthermore, the conventional method has significant disadvantages in terms of data integration, real-time data processing, and dynamic energy optimization, which limits the increase in energy efficiency and further reduction in cost. There is a lack of efficient tools in the art to analyze and exploit complex relationships between device load characteristics and a variety of external factors. Conventional energy management systems often cannot handle large data streams in real time and have difficulty responding quickly to sudden changes in the external environment. The existing energy management solution generally uses a model fixedly, lacks flexibility, cannot be automatically adjusted when facing to the change of data characteristics, and adopts a static or manual decision-making mode on energy distribution and consumption strategy formulation by the traditional method, so that the efficiency is low and the method is not accurate enough.

Aiming at the problems, the application provides an intelligent analysis method for energy consumption based on time sequence and load characteristics, which can effectively identify peak and valley periods of energy consumption by monitoring and predicting energy demand in real time, and can optimize energy distribution and adjust equipment operation plans by utilizing the information, thereby enhancing adaptability and response capability of energy management. The energy consumption intelligent analysis method and system based on time sequence and load characteristics for real-time data processing and response, self-adaptive prediction and model optimization and intelligent energy distribution strategy can accurately capture and analyze the complex data interactions, realize intelligent analysis of energy consumption and optimize energy use.

Fig. 1 is a schematic diagram of a framework of an intelligent analysis system for energy consumption according to an embodiment of the present application. As shown in fig. 1, the system includes: the system comprises a data collection module, a data processing and analyzing module, a model optimization and self-adaptive adjustment module, a real-time monitoring and response module and an energy strategy optimization module.

The data collection module comprises: multifunctional energy meter interface: the method is responsible for monitoring and recording the energy use conditions of various key devices in real time, and ensuring the accuracy of data and the consistency of time; sensor network: the method comprises the steps of deploying on key equipment, and collecting running state data of the equipment, wherein the running state data comprise start/stop time, running frequency, temperature, pressure and the like; a data transmission system: the collected data is synchronized to a central data processing center via a secure network connection.

The data processing and analyzing module comprises: pretreatment submodule: cleaning, normalizing and outlier detection are carried out on the collected data, and abnormal data points are processed by using methods such as Z-Score and the like; time series analysis tool: identifying a basic mode and periodic variation of energy consumption by utilizing technologies such as autoregressive, seasonal decomposition and the like; factor analyzer model: and establishing a prediction model according to the complex relation between the load characteristics of the equipment and external factors, and capturing and analyzing interaction among the characteristics.

The model optimization and self-adaption adjustment module comprises the following components: dynamic network adjustment tool: dynamically adjusting the number of layers and the number of neurons of the LSTM network according to seasonal and nonlinear characteristics of the data; parameter optimization algorithm: and the grid searching and cross-validation method is adopted to optimize model parameters, so that the prediction accuracy and the generalization capability of the model are improved.

The real-time monitoring and response module comprises: and a data stream processing platform: integrating a real-time data stream processing technology, receiving and analyzing new data in real time, and rapidly responding to external changes and emergencies; an intelligent scheduling system: based on the model predictions, the energy allocation and equipment operating strategies are dynamically adjusted, such as adjusting the load or switching to a backup energy system during the predicted high energy periods.

The energy strategy optimization module comprises: intelligent energy allocation strategy: optimizing energy distribution according to the prediction result, and implementing a differential pricing strategy and a demand response measure to reduce energy cost and improve efficiency; policy adjustment feedback mechanism: and continuously adjusting and optimizing the prediction model and the energy use strategy according to the actual energy consumption condition and the user feedback.

The following describes the technical scheme of the present application and how the technical scheme of the present application solves the above technical problems in detail with specific embodiments. The following embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. Embodiments of the present application will be described below with reference to the accompanying drawings.

Fig. 2 is a schematic flow chart of an intelligent analysis method for energy consumption based on time series and load characteristics according to an embodiment of the present application. The execution subject of the embodiment is an intelligent analysis system for energy consumption. As shown in fig. 2, the method includes:

S101: and acquiring at least one target external factor data influencing the energy consumption of the equipment in the period to be predicted, wherein the target external factor is determined according to the association relationship between the energy consumption of the equipment and the external factor.

The target external factor data refer to external factor data influencing the energy consumption of the equipment. For example, the external factor data may be weather forecast data of a period to be predicted, and day/non-day information of the period to be predicted.

It will be appreciated that external factors may have different effects on the energy consumption of the device, for example, on a weekday when the energy demand is high, the time the device is running may be longer than on a non-weekday when the energy consumption of the device is high, so that external factors are also an important consideration when predicting the energy demand. After the external factor data is acquired, the data is first processed to make it suitable for model input. The numerical data is normalized to be uniform to the range of [0,1] or [ -1,1], and the category data is converted into numerical values by single thermal coding. By way of example, the numeric data includes: the temperature and humidity, and the type data comprise information such as hours, days, months and the like, workdays, non-workdays, holidays and the like.

S102: and determining a prediction feature vector corresponding to the period to be predicted according to the at least one target external factor data and the equipment use mode of the key equipment.

Optionally, preprocessing the at least one target external factor data to obtain processed target external factor data, and determining a prediction feature vector corresponding to a period to be predicted according to the processed target external factor data and the equipment use mode.

It will be appreciated that from the collected external factor data and the known device usage patterns of the critical devices, feature vectors are generated at future points in time, which will be input to the factor analysis model for predicting future energy demands and device loads. The future point in time, i.e. the period to be predicted.

S103: and inputting the prediction feature vector into a factor analysis model to obtain the equipment prediction load of the period to be predicted and the energy demand corresponding to the equipment prediction load, wherein the factor analysis model is determined according to the first historical energy consumption data and the historical equipment load features.

It can be appreciated that the factor analysis model is used to predict the feature vector at the future time and output the predicted energy demand. The factor analysis model comprehensively considers the influence of each characteristic and the interaction effect thereof, predicts the equipment load conditions of different time points, analyzes the prediction results, particularly pays attention to the time period with abnormally high predicted energy consumption and the influence of important external factors, applies the prediction results to an energy management system, and can optimize the energy distribution and equipment operation strategies.

Optionally, after obtaining the equipment predicted load of the period to be predicted and the energy demand corresponding to the equipment predicted load, according to the prediction result, the energy distribution and consumption strategy can be optimized to improve the energy use efficiency.

And analyzing and processing the prediction result, and identifying key influence factors of the high energy consumption time period and the occurrence of the energy consumption peak. Analysis of the model prediction results, focusing on predicting periods of abnormally high energy consumption, due to specific external conditions (extreme weather) or specific times (weekday peak hours), and determining major external factors affecting the peak energy consumption. By way of example, the external factor may be a change in temperature, humidity or a special event.

Optimizing energy distribution according to the analysis result, and adjusting the operation time of the high-energy-consumption equipment for the period with high predicted energy consumption, for example, adjusting part of non-urgent tasks or operation to a low-peak period. In the case of accurate peak energy consumption predictions, the energy consumption during these periods is reduced by demand response strategies, such as adjusting temperature control settings, limiting certain operations, or starting up the standby energy system.

And meanwhile, optimizing the strategy of energy consumption, and executing a time difference pricing strategy according to the predicted result of energy demand so as to encourage the user to use more energy in the low peak period and reduce the energy use in the peak period. Based on the energy demand prediction, energy purchasing decision is made in advance, and optimal purchasing time and contracts are selected by using prediction information so as to reduce energy cost.

According to the intelligent analysis method for energy consumption based on the time sequence and the load characteristics, at least one target external factor data influencing the energy consumption of the equipment in a period to be predicted is obtained, wherein the target external factor is determined according to the association relationship between the energy consumption of the equipment and the external factor. And determining a prediction feature vector corresponding to the period to be predicted according to the at least one target external factor data and the equipment use mode of the key equipment. And inputting the predicted feature vector into a factor analysis model to obtain the equipment predicted load of the period to be predicted and the energy demand corresponding to the equipment predicted load, wherein the factor analysis model is determined according to the first historical energy consumption data and the historical equipment load features. The method can integrate real-time data flow and timely respond to external changes, and adopts external influence factors and equipment use modes to jointly predict the future energy demand, so that the prediction accuracy is improved.

Fig. 3 is a second schematic flow chart of an intelligent analysis method for energy consumption based on time sequence and load characteristics according to an embodiment of the present application. This embodiment is a detailed description of the time-loop model training process in the intelligent analysis method for energy consumption based on time series and load characteristics based on the embodiment of fig. 2. As shown in fig. 3, the method includes:

s201: and acquiring first historical energy consumption data of the key equipment, load characteristic data corresponding to the first historical energy consumption data and a plurality of historical external factor data.

It can be appreciated that the key devices are integrated by using the multifunctional energy meter to continuously record the energy consumption of the devices, wherein the energy comprises electric energy, water quantity, fuel gas and the like. The energy meter has the functions of data recording and time marking, and can ensure the accuracy of data and the consistency of time. Device operational status data including start/stop times, operating frequencies, temperatures, pressures, and other operating parameters of the device are collected by sensors mounted on the device.

The real-time data acquisition system automatically collects data from the energy meter and the sensor according to the data acquisition time interval, the collected data is used as a data set, the data is transmitted to the central data processing center through the safe network connection, the real-time data acquisition system and the energy meter and the sensor synchronously operate, and the data acquisition time interval is set to ensure that the data is updated according to the preset time period (each hour and each day), so that the analysis time sequence requirement is met. At the same time, in order to ensure high reliability and traceability of the data set, encryption and security measures are adopted to protect the safety and integrity of the data in the transmission process.

S202: and calculating the first historical energy consumption data, and determining a standard score of each first historical energy consumption data.

It will be appreciated that prior to analysis of the data set, statistical-based outlier detection is introduced to identify and process outliers in the energy consumption data, adding environmental and socioeconomic variables as input features in addition to temporal features, including holidays, weekdays and weekends, and processing and identifying outliers (outliers) is critical in the processing of the energy consumption data because they represent data collection errors, equipment failures or unusual usage patterns, which distort the results of subsequent analysis.

Outliers can be effectively identified by using a statistical-based outlier detection method, i.e. a Z-Score method, which is applicable to normally distributed datasets. Z-Score is a method of measuring the degree of deviation of individual data points from the mean, in which the Z-Score value for each point represents the standard deviation number from the mean of the dataset, and the calculation formula for Z-Score is:

wherein X is an observed value, Is the average value of the samples,Is the standard deviation of the sample, the average value and standard deviation of the data set are calculated according to the formula, and for each observed value in the data setIts standard Score Z-Score value was obtained.

S203: and judging whether the absolute value of the standard score is larger than a preset value, if so, executing the step S204, and if not, executing the step S205.

It will be appreciated that identifying outliers based on criteria, selecting a predetermined valueComparing the preset value with the standard score, and screening out abnormal points in the data set. For example, the preset value may be set to 2.5, 3, or 3.5. The setting of the preset value depends on how strictly outlier recognition is desired.

S204: if the absolute value of the standard score is larger than a preset value, determining that the first historical energy consumption data corresponding to the standard score is an abnormal point, and eliminating the abnormal point.

It will be appreciated that ifThenIs considered as an outlier, i.e. an outlier, which affects the training of the predictive model, so the data value is directly rejected.

S205: and if the absolute value of the standard score is not greater than a preset value, determining that the first historical energy consumption data corresponding to the standard score is the second historical energy consumption data.

It will be appreciated that ifThenAnd taking the data value as second historical energy consumption data, analyzing and processing the second historical energy consumption data after determining all the second historical energy consumption data, and training the model by utilizing the characteristics of the second historical energy consumption data.

S206: and analyzing and processing the second historical energy consumption data to obtain seasonal characteristics and nonlinear characteristics of the second historical energy consumption data.

It will be appreciated that the seasonal features of the data are analyzed using autocorrelation and partial autocorrelation functions (ACF and PACF), with ACF showing a distinct periodic pattern, indicating that the data is strongly seasonal. By visualizing the distribution of the data, or using a non-linearity test to evaluate the degree of non-linearity of the data, the BDS test may be employed to evaluate the degree of non-linearity of the data volume, for example.

S207: and determining the neuron number range, the hierarchical number range and the time window of the initial time cycle model according to the seasonal characteristic and the nonlinear characteristic, wherein the time window is a time length covering a complete seasonal period.

It will be appreciated that a range may be initially set based on seasonal and non-linear characteristics of the data, requiring more layers to capture these patterns if the data shows a stronger seasonal; if the data nonlinearities are strong, increasing the number of neurons per layer helps the model capture more complex relationships. The time circulation model is a model formed by long-term and short-term memory networks, and is called LSTM model for short.

The selection of a time window, which should cover at least one complete seasonal period based on seasonal features, if the data has a seasonal nature of day, the window is set to 24 hours, the ACF graph is observed based on an autocorrelation function to determine the furthest time step affecting the current consumption, if the ACF is still significant after 48 hours, which indicates that at least a 48 hour window is required, significant seasonal lengths in the data, e.g. 24, 48 or more hours, are determined and these are used as candidate windows, a model is trained for each window length, and its performance is evaluated by cross-validation, selecting a window length that performs better than the others.

S208: and determining a plurality of network configurations according to the neuron number range and the hierarchy number range, wherein one neuron number and one hierarchy number form one network configuration.

It will be appreciated that different network configurations are tested using a performance-based approach, such as grid search (english name GRID SEARCH), defining a range of neuron numbers and a range of layers, and then training and validating each configuration to select a model with the least validation error. For example, the number of neurons may range from 50 to 200, the number of layers may range from 1 to 3, one network configuration may be a number of neurons of 50, and the number of layers of 1.

S209: and training and verifying the network configurations to obtain the target network configuration with the minimum verification error.

It will be appreciated that the number of layers and the number of neurons of the LSTM network are dynamically adjusted based on seasonal and nonlinear characteristics of the data, for example, using deeper network structures during peak energy consumption in the winter and summer to improve learning and predictive capabilities of the model. On the basis of the adaptive selection, the number of LSTM layers and the number of neurons which are finally selected are defined. Training and verifying various network configurations, determining verification errors, and selecting the network configuration corresponding to the model with the minimum verification errors as a target network configuration through multiple configurations.

Optionally, the method for selecting the model parameter with the best output performance by adopting a grid search algorithm comprises the following steps:

Defining parameters and possible values to be tested, different numbers of layers and numbers of neurons per layer can be tested for the LSTM model. List of parameters to be tested: layer number example: layer 1, layer 2, layer 3; neuron number example: 50, 100, 150, 200. A model instance is created for each parameter combination, and a corresponding LSTM network is constructed for each combination of layer numbers and neuron numbers. Other parameters of the model, such as activation functions, optimizers, etc., are set. The models for each parameter combination are trained and evaluated, and cross-validation is used to evaluate the performance of each combination, ensuring the stability and reliability of the results. According to the evaluation result, the evaluation result may be, for example, the minimum mean square error, and the parameter combination excellent in performance may be selected. The configuration and performance index of the best model is recorded and reported. The model is retrained over the entire dataset using the selected optimal parameter combination. The final performance of the model is verified, ensuring that the model behaves similarly on future data as in cross-validation.

Wherein, the cross-validation specifically refers to: the dataset was divided into multiple parts (3 parts) and one part was used in turn as the test set and the other as the training set, so that multiple tests were performed on each combination to ensure the accuracy of the assessment.

S209: and training and optimizing the initial time circulation model according to the time window and the target network configuration, and determining a time circulation model.

It can be understood that after determining the time window and the network configuration, the length of the input sequence, that is, the size of the time window of the historical data which should be considered when the model predicts, is defined, and is matched with the periodic characteristics of energy consumption, the dropout rate is set to prevent overfitting, the ReLU or tanh is selected as an activation function, a RMSprop optimizer is adopted, the Mean Square Error (MSE) is used as a loss function, so as to form a preliminary LSTM model, and the performance of the LSTM model is evaluated by adopting a time sequence cross-validation method so as to simulate various situations in actual operation, and the learning rate is dynamically adjusted according to the performance of the model in the training process. The output of LSTM is combined with the application of statistical methods to enhance the smoothing and predictive accuracy of energy consumption trends.

The application statistical method comprises the following steps: moving average: calculating a moving average of LSTM output, selecting an average value of the latest N time steps as a final output, and being suitable for eliminating fluctuation in a period; and (3) exponential smoothing: if the data has more complex trend and seasonality, the data is smoothed by using an index, and the recent data is given higher weight, so that the data is more flexibly adapted to trend change; parameter optimization: parameters (such as smoothing coefficients) of the statistical method are adjusted according to historical data, an optimal smoothing effect is ensured, an LSTM model is integrated into a real-time data stream processing platform, an emergency is responded quickly, the data stream processing platform comprises APACHE KAFKA or Apache Storm, so that the model can receive and analyze the latest data in real time, and the emergency or obvious consumption change is responded timely.

Optionally, to ensure accurate transmission of data, a data stream interface is provided, and a data stream management system (APACHE KAFKA) is used as an intermediate layer of data transmission to collect and distribute energy consumption data in real time. The data stream is configured to access the model services, ensuring that the data can be transferred to the model processing unit in real time.

Optionally, the trained LSTM model is deployed as a service, and the API service is created using TensorFlow Serving so that it can receive real-time data and return the prediction results. Ensuring that the model can be loaded and process real-time data quickly, taking into account delay and processing power requirements. The input data is processed in real time, including standardization and format conversion, so as to adapt to the input requirement of the model, and necessary abnormal value detection and correction are implemented, so that the quality of the input data is ensured. And establishing a feedback loop between the model output and the energy management system, adjusting an energy use strategy based on model prediction, and adjusting and optimizing the model according to the actual energy use condition.

According to the intelligent analysis method for energy consumption based on the time sequence and the load characteristics, provided by the embodiment, seasonal characteristics and nonlinear characteristics of the second historical energy consumption data are obtained through screening and analysis processing of the first historical energy consumption data. And determining the neuron number range, the hierarchical number range and the time window of the initial time cycle model according to the seasonal characteristic and the nonlinear characteristic. And determining a plurality of network configurations according to the neuron number range and the hierarchical number range. Training and verifying the multiple network configurations to obtain the target network configuration with the minimum verification error. And training and optimizing the initial time circulation model according to the time window and the target network configuration, and determining the time circulation model. The model can receive and analyze the latest data in real time and respond to emergencies or obvious consumption changes in time.

Fig. 4 is a flowchart of an intelligent analysis method for energy consumption based on time series and load characteristics according to an embodiment of the present application. This embodiment is a detailed description of the factor analysis model training process in the intelligent analysis method for energy consumption based on time series and load characteristics based on the embodiment of fig. 2. As shown in fig. 4, the method includes:

S301: and acquiring first historical energy consumption data of the key equipment, load characteristic data corresponding to the first historical energy consumption data and a plurality of historical external factor data.

Step S301 is similar to step S201, and will not be described here.

S302: and inputting the first historical energy consumption data into a time circulation model to obtain an energy consumption mode output by the time circulation model.

It can be understood that the historical energy consumption data is input into a trained time circulation model, an early warning system is set in the time circulation model, when the model predicts an abnormal consumption mode, relevant personnel can be immediately notified or measures can be automatically taken, and the time circulation model not only can more accurately identify and predict the energy consumption mode, but also can effectively cope with irregular consumption behaviors and extreme events. In the actual application process, when abnormal energy consumption is detected, the abnormal energy consumption mode is input into the general analysis model, and external influence factors of abnormal energy consumption can be obtained, so that the adjustment of consumption strategies by related personnel is facilitated.

S303: and initializing the general analysis model to obtain the configured general analysis model.

It can be appreciated that the generic analytical model is first model-configured, the factor analyzer model is initialized, core parameters are configured, including factor dimensions, which determine the complexity of the interaction terms. The factor dimension is typically set to a lower dimension, e.g., 10-50. Meanwhile, super parameters such as learning rate, iteration times and the like are set, and the configuration of the general analysis model is completed by using a proper loss function.

S304: and carrying out normalization processing on the load characteristic data and the plurality of historical external factor data to obtain characteristic data.

It can be understood that the load characteristic data and the historical external factor data are divided into two types, namely, numerical data and category data, the numerical data are respectively processed according to the data categories, normalization processing is performed on the numerical data, single-heat code conversion processing is performed on the category data, and the processed data set forms characteristic data, and a classification method and a processing method are similar to step S101 and are not repeated here.

S305: and training the configured general analysis model according to the energy consumption mode and the characteristic data to obtain a trained factor analysis model, wherein the factor analysis model is used for indicating the association relation between the equipment energy consumption and external factors.

It will be appreciated that the energy consumption pattern identified by time series analysis output by the time-loop model, and the feature data are used as feature sets to train the configured generic analysis model, during which the model will learn not only the impact of a single feature on energy consumption, but also the impact of inter-feature interactions, applying cross-validation techniques to evaluate the performance of the model and prevent overfitting. The factor analysis machine can process the interaction among the features, analyze which factors and the interaction relationship thereof have obvious influence on the energy consumption data by using the model parameters and the output, and utilize the trained factor analysis model to conduct the energy consumption prediction and the influence analysis of external factors so as to provide decision support for energy management and optimization.

The trained factor analysis model comprises the correlation analysis result between the equipment energy consumption and the external factors, and the dot product of the factor vector is used for capturing interaction between the features, and the contribution of each factor to the energy consumption can be remarkably represented by using parameters of the model (comprising bias terms, weights of single features and weights of interaction between the features, which are obtained through historical data training).

Optionally, the factor analyzer is a machine learning algorithm that effectively processes a high-dimensional sparse data set and captures all interactions between variables, and the core idea is to approximate the interaction weights between each pair of features using a factorization technique, so that the factor analyzer is particularly suitable for a scenario containing a large number of category features, and in the correlation analysis between the device load characteristics and the external factors in the embodiment of the present application, the prediction model of the factor analyzer is expressed as the following formula:

wherein, Is the predictive output of the model, x is the feature vector, whereIs the value of the i-th feature,Is a global bias that is set to a value,Is the weight of the i-th feature,AndIs the factor vector of the i and j th features, whose dimension is k, representing the number of potential factors,Is thatAndFor modeling the interaction between feature i and feature j.

Bias and linearity terms: And Is a component of the traditional linear regression model and is responsible for capturing independent contributions of features;

Interaction item: Is the core of FM and can simulate the interaction effect between any two characteristics, and each characteristic i and j passes through respective factor vectors AndTo represent the dot product of these factor vectorsA value is provided that indicates how the two features jointly affect the target variable.

Compared with a model which needs to learn a weight for each pair of feature combinations independently, the factor analyzer greatly reduces the parameter quantity of the model through a low-rank matrix v mode, simultaneously maintains the capability of capturing complex interactions, can process any real-valued feature, and naturally expands to multidimensional and multi-modal data. When applying a factor analyzer to model the device load characteristics in relation to external factors, the emphasis is on selecting the appropriate factor dimension k and the appropriate optimization algorithm (random gradient descent) to train the model. Furthermore, the determination of the optimal hyper-parameters (k, learning rate) by cross-validation is critical to establishing an accurate and reliable predictive model.

According to the intelligent analysis method for energy consumption based on the time sequence and the load characteristics, the first historical energy consumption data of the key equipment, the load characteristic data corresponding to the first historical energy consumption data and the plurality of historical external factor data are obtained. And inputting the first historical energy consumption data into a time circulation model to obtain an energy consumption mode output by the time circulation model. And initializing the general analysis model to obtain the configured general analysis model. And carrying out normalization processing on the load characteristic data and the plurality of historical external factor data to obtain characteristic data. And training the configured general analysis model according to the energy consumption mode and the characteristic data to obtain a trained factor analysis model. The model effectively handles complex interactions between equipment energy consumption and various external factors, and allows adaptation to different types of data input and changing environmental conditions, thereby providing continuous and dynamic energy management optimization support.

Fig. 5 is a flow chart diagram of an intelligent analysis method for energy consumption based on time series and load characteristics according to an embodiment of the present application. This embodiment is a detailed process description of the implementation of the feedback collection system in the intelligent analysis method of energy consumption based on time series and load characteristics based on the embodiment of fig. 2. As shown in fig. 5, the method includes:

s401: and monitoring and acquiring actual energy consumption data in real time by adopting a feedback collection system.

It can be understood that the prediction model and the optimization strategy can be adjusted according to the actual energy consumption situation through the feedback mechanism, so that continuous energy management optimization is realized. First, a feedback collection system is established, automated systems and sensor technology are implemented to continuously monitor and record actual energy consumption data, including equipment usage, peak energy consumption periods, total energy consumption, and periodically collect user feedback regarding energy usage, including any reports regarding equipment performance or energy deficiency.

S402: and determining an energy deviation value according to the energy demand quantity corresponding to the actual energy consumption data and the equipment prediction load.

It will be appreciated that the prediction model performance may be assessed from actual energy consumption data, the prediction results compared with the actual energy consumption data periodically, the prediction accuracy and bias analyzed, the type and pattern of common errors in model predictions identified, including systematic or random errors, and the correlation of these errors with specific external conditions or modes of operation.

S403: and updating the model parameters of the factor analysis model according to the energy deviation value, and adjusting an energy distribution strategy and a consumption strategy.

It can be appreciated that the model parameters, including weights and factor vectors in the factor analyzer, are adjusted based on the error analysis results (energy bias values) to adjust and optimize the model to better reflect actual energy usage. And re-evaluating the existing energy distribution strategy and consumption strategy according to the result of the model adjustment and new energy use data, and dynamically adjusting the energy management strategy according to the latest model prediction and market or environment change, including modifying the peak operation strategy or adjusting the energy purchase plan.

According to the intelligent analysis method for energy consumption based on time sequence and load characteristics, the feedback collection system is adopted to monitor and acquire actual energy consumption data in real time. And determining an energy deviation value according to the energy demand quantity corresponding to the actual energy consumption data and the equipment prediction load. And updating the model parameters of the factor analysis model according to the energy deviation value, and adjusting an energy distribution strategy and a consumption strategy. The implementation of the method enables the energy management system to continuously learn and adapt, and changes of environment and operation conditions are dealt with by continuously adjusting the prediction model and the optimization strategy, so that long-term energy management optimization is maintained.

Fig. 6 is a schematic structural diagram of an intelligent analysis device for energy consumption based on time series and load characteristics. As shown in fig. 6, the intelligent analysis device 500 for energy consumption based on time series and load characteristics provided by the present application includes:

An obtaining module 501, configured to obtain at least one target external factor data affecting the device energy consumption in a period to be predicted, where the target external factor is determined according to an association relationship between the device energy consumption and the external factor;

a determining module 502, configured to determine a prediction feature vector corresponding to a period to be predicted according to the at least one target external factor data and a device usage mode of the key device;

The input module 503 is configured to input the prediction feature vector into a factor analysis model, to obtain a device predicted load of a period to be predicted and an energy demand corresponding to the device predicted load, where the factor analysis model is determined according to the first historical energy consumption data and the historical device load feature.

Optionally, the apparatus further includes: a training module 504;

The acquiring module 501 is further configured to acquire first historical energy consumption data of the key device, load characteristic data corresponding to the first historical energy consumption data, and a plurality of historical external factor data;

The input module 503 is further configured to input the first historical energy consumption data into a time cycle model, so as to obtain an energy consumption mode output by the time cycle model;

The training module 504 is configured to train a general analysis model according to the energy consumption mode, the load characteristic data, and a plurality of historical external factor data, to obtain the trained factor analysis model, where the factor analysis model is used to indicate an association relationship between equipment energy consumption and external factors.

Optionally, the apparatus further includes: a processing module 505;

the processing module 505 is configured to perform screening processing on the first historical energy consumption data to obtain second historical energy consumption data;

The processing module 505 is further configured to analyze the second historical energy consumption data to obtain seasonal features and nonlinear features of the second historical energy consumption data;

the determining module 502 is further configured to determine, according to the seasonal characteristic and the nonlinear characteristic, a neuron number range, a hierarchical number range, and a time window of an initial time cycle model, where the time window is a time length covering a complete seasonal period;

The determining module 502 is further configured to determine a plurality of network configurations according to the neuron number range and the hierarchy number range, where one neuron number and one hierarchy number form one network configuration;

The processing module 505 is further configured to perform training and verification processing on the plurality of network configurations, so as to obtain a target network configuration with a minimum verification error;

The determining module 502 is further configured to perform training optimization on the initial time loop model according to the time window and the target network configuration, and determine a time loop model.

Optionally, the apparatus further includes: a judgment module 506;

The processing module 505 is further configured to perform calculation processing on the first historical energy consumption data, and determine a standard score of each first historical energy consumption data;

the judging module 506 is configured to judge whether an absolute value of the standard score is greater than a preset value;

the determining module 502 is further configured to determine that the first historical energy consumption data corresponding to the standard score is an outlier if the absolute value of the standard score is greater than a preset value, and reject the outlier;

the determining module 502 is further configured to determine that the first historical energy consumption data corresponding to the standard score is the second historical energy consumption data if the absolute value of the standard score is not greater than a preset value.

Optionally, the processing module 505 is further configured to initialize the general analysis model to obtain a configured general analysis model;

the processing module 505 is further configured to normalize the load characteristic data and the plurality of historical external factor data to obtain feature data;

The processing module 505 is further configured to perform training processing on the configured general analysis model according to the energy consumption mode and the feature data, so as to obtain a trained factor analysis model.

Optionally, the processing module 505 is further configured to pre-process the at least one target external factor data to obtain processed target external factor data;

the determining module 502 is further configured to determine a prediction feature vector corresponding to the period to be predicted according to the processed target external factor data and the device usage mode.

Optionally, the acquiring module 501 is further configured to monitor and acquire actual energy consumption data in real time by using a feedback collecting system;

the determining module 502 is further configured to determine an energy deviation value according to an energy demand amount corresponding to the actual energy consumption data and the predicted load of the device;

the processing module 505 is further configured to update the model parameters of the factor analysis model according to the energy deviation value, and adjust an energy allocation policy and a consumption policy.

Fig. 7 is a schematic structural diagram of an intelligent analysis device for energy consumption based on time sequence and load characteristics. As shown in fig. 7, the present application provides an intelligent analysis device for energy consumption based on time series and load characteristics, the intelligent analysis device 600 for energy consumption based on time series and load characteristics comprising: a receiver 601, a transmitter 602, a processor 603 and a memory 604.

A receiver 601 for receiving instructions and data;

A transmitter 602 for transmitting instructions and data;

memory 604 for storing computer-executable instructions;

The processor 603 is configured to execute the computer-executable instructions stored in the memory 604, so as to implement the steps executed by the intelligent analysis method for energy consumption based on time sequence and load characteristics in the above embodiment. Reference may be made in particular to the description of the foregoing embodiments of the intelligent analysis method for energy consumption based on time series and load characteristics.

Alternatively, the memory 604 may be separate or integrated with the processor 603.

When the memory 604 is provided separately, the electronic device further comprises a bus for connecting the memory 604 and the processor 603.

The application also provides a computer readable storage medium, wherein the computer readable storage medium stores computer execution instructions, and when the processor executes the computer execution instructions, the energy consumption intelligent analysis method based on the time sequence and the load characteristic, which is executed by the energy consumption intelligent analysis equipment based on the time sequence and the load characteristic, is realized.

Those of ordinary skill in the art will appreciate that all or some of the steps, systems, functional modules/units in the apparatus, and methods disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. In a hardware implementation, the division between the functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be performed cooperatively by several physical components. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as known to those skilled in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by a computer. Furthermore, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.

Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.

It is to be understood that the application is not limited to the precise arrangements and instrumentalities shown in the drawings, which have been described above, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims

1. An intelligent analysis method for energy consumption based on time sequence and load characteristics, which is characterized by comprising the following steps:

2. The method of claim 1, wherein prior to the obtaining at least one target external factor data affecting device energy consumption during the period to be predicted, the method further comprises:

3. The method of claim 2, wherein before inputting the first historical energy consumption data into the time-loop model to obtain the energy consumption pattern output by the time-loop model, the method further comprises:

4. The method of claim 2, wherein the filtering the first historical energy consumption data to obtain second historical energy consumption data comprises:

5. The method of claim 2, wherein training the generic analytical model based on the energy consumption pattern, the load characteristic data, and a plurality of historical external factor data to obtain the trained factor analytical model comprises:

6. The method according to claim 1, wherein the determining a prediction feature vector corresponding to the period to be predicted according to the at least one target external factor data and the device usage pattern of the key device includes:

7. The method according to claim 1, wherein the method further comprises:

8. An intelligent energy consumption analysis device based on time series and load characteristics, characterized in that the device comprises:

9. An intelligent energy consumption analysis device based on time series and load characteristics, comprising:

a memory;

A processor;

wherein the memory stores computer-executable instructions;

the processor executes the computer-executable instructions stored by the memory to implement the intelligent analysis method for energy consumption based on time series and load characteristics as claimed in any one of claims 1 to 7.

10. A computer readable storage medium, wherein computer executable instructions are stored in the computer readable storage medium, which when executed by a processor is configured to implement the intelligent analysis method for energy consumption based on time series and load characteristics according to any one of claims 1-7.