CN113449904A

CN113449904A - Multi-energy load prediction method, device and equipment

Info

Publication number: CN113449904A
Application number: CN202110519142.9A
Authority: CN
Inventors: 杨海跃; 李国翊; 赵海洲; 武恺馨; 于潞; 高瑞超; 武志伟; 王洋
Original assignee: Tianjin University; State Grid Corp of China SGCC; State Grid Hebei Electric Power Co Ltd; Hengshui Power Supply Co of State Grid Hebei Electric Power Co Ltd
Current assignee: Tianjin University; State Grid Corp of China SGCC; State Grid Hebei Electric Power Co Ltd; Hengshui Power Supply Co of State Grid Hebei Electric Power Co Ltd
Priority date: 2021-05-12
Filing date: 2021-05-12
Publication date: 2021-09-28

Abstract

The invention discloses a method, a device and equipment for predicting a multi-energy load, wherein the method comprises the following steps: acquiring target energy load values of a target area at a plurality of historical moments; generating a trend curve and a fluctuation curve of each target energy source; extracting trend curve characteristics and fluctuation curve characteristics corresponding to a plurality of target moments from the trend curve and the fluctuation curve of each target energy; constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment, and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment; and predicting each target energy load value at the time to be predicted according to the constructed model and each target energy load value of a plurality of specified historical times corresponding to the target area at the time to be predicted, so as to obtain a prediction result. By constructing a trend load prediction model and a fluctuation load prediction model, the influence of the time-varying characteristic of the multi-energy system and the coupling between the energy sources are considered, and the prediction accuracy of load prediction is improved.

Description

Multi-energy load prediction method, device and equipment

Technical Field

The application belongs to the technical field of load prediction, and particularly relates to a multi-energy load prediction method, device and equipment.

Background

A Regional Integrated Energy System (RIES) integrates various heterogeneous Energy sources such as: electric energy, heat energy, cold energy and the like, and realizes diversified coordination planning, cooperative management and optimized operation of energy. With the popularization and development of the RIES, the load prediction is taken as an important premise of the optimization management of the energy system, and higher requirements are imposed on the accuracy, the real-time performance, the reliability and the intelligence of the load prediction.

In the prior art, a correlation analysis method, a neural network method and the like are generally adopted, load prediction is only carried out on a certain type of energy, and the prediction precision is low.

Disclosure of Invention

In view of this, the invention provides a method, a device and equipment for predicting a multi-energy load, and aims to solve the problem of low prediction accuracy of load prediction.

A first aspect of an embodiment of the present invention provides a method for predicting a multi-energy load, including:

acquiring target energy load values of a target area at a plurality of historical moments; the target energy sources are at least two;

generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve;

extracting trend curve characteristics corresponding to a plurality of target moments from the trend curve of each target energy, and extracting fluctuation curve characteristics corresponding to each target moment from the fluctuation curve corresponding to each target energy;

constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment, and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment;

and predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of specified historical times corresponding to the time to be predicted in the target area so as to obtain a prediction result.

A second aspect of an embodiment of the present invention provides a multi-energy load prediction apparatus, including:

the acquisition module is used for acquiring target energy load values of a target area at a plurality of historical moments; the target energy sources are at least two;

the preprocessing module is used for generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve;

the extraction module is used for extracting trend curve characteristics corresponding to a plurality of target moments from the trend curve of each target energy source and extracting fluctuation curve characteristics corresponding to each target moment from the fluctuation curve corresponding to each target energy source;

the training module is used for constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment;

and the prediction module is used for predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of specified historical times corresponding to the time to be predicted of the target area so as to obtain a prediction result.

A third aspect of the embodiments of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of the multi-energy load prediction method according to the first aspect when executing the computer program.

A fourth aspect of embodiments of the present invention provides a computer-readable storage medium, which stores a computer program that, when executed by a processor, implements the steps of the multi-energy load prediction method according to the first aspect.

Compared with the prior art, the invention has the following beneficial effects:

according to the embodiment of the invention, the target energy load values of the target area at a plurality of historical moments are obtained; generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve; extracting trend curve characteristics corresponding to a plurality of target moments from the trend curve of each target energy, and extracting fluctuation curve characteristics corresponding to each target moment from the fluctuation curve corresponding to each target energy; constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment, and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment; and predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of specified historical times corresponding to the time to be predicted in the target area so as to obtain a prediction result. By constructing a trend load prediction model and a fluctuation load prediction model of each target energy, the influence of time-varying characteristics of prediction on a multi-energy system and the coupling between each energy are considered, and the prediction precision of load prediction is improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.

Fig. 1 is an application scenario diagram of a multi-energy load prediction method according to an embodiment of the present invention;

FIG. 2 is a flowchart illustrating an implementation of a method for multi-energy load prediction according to an embodiment of the present invention;

FIG. 3 is a flowchart illustrating an implementation of a method for multi-energy load prediction according to another embodiment of the present invention;

FIG. 4 is a graph of cold, hot, electrical raw load curves, ripple curves, and trend curves provided by an embodiment of the present invention;

FIG. 5 is a comparison of predicted curves and actual curves for the prediction of cold, heat, and electrical loads on a test set for four models, M1-M4, spanning a week in time;

FIG. 6 is an enlarged view of the rectangular frame portion of FIG. 5;

FIG. 7 is a plot of root mean square error versus average absolute percent error for cold load predictions based on the M1-M4 model;

FIG. 8 is a plot of the root mean square error versus the mean absolute percent error for predictions of thermal loads based on the M1-M4 model;

FIG. 9 is a plot of the root mean square error versus the mean absolute percentage error for predictions made for electrical loads based on the M1-M4 model;

FIG. 10 is a graph of the relative error percentage distribution box for cold, hot, and electrical loads based on the M1-M4 model;

FIG. 11 is a schematic structural diagram of a multi-energy load prediction apparatus according to an embodiment of the present invention;

fig. 12 is a schematic diagram of an electronic device provided by an embodiment of the invention.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.

In order to make the objects, technical solutions and advantages of the present invention more apparent, the following description is made by way of specific embodiments with reference to the accompanying drawings.

The multi-energy load prediction refers to load prediction of a power system (such as an RIES) powered by various energy sources, compared with the traditional single energy system, the RIES integrates energy supply, energy conversion and energy storage devices in various forms, coupling performance of different energy sources is improved in different links such as sources, networks and loads, flexibility of energy utilization of the whole system is improved, and the multi-energy load prediction is an important component of a new-generation energy system for constructing a clean, low-carbon, safe and efficient modern energy system. Therefore, the accurate multi-energy load prediction model can provide more accurate prediction information for multi-energy planning and scheduling of the RIES, and has important practical significance and economic value.

In the prior art, a correlation analysis method, a neural network method and the like are generally adopted, load prediction is only carried out on a certain type of energy, and the prediction precision is low. However, in the prior art, when multi-energy load prediction is performed on a multi-energy system, only the time correlation of the whole historical load is generally considered, and the prediction precision is low.

The multi-energy load prediction method provided by the invention improves the prediction accuracy of load prediction based on the long-term characteristic and short-term characteristic influence of the influence factors on the multi-energy load, the multi-energy change characteristic influence and the multi-energy coupling of the change characteristic under the condition of comprehensive energy physical interconnection and information interaction.

Fig. 1 is an application scenario diagram of a multi-energy load prediction method according to an embodiment of the present invention. The multi-energy load prediction method provided by the embodiment of the invention can be applied to the application environment but is not limited to the application environment. As shown in fig. 1, the application environment includes: power data acquisition equipment 11, electronic equipment 12 and dispatch center 13.

The dispatch center 13 is used to send prediction instructions to the electronic device 12. The electronic device 12 is configured to send a collection instruction to the power data collection device 11 in the target area after receiving the prediction instruction. The power data acquisition device 11 is configured to acquire target energy load values at a plurality of designated historical times of the target area after receiving the acquisition instruction, and send the load values to the electronic device 12. The electronic device 12 is further configured to predict each target energy load value at a time to be predicted after receiving each target energy load value, and send an obtained prediction result to the scheduling center 13.

The power data collection device 11 may be an electromechanical integrated electric meter, an all-electronic electric meter, etc., and is not limited herein. The electronic device 12 may be a server, a terminal, etc., and is not limited thereto. The server may be implemented as a stand-alone server or as a server cluster comprised of multiple servers. The terminal may include, but is not limited to, a desktop computer, a laptop computer, a tablet computer, and the like. The power data acquisition device 11, the electronic device 12, and the dispatching center 13 may perform data interaction through a line, or may perform data interaction through a network or a bluetooth, which is not limited herein. The electronic device 12 may be a device installed independently, or may be a device installed in the scheduling center 13, and is not limited herein.

Fig. 2 is a flowchart illustrating an implementation of a method for predicting a multi-energy load according to an embodiment of the present invention. In this embodiment, the method is applied to the electronic device in fig. 1 as an example. As shown in fig. 2, the method includes:

s201, acquiring target energy load values of a target area at a plurality of historical moments; the target energy sources are at least two.

In the present embodiment, the historical time may be in units of hours and minutes, and is not limited herein. For example, all the hour points in a certain target area are selected, or a preset hour point is selected, which is not limited herein.

S202, generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve.

In this embodiment, each target energy corresponds to one trend curve and one fluctuation curve. The multi-energy load is affected by the long-term and short-term characteristics of the influencing factors. The long-term characteristics such as regional characteristics, periodic climate change, popular habits and the like cause the load to have trend characteristics. Short-term characteristics such as changes in current weather conditions, human life activities, etc., cause the load to have short-term volatility characteristics. The trend can effectively represent the variation trend of the total load curve, and the fluctuation can highlight the load fluctuation condition caused by accidental factors. When multi-energy load prediction is carried out, the variation characteristics are considered, the variation morphological characteristics of the multi-energy sources can be effectively mastered, the prediction accuracy is improved, the correlation information of each energy source under the variation characteristics can be further researched, and the multi-energy coupling characteristic learning is enhanced.

And S203, extracting trend curve characteristics corresponding to a plurality of target moments from the trend curve of each target energy source, and extracting fluctuation curve characteristics corresponding to each target moment from the fluctuation curve corresponding to each target energy source.

In this embodiment, the target time may be selected from historical times. The time intervals of the respective target time instants are the same.

And S204, constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment, and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment.

In this embodiment, a plurality of designated historical moments are selected according to a certain step length before the target moment, and then a trend curve feature and a fluctuation curve feature corresponding to each target moment are generated. And (3) taking the trend curve characteristics corresponding to the target moments as input samples for training, taking the load components of the target moments as real values of the samples, and training a trend load prediction model. And taking the fluctuation curve characteristics corresponding to a plurality of target moments as input samples for training, taking the load components of the target moments as real values of the samples, and training a fluctuation load prediction model.

And S205, predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of designated historical times corresponding to the time to be predicted of the target area to obtain a prediction result.

In this embodiment, the plurality of designated historical moments may be moments selected according to a certain step length before the moment to be predicted. For example, the time to be predicted may be a time 1 hour after the current time, and the specified historical time may be a time 1 hour before the current time, or a time 2 hours before the current time. During prediction, a trend curve and a fluctuation curve of each target energy source can be generated according to the load value of each target energy source at a plurality of designated historical moments, then a trend curve characteristic corresponding to the moment to be predicted is extracted from the trend curve of each target energy source, a fluctuation curve characteristic corresponding to the moment to be predicted is extracted from the fluctuation curve corresponding to each target energy source, and the trend curve characteristic and the fluctuation curve characteristic corresponding to the moment to be predicted are input into corresponding models, so that a prediction result is obtained.

In the embodiment, each target energy load value of a target area at a plurality of historical moments is obtained; generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve; extracting trend curve characteristics corresponding to a plurality of target moments from the trend curve of each target energy, and extracting fluctuation curve characteristics corresponding to each target moment from the fluctuation curve corresponding to each target energy; constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment, and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment; and predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of specified historical times corresponding to the time to be predicted in the target area so as to obtain a prediction result. By constructing a trend load prediction model and a fluctuation load prediction model of each target energy, the influence of time-varying characteristics of prediction on a multi-energy system and the coupling between each energy are considered, and the prediction precision of load prediction is improved.

Fig. 3 is a flowchart of an implementation of a method for predicting a multi-energy load according to another embodiment of the present invention. In some embodiments, based on the embodiment shown in fig. 2, as shown in fig. 3, constructing a trend load prediction model based on the trend curve characteristics corresponding to each target time, and constructing a fluctuating load prediction model based on the fluctuating curve characteristics at each target time respectively may include:

aiming at each target energy, establishing a trend load prediction model of the target energy according to a long-term and short-term memory algorithm;

training a trend load prediction model of each target energy source based on the trend curve characteristics corresponding to each target moment;

aiming at each target energy, establishing a fluctuating load prediction model of the target energy according to a least square support vector regression algorithm;

and respectively training the fluctuating load prediction model of each target energy source based on the fluctuating curve characteristics corresponding to each target moment.

In this embodiment, for each target energy, a single-layer LSTM structure may be adopted according to a Long Short-Term Memory algorithm (LSTM), so as to establish a Long Short-Term Memory neural network model, and the model is used as a trend load prediction model of the target energy. The number of each energy source neuron is (64, 64, 64), the activation function is Relu, and the optimizer is Adam. The cell unit of the LSTM model comprises a forgetting gate, an input gate and an output gate, and long-term dependence is realized through a gate control mechanism. The change of each state of the cell unit can be realized by the following formula:

f_t＝σ(W_fhh_t-1+W_fxx_t+b_f) (1)

i_t＝σ(W_ihh_t-1+W_ixx_t+b_i) (2)

o_t＝σ(W_ohh_t-1+W_oxx_t+b_o) (3)

h_t＝o_t*tanh(C_t) (6)

wherein, W_fh，W_ih，W_oh，W_ch，W_fx，W_ix，W_ox，W_cxIs a weight matrix; b_f，b_i，b_o，b_cIs a bias term; sigma is sigmoid function; x is the number of_tAn input sequence of the current time step t; f. of_t，i_t，o_tOutput vectors of the forgetting gate, the input gate and the output gate are respectively; c_tIs the cell state at the current time step t; h is_tIs the output of the final hidden layer.

In this embodiment, because the frequency of the time series of the fluctuation curve fluctuates greatly and the long-term dependence on time is not obvious, a Least Square Support Vector Regression (LS-SVR) algorithm is used to establish a fluctuation load prediction model of the target energy for each target energy. Compared with a heuristic mode of a neural network and empirical components in implementation, the LSSVR has stricter theoretical and mathematical basis, no local minimum exists, and the generalization performance is stronger, so that the LSSVR algorithm is more suitable for the prediction of a fluctuation curve and has better prediction performance. In this embodiment, a gaussian kernel Function (RBF) is selected for use.

In the embodiment, algorithms are respectively selected for processing the trend curve characteristics and the fluctuation curve characteristics to establish different models, so that the prediction error caused by the variation characteristics can be effectively reduced, and the prediction precision of load prediction is improved.

Further, training a trend load prediction model of each target energy source based on the trend curve characteristics corresponding to each target time includes:

and training a trend load prediction model of each target energy source based on a multi-task learning algorithm of a hard parameter sharing mechanism and trend curve characteristics corresponding to each target moment.

In this embodiment, a Multi-Task Learning algorithm (MTL) of a hard parameter sharing mechanism is adopted, a plurality of loss functions are weighed by considering the homogeneity uncertainty of each Task, a concept of an induced migration mechanism is introduced, a plurality of tasks, all or a part of which are related but not identical, are processed, shared information included in the plurality of tasks is fully utilized, and the plurality of tasks are trained in parallel, so that the goal of improving the performance of each Task is achieved. The multi-task learning can improve the learning efficiency and the application performance of each task and can also reduce the scale of the model to a great extent. A hard parameter sharing mechanism in the MTL parameter sharing mechanism is adopted, namely one or more hidden layers which are common in a network are shared among a plurality of tasks, and meanwhile output layers of the tasks related to problems are reserved, so that the efficiency is improved, and the risk of overfitting is reduced.

In some embodiments, based on any of the above embodiments, a variation characteristic curve of each target energy is generated according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve, comprising:

for each target energy, forming a load time sequence of the target energy according to the load value of the target energy, and decomposing the load time sequence of the target energy according to a pole symmetric modal decomposition algorithm to obtain a plurality of modal component sequences;

and aiming at each target energy, calculating a load time sequence of the target energy and a sample entropy value of each modal component sequence, deleting the modal component sequences of which the sample entropy values relative to the load time sequence exceed a preset threshold value in all the modal component sequences, merging and reconstructing the modal component sequences meeting preset conditions in the sample entropy values into a fluctuation curve, and merging and reconstructing the other modal component sequences into a trend curve.

In the embodiment, the influence of the variation characteristic on load prediction is considered, and a trend curve and a fluctuation curve of each energy load characteristic are constructed by a pole symmetric modal decomposition-sample entropy algorithm.

The pole Symmetric Mode Decomposition algorithm (ESMD) is an improvement on the basis of an Empirical Mode Decomposition (EMD) Method. The EMD algorithm is a data self-adaptive analysis method, is suitable for the analysis of nonlinear non-stationary signal sequences, and is used for carrying out stabilization processing on complex signals and extracting modal component Sequences (IMFs) with different characteristic scales or inherent periods and trend remainder R₀. However, the decomposed trend function is too rough, the screening times are difficult to determine, and the obtained modal component sequence contains the inherent mode and the abnormal event or contains the components of adjacent characteristic time scales, so that different modal components cannot be effectively separated according to the time characteristic scales, and modal aliasing is caused. The ESMD algorithm replaces the outer envelope interpolation of the EMD algorithm with the internal pole symmetric interpolation, optimizes the residual component by the principle of least square to enable the residual component to become the optimal adaptive global mean line of the whole data sequence, and can better reflect the change trend of data, thereby determining the optimal screening times and solving the problem of modal aliasing of the EMD algorithm. Because the ESMD algorithm has better self-adaptability and is based on the local variability of signals, the ESMD algorithm has good processing capacity aiming at the non-stable, non-linear and periodic random sequence of the load time sequence, so the ESMD algorithm is selected as the basis, and the load time sequence of each energy source is decomposed according to the frequency characteristic to obtain the modal component sequence IMF of each energy source characteristic_iAnd trend remainder R₀。

This implementationIn one example, the loading time series of each target energy and the corresponding modal component series IMF of each energy feature are used_iAnd trend remainder R₀And performing Sample Entropy (SE) calculation to quantitatively describe the complexity and the regularity of the system load from the view point of time series complexity. And comprehensively analyzing the SE value, and recombining all components of the ESMD decomposition into a variation characteristic curve comprising a trend curve and a fluctuation curve.

In this embodiment, the preset threshold may be a preset fixed value, or may also be a preset coefficient multiplied by a sample entropy value of the load time series of each target energy, for example, the preset coefficient is 1, which is not limited herein. And discarding the modal component sequence with the sample entropy value of the relative load time sequence exceeding the preset threshold in the modal component sequence as the noise part of the sequence, so as to avoid larger errors. The preset condition may be a fixed value larger than a preset value, or may be a sample entropy value larger than a load time series of each target energy source, which is not limited herein.

Optionally, the modal component sequence may be recombined into a plurality of different curves by analyzing the sample entropy according to the variation characteristic of the load time sequence of each target energy source. In this embodiment, the number and type of the recombined curves may be determined according to the actual demand of load prediction, which is not limited herein.

Further, extracting trend curve features corresponding to a plurality of target moments from the trend curve of each target energy, and extracting fluctuation curve features corresponding to each target moment from the fluctuation curve corresponding to each target energy, includes:

determining a trend data sequence of the target time of each target energy source according to the trend curve of each target energy source, and determining fluctuation data sequences of various candidate times of each target energy source according to the fluctuation curve of each target energy source;

calculating the correlation coefficient between the fluctuation data sequence of each type of candidate time of the target energy and the fluctuation data sequence of the target time, the correlation coefficient between the trend data sequence of each type of candidate time of the target energy and the trend data sequence of the target time, the fluctuation data sequence of each type of candidate time of the target energy and the fluctuation data sequences of the target time of the rest of target energy, and the correlation coefficient between the trend data sequence of each type of candidate time of the target energy and the trend data sequence of the target time of the rest of target energy for each type of target energy; selecting candidate moments with the correlation meeting preset correlation conditions from various candidate moments as specified historical moments corresponding to the target moments according to the correlation coefficients;

for each target energy, extracting the load component of the appointed historical moment corresponding to each target moment from the trend curve of the target energy, and forming the trend curve characteristic of the target energy corresponding to each target moment; and extracting the load component of the appointed historical time corresponding to each target time from the fluctuation curve of the target energy, and forming the fluctuation curve characteristic of the target energy corresponding to each target time.

In this embodiment, the correlation coefficient may be a pearson correlation coefficient, and the correlation satisfying the preset correlation condition may be that all or most of the correlation coefficients corresponding to each target energy are greater than a preset value. The designated historical time corresponding to the target time is before the target time.

In the embodiment, by calculating the correlation coefficient, a data sequence with strong correlation can be selected as an input sequence during model training, so that the multi-functional coupling of the trend curve and the fluctuation curve can be measured, and the prediction accuracy of load prediction is improved.

Further, determining a trend data sequence of the target time of each target energy source according to the trend curve of each target energy source, and determining a fluctuation data sequence of various candidate times of each target energy source according to the fluctuation curve of each target energy source, includes:

selecting a plurality of target moments according to a first preset time step, and aiming at each target energy, forming the load component of each target moment on the trend curve of the target energy into a trend data sequence of the target moment, and forming the load component of each target moment on the fluctuation curve of the target energy into a fluctuation data sequence of the target moment;

aiming at each target moment, selecting a plurality of candidate moments before the target moment according to a second preset time step;

and dividing the candidate moments with the same step size relative to the target moment into one class, aiming at each class of candidate moments, forming the load components of the class of candidate moments on the trend curve of the target energy into a trend data sequence of the class of candidate moments, and forming the load components of the class of candidate moments on the fluctuation curve of the target energy into a fluctuation data sequence of the class of candidate moments.

In this embodiment, the first preset time step and the second preset time step may be time lengths set arbitrarily. Preferably, the first preset time step may be set to one day, and the second preset time step may be set to one hour.

In some embodiments, on the basis of any of the above embodiments, before the constructing the trend load prediction model based on the trend curve characteristics corresponding to each target time, and the constructing the fluctuating load prediction models based on the fluctuating curve characteristics at each target time, further includes:

acquiring meteorological data of each target moment of a target area to obtain meteorological characteristics; and adding the meteorological features into the trend curve features and the fluctuation curve features corresponding to each target moment.

Optionally, the meteorological data may include at least one of: dry bulb temperature, wet bulb temperature, and relative humidity. Meteorological data may also include wind speed, irradiance, barometric pressure.

In this embodiment, the meteorological data of each target time of the target area may be acquired from the network. By adding meteorological features, the influence of meteorological changes on load prediction is considered, and the prediction accuracy of the load prediction is improved.

In some embodiments, on the basis of any of the above embodiments, after the trend load prediction model is constructed based on the trend curve characteristics corresponding to each target time, and the fluctuating load prediction models are respectively constructed based on the fluctuating curve characteristics at each target time, the method further includes:

performing model precision evaluation on the trend load prediction model and the fluctuation load prediction model of each target energy according to preset indexes;

if the evaluation result is qualified, predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of appointed historical times corresponding to the target area at the time to be predicted so as to obtain a prediction result;

otherwise, continuing to train the trend load prediction model and the fluctuation load prediction model of each target energy.

In this embodiment, a set of load components at historical times is input to the trend load prediction model and the fluctuating load prediction model of each target energy, output values of all the models are obtained, and then one-to-one weighted reconstruction is performed on the output values, so as to obtain predicted load values. And determining the numerical value of the preset index according to the predicted load value and the actual load value of the moment to be predicted corresponding to the group of historical moments. And if the numerical value of the preset index is greater than or equal to the preset qualified index value, the evaluation result is qualified. The value of the preset index may be determined by the following formula:

wherein epsilon_RMSEIs a root mean square error index value, epsilon_MAPEIs the mean absolute percentage error index value, ε_REIs a relative error value, n is the number of the input historical time, y_iIs the actual load value at the moment to be predicted,

is a predicted load value. The preset indicators may also include mean square error, mean absolute error, symmetric mean absolute percentage error.

In this embodiment, by evaluating the trained model, a model with low prediction accuracy can be removed, and the prediction accuracy of load prediction can be effectively improved.

In some embodiments, on the basis of any one of the above embodiments, predicting, according to the trend load prediction model, the fluctuating load prediction model, and each target energy load value of the target area at a plurality of specified historical times corresponding to the time to be predicted, each target energy load value at the time to be predicted to obtain a prediction result, includes:

predicting each target energy load component at the moment to be predicted according to the trend load prediction model and each target energy load value of a plurality of designated historical moments corresponding to the target area at the moment to be predicted so as to obtain a trend prediction result of each target energy;

predicting each target energy load component at the moment to be predicted according to the fluctuating load prediction model and each target energy load value of a plurality of specified historical moments corresponding to the target area at the moment to be predicted so as to obtain a trend prediction result of each target energy;

and performing one-to-one weighted reconstruction on the trend prediction result of each target energy and the trend prediction result of each target energy to obtain the prediction result of each target energy.

In this embodiment, after the target energy load values at the multiple designated historical times corresponding to the time to be predicted need to be decomposed and recombined, the obtained load components of the target energy at the multiple designated historical times are input to the corresponding model as input data to predict the target energy load components at the time to be predicted.

In some embodiments, on the basis of any of the above embodiments, obtaining each target energy load value of the target area at a plurality of historical time instants includes:

and acquiring and normalizing target energy load values of the target area at a plurality of historical moments.

In the embodiment, the acquired load value is subjected to normalization processing, so that the efficiency of a prediction algorithm can be improved, single data is prevented from overflowing in the calculation process, and the prediction accuracy of load prediction is improved.

Optionally, the specified historical time corresponding to each target time may include at least one of the following: the first 3 hours, the first 2 hours, and the first 1 hour of the target time. Optionally, the target energy source may comprise at least one of: electric energy, heat energy, cold energy.

The above-described multi-energy load prediction method will be described below by way of an example of embodiment, but is not limited thereto. In this embodiment, the selected target area occupies 160 ten thousand square meters, including more than 160 buildings. The time span of the training set is selected from 2011, 9, 10, 1 to 2012, 3, 9, 24, and the time span of the test set is selected from 2012, 3, 10, 1 to 2012, 3, 16, 24.

Step 1, acquiring target energy load values of a plurality of moments in a training set by taking an hour as each target energy load value of a plurality of historical moments, and normalizing the target energy load values. The target energy sources are electric energy, heat energy and cold energy.

Step 2, generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve. Fig. 4 is a graph of cold, hot, electrical raw load curves, ripple curves, and trend curves provided by an embodiment of the present invention. Step 2 may include:

and 2.1, aiming at each target energy, forming a load time sequence of the target energy according to the load value of the target energy, and decomposing the load time sequence of the target energy according to a pole symmetric modal decomposition algorithm to obtain a plurality of modal component sequences.

And 2.2, calculating the load time sequence of each target energy source and the sample entropy value of each modal component sequence aiming at each target energy source. The calculation results are shown in table 1.

Table a load time series of each target energy source and a sample entropy value table of each modal component series

As can be seen from Table 1, the IMF of each target energy source₁The sample entropy of the sequence is too large and far exceeds the original load characteristic sequence, so that the sample entropy is regarded as the noise part of the sequence to be abandoned, and larger errors are avoided. IMF is calculated according to the entropy of each energy sample₂And IMF₃And merging and reconstructing the sequences into a fluctuation curve, and merging and reconstructing the rest component sequences into a trend curve. The reconstructed curve is shown in fig. 4.

And 3, extracting trend curve characteristics corresponding to a plurality of target moments from the trend curve of each target energy source, and extracting fluctuation curve characteristics corresponding to each target moment from the fluctuation curve corresponding to each target energy source. Step 3 may include:

and 3.1, setting the first preset time step as t and the second preset time step as one hour.

And selecting the candidate time of the previous 3 hours, the previous 2 hours and the previous 1 hour of the target time as the candidate time for each target time, namely if the target time is t, the candidate time is t-3, t-2 and t-1. The sequence of the load values of the cold, hot and electric energy sources at all the candidate moments is as follows: LF ═ L_c,t-3，L_h,t-3，L_e,t-3，L_c,t-2，L_h,t-2，L_e,t-2，L_c,t-1，L_h,t-1，L_e,t-1Units are cold tons (ton), pound mass per hour (lbm/hr), and Megawatts (MW), respectively. Wherein L represents the respective energy load value, c, h, e represent the cooling load, the heating load and the electrical load, respectively, for example L_c,t-1Representing the cooling load at time t-1.

And 3.2, dividing the candidate moments with the same step size relative to the target moment into one class, obtaining three classes of candidate moments in total, and determining a trend data sequence of the target moment of each target energy source and a fluctuation data sequence of the candidate moments of each target energy source. And calculating the correlation coefficient between the fluctuation data sequence of each type of candidate time of the target energy and the fluctuation data sequence of the target time, the correlation coefficient between the trend data sequence of each type of candidate time of the target energy and the trend data sequence of the target time, the fluctuation data sequence of each type of candidate time of the target energy and the fluctuation data sequences of the target time of the rest of target energy, and the correlation coefficient between the trend data sequence of each type of candidate time of the target energy and the trend data sequence of the target time of the rest of target energy. The calculation results are shown in table 2.

TABLE 2 Pearson correlation coefficient value Table

As can be seen from table 2, in the trend curve, for the load at the candidate time and the target time, the correlation coefficient of each energy source is not less than 0.98, the correlation coefficient between the thermal load and the electrical load is not less than 0.80, and the correlation coefficient between the remaining energy sources is not less than 0.88. It can be seen that in the trend curve, the correlation between the energy sources is very strong. In the fluctuation curve portion, the correlation coefficient between each energy source itself or energy source at the candidate time and the target time gradually increases, and the closer to the target time, the stronger the time correlation thereof. Except that the correlation of the heat load at the historical moment to the cold and electric loads at the current moment is weak, the correlation of the heat load at the historical moment to the electric loads is close to or greater than 0.4, and the correlation is basically achieved or greater than a medium degree. Thus, all the above candidate times are taken as the designated history times corresponding to the target time.

And 3.3, taking the load components of the target energy sources at the designated historical moments corresponding to the target moments on the trend curves as trend curve features corresponding to each target moment, and taking the load components of the target energy sources at the designated historical moments corresponding to the target moments on the fluctuation curves as fluctuation curve features corresponding to each target moment.

Step 4, obtainingObtaining meteorological data of each target time in a target area to obtain a meteorological feature matrix MF ═ DT_t，WT_t，H_t}. Wherein, DT_tFor dry bulb temperature, WT_tIs the wet bulb temperature, H_tIs a relative humidity value.

Adding the meteorological features into the trend curve features and the fluctuation curve features corresponding to each target moment; respectively obtaining the fused trend curve characteristics TF ═ TL_c,t-3，TL_h,t-3，TL_e,t-3，TL_c,t-2，TL_h,t-2，TL_e,t-2，TL_c,t-1，TL_h,t-1，TL_e,t-1，DT_t，WT_t，H_tAnd the fused fluctuation curve characteristic FF ═ FL_c,t-3，FL_h,t-3，FL_e,t-3，FL_c,t-2，FL_h,t-2，FL_e,t-2，FL_c,t-1，FL_h,t-1，FL_e,t-1，DT_t，WT_t，H_t}。

And 5, constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment, and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment. Step 5 may include:

and 5.1, adding multi-task learning into the LSTM model to obtain an MTL-LSTM model, inputting the fused trend curve characteristic TF serving as a time sequence sample into the MTL-LSTM model for model training and testing, and optimizing through an Adam optimizer.

And 5.2, inputting the fused fluctuation curve characteristic FF serving as a time sequence sample into an LSSVR model for model training and testing.

Step 6, performing model precision evaluation on the trend load prediction model and the fluctuation load prediction model of each target energy according to the root mean square error index, the average absolute percentage error index and the relative error; if the evaluation result is qualified, executing the step 7; otherwise, continuing to train the trend load prediction model and the fluctuation load prediction model of each target energy.

Step 7, predicting each target energy load component at the moment to be predicted according to the trend load prediction model and each target energy load value of the target area 3 hours, 2 hours and 1 hour before the moment to be predicted so as to obtain a trend prediction result of each target energy;

predicting each target energy load component at the moment to be predicted according to the fluctuating load prediction model and each target energy load value of the target area 3 hours, 2 hours and 1 hour before the moment to be predicted so as to obtain a trend prediction result of each target energy;

In testing and analyzing the experiment of the invention, the test and verification of the model prediction effect are carried out on the selected test set data by training the selected training set data. Setting M1 as a basic LSTM model, M2 as a basic LSSVR model, M3 as a prediction model for multi-energy load prediction using a pole symmetric modal decomposition algorithm, sample entropy, long-short term memory algorithm, and least square vector regression algorithm, that is, an ESMD-SE-LSTM-LSSVR model, provided in an embodiment of the present invention, and M4 as a prediction model for multi-energy load prediction using a pole symmetric modal decomposition algorithm, sample entropy, multi-task learning-long-short term memory algorithm, and least square vector regression algorithm, that is, a MELF _ MECVCC model, provided in an embodiment of the present invention.

FIG. 5 is a comparison of predicted curves and actual curves for the prediction of cold, heat, and electrical loads on a test set for four models, M1-M4, spanning a week in time. Fig. 6 is an enlarged view of a rectangular frame portion in fig. 5. The time span of fig. 6 is one day, and the specific time is 3 months and 13 days 2012.

As shown in fig. 5, the fitting effect of the four groups of models to the actual load curve is better, and the following description is specifically given with reference to fig. 6 of subsequence samples thereof. As shown in fig. 6, the fitting of M3 and M4 to the actual load curves of cold, heat and electricity is relatively better than that of M1 and M2, the M1 model is too gentle, the fitting to the fluctuation condition of the actual load curve is relatively poor, the M2 model has relatively large fluctuation amplitude, for example, at the moment of 7:00, the electric load predicted value fluctuates greatly, but it is also obvious from an overall perspective that LSSVR has a good generalization characteristic to sequences with relatively large fluctuation. And as can be clearly seen from the figure, when the actual load fluctuates, the fitting of the actual load curve by the M1 and the M2 has a hysteresis phenomenon compared with the fitting of the M3 and the M4, and the fitting effect of the prediction curve is improved by the M3 and the M4 models through consideration of the multipotential coupling of the load variation characteristic.

FIG. 7 is a plot of root mean square error versus average absolute percent error for cold load predictions based on the M1-M4 model. FIG. 8 is a plot of the root mean square error versus the average absolute percent error for predictions of thermal load based on the M1-M4 model. FIG. 9 is a plot of the root mean square error versus the average absolute percentage error for predictions made for electrical loads based on the M1-M4 model.

As shown in fig. 7-9, the numbers at the top of the rectangles in the figures indicate the corresponding load errors. It can be shown that for the predicted root mean square error indicator RMSE and the mean absolute percent error MAPE, the M1-M4 prediction error goes from large to small as M1> M2> M3> M4. And the MAPE of M3 reduced cold by 25.94%, 20.24%, heat by 27.75%, 14.81%, and electricity by 24.14%, 6.38% compared to M1 and M2. In the M4, on the basis that M3 considers the multipotency coupling of the variation characteristic curve, after a multitask learning method is introduced into the trend curve part, the cold and heat are respectively reduced by 4.06%, 23.19% and 6.06% compared with M3. M4 makes full use of the shared information in the model, the prediction accuracy is further improved, the error value is the smallest in the four models, and the fitting degree of the actual load curve is the best.

FIG. 10 is a graph of the relative error percentage distribution box for cold, hot, and electrical loads based on the M1-M4 model. As shown in fig. 10, the relative error between M3 and M4 is smaller than that between M1 and M2, and the error distribution range is significantly smaller, which indicates that the relative error of the prediction model can be reduced by considering the multi-energy coupling of the variation characteristic curve, and the relative error of M4 relative to M3 is smaller, so the M4 model has a certain effect on improving the multi-energy load prediction accuracy, and the prediction accuracy is higher.

It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

Fig. 11 is a schematic structural diagram of a multi-energy load prediction apparatus according to an embodiment of the present invention. As shown in fig. 11, the multi-energy load prediction apparatus 11 includes:

an obtaining module 1110, configured to obtain target energy load values of a target area at a plurality of historical moments; the target energy sources are at least two.

A preprocessing module 1120, configured to generate a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve.

The extracting module 1130 is configured to extract trend curve features corresponding to a plurality of target moments from the trend curves of the target energy sources, and extract a fluctuation curve feature corresponding to each target moment from the fluctuation curves corresponding to the target energy sources.

And the training module 1140 is used for constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment, and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment.

The prediction module 1150 is configured to predict, according to the trend load prediction model, the fluctuating load prediction model, and each target energy load value of a plurality of designated historical times corresponding to the target area at the time to be predicted, each target energy load value at the time to be predicted, so as to obtain a prediction result.

The multi-energy load prediction device adopted in the embodiment comprises: the acquisition module is used for acquiring target energy load values of a target area at a plurality of historical moments; the preprocessing module is used for generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve; the extraction module is used for extracting trend curve characteristics corresponding to a plurality of target moments from the trend curve of each target energy source and extracting fluctuation curve characteristics corresponding to each target moment from the fluctuation curve corresponding to each target energy source; the training module is used for constructing a trend load prediction model based on the trend curve characteristics corresponding to each target moment and respectively constructing a fluctuation load prediction model based on the fluctuation curve characteristics of each target moment; and the prediction module is used for predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of specified historical times corresponding to the time to be predicted in the target area so as to obtain a prediction result. By constructing a trend load prediction model and a fluctuation load prediction model of each target energy, the influence of time-varying characteristics of prediction on a multi-energy system and the coupling between each energy are considered, and the prediction precision of load prediction is improved.

Optionally, the training module 1140 is configured to, for each target energy, establish a trend load prediction model of the target energy according to a long-term and short-term memory algorithm;

Optionally, the training module 1140 is configured to train a trend load prediction model of each target energy source based on a multi-task learning algorithm of a hard parameter sharing mechanism and a trend curve feature corresponding to each target time.

Optionally, the preprocessing module 1120 is configured to, for each target energy, form a load time sequence of the target energy according to the load value of the target energy, and decompose the load time sequence of the target energy according to a pole symmetric modal decomposition algorithm to obtain a plurality of modal component sequences;

Optionally, the extracting module 1130 is configured to obtain a trend curve feature corresponding to each target time, where the trend curve feature includes a load component of each target energy at a plurality of designated historical times corresponding to the target time on each trend curve;

the fluctuation curve characteristics corresponding to each target time include the load components of the respective target energy sources at a plurality of designated historical times corresponding to the target time on the respective fluctuation curves.

Optionally, the extracting module 1130 is configured to determine a trend data sequence of the target time of each target energy according to the trend curve of each target energy, and determine a fluctuation data sequence of various candidate times of each target energy according to the fluctuation curve of each target energy;

calculating the correlation coefficient between the fluctuation data sequence of each type of candidate time of the target energy and the fluctuation data sequence of the target time, the correlation coefficient between the trend data sequence of each type of candidate time of the target energy and the trend data sequence of the target time, the fluctuation data sequence of each type of candidate time of the target energy and the fluctuation data sequences of the target time of the rest of target energy, and the correlation coefficient between the trend data sequence of each type of candidate time of the target energy and the trend data sequence of the target time of the rest of target energy for each type of target energy;

selecting candidate moments with the correlation meeting preset correlation conditions from various candidate moments as specified historical moments corresponding to the target moments according to the correlation coefficients;

Optionally, the extracting module 1130 is configured to select a plurality of target moments according to a first preset time step, and for each target energy, configure the load components of the target moments on the trend curve of the target energy into a trend data sequence of the target moment, and configure the load components of the target moments on the fluctuation curve of the target energy into a fluctuation data sequence of the target moment;

Optionally, the multi-energy load prediction apparatus 11 further includes: a weather module 1160.

A meteorological module 1160, configured to obtain meteorological data of each target time in a target area to obtain meteorological features;

and adding the meteorological features into the trend curve features and the fluctuation curve features corresponding to each target moment.

Optionally, the meteorological data comprises at least one of: dry bulb temperature, wet bulb temperature, and relative humidity.

Optionally, the multi-energy load prediction apparatus 11 further includes: an evaluation module 1170.

The evaluation module 1170 is used for performing model precision evaluation on the trend load prediction model and the fluctuation load prediction model of each target energy according to preset indexes;

Optionally, the predicting module 1150 is configured to predict, according to the trend load prediction model and target energy load values of the target area at multiple designated historical times corresponding to the time to be predicted, target energy load components at the time to be predicted, so as to obtain a trend prediction result of each target energy;

Optionally, the obtaining module 1110 is configured to obtain and normalize target energy load values of the target area at multiple historical times.

Optionally, the specified historical time corresponding to each target time includes at least one of the following: the first 3 hours, the first 2 hours, and the first 1 hour of the target time. Optionally, the target energy source comprises at least one of: electric energy, heat energy, cold energy.

The multi-energy load prediction apparatus provided in this embodiment may be used to implement the method embodiments described above, and the implementation principle and the technical effect are similar, which are not described herein again.

Fig. 12 is a schematic diagram of an electronic device provided by an embodiment of the invention. As shown in fig. 12, an embodiment of the present invention provides an electronic device 12, where the electronic device 12 of the embodiment includes: a processor 1200, a memory 1210, and a computer program 1220 stored in the memory 1210 and executable on the processor 1200. The processor 1200, when executing the computer program 1220, implements the steps of the above-described embodiments of the multi-energy load prediction method, such as the steps 201 to 205 shown in fig. 2. Alternatively, the processor 1200, when executing the computer program 1220, implements the functions of the modules/units in the above-described device embodiments, such as the functions of the modules 1110 to 1150 shown in fig. 11.

Illustratively, the computer program 1220 may be divided into one or more modules/units, which are stored in the memory 1210 and executed by the processor 1200 to implement the present invention. One or more of the modules/units may be a series of computer program instruction segments capable of performing certain functions that describe the execution of the computer program 1220 on the electronic device 12.

The electronic device 12 may be a desktop computer, a notebook, a palm top computer, a cloud server, or other computing device. The terminal may include, but is not limited to, a processor 1200, a memory 1210. Those skilled in the art will appreciate that fig. 12 is merely an example of electronic device 12 and does not constitute a limitation of electronic device 12 and may include more or fewer components than shown, or some components may be combined, or different components, e.g., a terminal may also include input-output devices, network access devices, buses, etc.

The Processor 1200 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The storage 1210 may be an internal storage unit of the electronic device 12, such as a hard disk or a memory of the electronic device 12. The memory 1210 may also be an external storage device of the electronic device 12, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like provided on the electronic device 12. Further, the memory 1210 may also include both internal storage units of the electronic device 12 and external storage devices. The memory 1210 is used to store computer programs and other programs and data required by the terminal. The memory 1210 may also be used to temporarily store data that has been output or is to be output.

An embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the steps in the foregoing embodiments of the multi-energy load prediction method are implemented.

The computer-readable storage medium stores a computer program 1220, the computer program 1220 includes program instructions, and when the program instructions are executed by the processor 1200, all or part of the processes in the method according to the above embodiments may be implemented by the computer program 1220 instructing related hardware, and the computer program 1220 may be stored in a computer-readable storage medium, and when the computer program 1220 is executed by the processor 1200, the steps of the above embodiments of the method may be implemented. Computer program 1220 includes, among other things, computer program code, which may be in the form of source code, object code, an executable file or some intermediate form. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, U.S. disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution media, and the like. It should be noted that the computer readable medium may include any suitable increase or decrease as required by legislation and patent practice in the jurisdiction, for example, in some jurisdictions, computer readable media may not include electrical carrier signals and telecommunications signals in accordance with legislation and patent practice.

The computer readable storage medium may be an internal storage unit of the terminal of any of the foregoing embodiments, for example, a hard disk or a memory of the terminal. The computer readable storage medium may also be an external storage device of the terminal, such as a plug-in hard disk provided on the terminal, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like. Further, the computer-readable storage medium may also include both an internal storage unit and an external storage device of the terminal. The computer-readable storage medium is used for storing a computer program and other programs and data required by the terminal. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules, so as to perform all or part of the functions described above. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal and method may be implemented in other ways. For example, the above-described apparatus/terminal embodiments are merely illustrative, and for example, a module or a unit may be divided into only one logical function, and may be implemented in other ways, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow in the method according to the embodiments of the present invention may also be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of the embodiments of the method. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, U.S. disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution media, and the like. It should be noted that the computer readable medium may include any suitable increase or decrease as required by legislation and patent practice in the jurisdiction, for example, in some jurisdictions, computer readable media may not include electrical carrier signals and telecommunications signals in accordance with legislation and patent practice.

The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.

Claims

1. A method for predicting a multi-energy load, comprising: acquiring target energy load values of a target area at a plurality of historical moments; the target energy sources are at least two; generating a variation characteristic curve of each target energy according to each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve;

2. The multi-energy load prediction method according to claim 1, wherein the step of constructing a trend load prediction model based on the trend curve characteristics corresponding to each target time, and the step of constructing a fluctuating load prediction model based on the fluctuating curve characteristics of each target time respectively comprises:

3. The multi-energy load prediction method of claim 2, wherein training a trend load prediction model of each target energy source based on trend curve characteristics corresponding to each target time comprises:

4. The multi-energy load prediction method according to claim 1, wherein a variation characteristic curve of each target energy is generated based on each target energy load value; wherein the variation characteristic curve comprises a trend curve and a fluctuation curve, comprising:

5. The multi-energy load prediction method according to claim 1, wherein the trend curve feature corresponding to each target time comprises a load component of each target energy source at a plurality of designated historical times corresponding to the target time on each trend curve;

6. The multi-energy load prediction method according to claim 5, wherein extracting trend curve features corresponding to a plurality of target time points from the trend curve of each target energy source, and extracting a fluctuation curve feature corresponding to each target time point from the fluctuation curve corresponding to each target energy source comprises:

7. The multi-energy load prediction method of claim 6, wherein determining a trend data series of the target time of each target energy source according to the trend curve of each target energy source, and determining a fluctuation data series of various candidate times of each target energy source according to the fluctuation curve of each target energy source comprises:

8. The multi-energy load prediction method according to any one of claims 1 to 7, wherein before the trend load prediction model is constructed based on the trend curve characteristics corresponding to each target time, and before the fluctuating load prediction model is constructed based on the fluctuating curve characteristics at each target time, the method further comprises:

acquiring meteorological data of each target moment of the target area to obtain meteorological characteristics;

adding the meteorological features into the trend curve features and the fluctuation curve features corresponding to each target moment;

wherein the meteorological data comprises at least one of: dry bulb temperature, wet bulb temperature and relative humidity;

after a trend load prediction model is built based on the trend curve characteristics corresponding to each target moment and a fluctuation load prediction model is built based on the fluctuation curve characteristics of each target moment, the method further comprises the following steps:

if the evaluation result is qualified, predicting each target energy load value at the time to be predicted according to the trend load prediction model, the fluctuation load prediction model and each target energy load value of a plurality of appointed historical times corresponding to the time to be predicted of the target area to obtain a prediction result;

otherwise, continuing to train a trend load prediction model and a fluctuating load prediction model of each target energy;

the predicting, according to the trend load prediction model, the fluctuation load prediction model, and each target energy load value of a plurality of designated historical moments corresponding to the target area at the moment to be predicted, each target energy load value at the moment to be predicted to obtain a prediction result, includes:

performing one-to-one weighted reconstruction on the trend prediction result of each target energy and the trend prediction result of each target energy to obtain the prediction result of each target energy;

the acquiring of the target energy load values of the target area at a plurality of historical moments includes:

acquiring target energy load values of a target area at a plurality of historical moments and carrying out normalization processing on the target energy load values;

the appointed historical time corresponding to each target time comprises at least one of the following items: the first 3 hours, the first 2 hours, and the first 1 hour of the target time;

the target energy source comprises at least one of: electric energy, heat energy, cold energy.

9. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the multi-energy load prediction method according to any one of the preceding claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method for multi-energy load prediction according to any one of claims 1 to 8.