CN116266700A - Prediction method for adjustable potential of industrial load demand response - Google Patents
Prediction method for adjustable potential of industrial load demand response Download PDFInfo
- Publication number
- CN116266700A CN116266700A CN202111530350.5A CN202111530350A CN116266700A CN 116266700 A CN116266700 A CN 116266700A CN 202111530350 A CN202111530350 A CN 202111530350A CN 116266700 A CN116266700 A CN 116266700A
- Authority
- CN
- China
- Prior art keywords
- prediction
- industrial
- load
- energy consumption
- response
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000004044 response Effects 0.000 title claims abstract description 46
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000005265 energy consumption Methods 0.000 claims abstract description 16
- 230000005611 electricity Effects 0.000 claims abstract description 13
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 11
- 230000001105 regulatory effect Effects 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 3
- 230000015654 memory Effects 0.000 abstract description 5
- 238000010801 machine learning Methods 0.000 abstract description 4
- 238000004519 manufacturing process Methods 0.000 abstract description 2
- 238000009776 industrial production Methods 0.000 abstract 1
- 238000005065 mining Methods 0.000 abstract 1
- 238000004458 analytical method Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Power Engineering (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of short-term load prediction of power demand response, and particularly relates to a prediction method based on a two-way long-short-term memory (LSTM) network prediction method process for industrial load demand response regulation potential, which is used for supporting a power grid to develop industrial renewable energy consumption. Load prediction is important for balancing power supply and demand and new energy consumption. At present, most load prediction based on machine learning belongs to a unidirectional LSTM network, and prediction accuracy and precision are limited, so that the load prediction method is improved through the bidirectional LSTM network. Industrial load prediction scenes are oriented to consider industrial electric equipment and production process flows, and the adjustable potential of different industrial production typical factories is predicted based on historical electricity consumption data. The universality and the prediction accuracy of the method are improved on the premise of considering the new energy consumption of the area, and the value mining of the adjustable load resources is realized.
Description
Technical Field
The invention designs a prediction method for adjustable potential of industrial load demand response.
Technical Field
A large amount of clean renewable energy sources are put into use, carbon emission can be reduced, however, the output of the renewable energy sources has intermittence, so that the renewable energy sources are matched with the power demand, the adjustable potential of a power consumer is excavated, new energy source consumption can be carried out on the side of the demand, and the stability of a power grid can be kept.
The load prediction can improve the new energy consumption reliability of the novel power system. Due to the modernization of power systems, the application of load prediction is becoming more important in the current demand response context. The advantage of demand response to new energy consumption also relies on load prediction techniques to analyze demand and regulatory potential. Short-term load prediction has a certain research basis, however, due to the nonlinear characteristics of univariate time series historical load data, a prediction method based on machine learning is widely used, wherein a long short-term memory (LSTM) network is widely used in the scenes of short-term load demand, power fluctuation prediction tasks and the like. However, the unidirectional LSTM network prediction method may lack part of the hidden layer data because it uses only unidirectional memories, and the previous and future hidden layer data may be used by two directional memories (feedforward and feedback loops) through the bidirectional LSTM network prediction method used to improve the prediction accuracy. In this way, the prediction method provided can effectively extract all hidden layer characteristics and improve the prediction accuracy.
Disclosure of Invention
The invention belongs to the technical field of demand side user adjustable potential analysis, and particularly relates to an industrial load demand response adjustment potential oriented prediction method which is used for supporting a power grid to develop large-scale renewable energy consumption based on industrial users. With the implementation of national energy strategies, renewable energy sources face a history of larger scale development, and at the same time, a certain degree of electricity limiting problems have begun to appear in various places. The industrial high-energy-consumption load occupies a large area, and the seasonal variation has certain fluctuation, and can be used as a new energy consumption object to analyze the adjustable potential of the industrial high-energy-consumption load, so that the nonlinear characteristics of single-variable time sequence historical load data are considered, the LSTM improved bidirectional network algorithm is adopted to combine the electricity utilization characteristics of different industrial typical electric equipment, the adjustable indexes of new energy consumption are analyzed and quantized through the historical data, the new energy consumption requirement is met when the scene of industrial adjustable potential analysis is met, the power supply and demand balance can be blocked, and the operation reliability of an electric power system is improved.
Drawings
FIG. 1 is a model diagram of an industrial load demand response adjustment potential oriented in accordance with an embodiment of the present invention; and
FIG. 2 is a flow chart of a method for predicting the regulatory potential of an industrial load demand response in accordance with an embodiment of the present invention.
Detailed Description
The exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, however, the present invention may be embodied in many different forms and is not limited to the examples described herein, which are provided to fully and completely disclose the present invention and fully convey the scope of the invention to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the accompanying drawings is not intended to be limiting of the invention. In the drawings, like elements/components are referred to by like reference numerals.
Unless otherwise indicated, terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. In addition, it will be understood that terms defined in commonly used dictionaries should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense.
Considering that the industrial enterprises on the demand side are numerous, the electric equipment of the same kind is numerous, and the production flows of the same industry are different, the electric equipment power utilization characteristics adopted in the process of inputting the model are set as variable parameters for simplicity.
Step 1: each input load history data sequence per day is divided into different clusters. For the historical data to predict each cluster, a separate input vector for each variable is considered. Since the trend of each prediction parameter is different, a specific structure is defined for each prediction parameter, i.e. a specific prediction network is specified. At this stage, a clustering objective is set for daily electricity usage, and the input data sequence of the predictive task is designated as its predictive network in a supervised manner.
The clustering task is completed under an unsupervised loop through a machine learning method. Giving an evaluation cluster performance DB parameter, defining the DB parameter as the inter-cluster distance in a cluster, and adopting the following formula:
the optimal clustering scheme, namely the DB value is minimum, the daily industrial electricity data is trained through clustering, and the mass center of each cluster is regarded as the target of the cluster. Data classification is then performed by the usual machine learning method of SVM. The SVM core idea is a kernel function that converts nonlinear features into linear features by mapping data to a new feature space with higher dimensions and then establishing a linear relationship between input and output variables within the new feature space. Let the training sample set be s= { (x) i ,y i )，x i ∈R n ,y i E R, i=1, 2,..i }, where x i Input for training samples in n dimensions, y i For training sample output, l is the number of samples. The SVM fits the estimate with the following equation:
wherein the method comprises the steps of
A nonlinear mapping of the input space to a new high-dimensional feature space; omega is weight vectorAn amount of; b is the bias. Assuming that all SCM training data uses linear function fitting, i.e., solving a minimized optimization problem:
at this time, K (x, x) is further calculated by a kernel function solving method i )＝ω T ω, the nonlinear function that needs to be solved at this time can be expressed as a high-dimensional feature space linear map:
in this step, the SVM is used to supervise and classify the daily electricity consumption data distribution prediction network.
Step 2:
the indexes for quantitatively adjusting the new energy consumption demand side are taken as model coefficients, and the load curves in the historical industrial demand response offers are adopted for analysis and are divided into the following 3 types:
1. predicting a baseline load: predicting the actual electricity consumption base line load of the response day if the demand response is not implemented according to the historical electricity consumption data, and recording as Q 0 。
2. Calculating the maximum response load: deriving the maximum response load of the day of the response from the baseline load and the offer response load, i.e. the load curve when the offer response reaches the maximum response capacity, denoted as Q m 。
3. The response standard capacity is given: the power grid gives the response standard of the current day of response according to the baseline load and the offer response load, namely the response capacity which is required to be achieved by a user according to the contract is recorded as Q 1 。
According to the 3 load curves, the required response degree alpha of the power grid can be quantized 1 (t) and the actual degree of response α (t):
in which Q 0 (t) represents a baseline load power for the industrial user offer to respond to the time of day t; q (Q) 1 (t) responding to standard capacity for the time period t of the day of the industrial user offer response; q (Q) m (t) is the maximum response load power for the time t period of the day of the industrial user offer response. From the definition of the response degree, to quantify the response behavior of the industrial user, the response completion degree of the industrial user offer and the income ratio of the user participation offer can be comprehensively described, and the higher the income ratio of the industrial user participation response is, the stronger the response will of the offer is.
The following three points are comprehensively considered to calculate the equivalent response degree of the user: response completion for 1.t time period; 2. average offer response completion of industrial users; the equivalent response completion of the user is affected by the response demands of the power grid to the user at different times. According to the three principles, defining the response completion degree R of the industrial user c And maximum response completion degree R c,max ：
The greater the response completions obtainable from the above analysis, the greater the maximum response completions, the greater the industry user demand response potential.
The index for representing the adjustment of the new energy consumption requirement side offer by the industrial user is obtained and then is used as a model coefficient to be input.
Step 3: the unidirectional LSTM network is constructed as a basis for prediction by constructing the bidirectional LSTM network in the next step. LSTM networks solve the problem of rapid gradient norms decrease based on long time components during cyclic neural network (RNN) training, with various gates more than RNN as powerful memory units. Each LSTM network has three main gates as input, output and forget gates. The proposed bidirectional LSTM predictive network is formed by stacking various LSTM networks. In this case, the input data of each LSTM network at time t is considered the output of the same network at time t-1, and at time t is considered the output of the previous network. Features will be passed between these networks during training, according to the stacked configuration of the LSTM networks. The main formulas for LSTM networks are as follows:
S t ＝O t tanh(c t ) (14)
here, the model parameters are defined as W it ，
W ht ，/>
b i ，b f ，b c ，
The parameters can be optimally adjusted in the training process through an optimization algorithm.
Step 4: and establishing a bidirectional LSTM network as a prediction model on the basis of the unidirectional LSTM network structure. In order to solve the problem that information data containing future characteristics is ignored when unidirectional LSTM network data are processed in a time mode, bidirectional LSTM network is adopted for prediction. Such a bi-directional LSTM network can not only use historical features, but also identify information containing future features, in such a way that predictions can relate to data features over time. As shown in fig. 1, the overall structure of a bi-directional LSTM network consists of two main hidden layers. In each layer, consider stacking identical LSTM networks, one as a backward hidden layer and the other as a forward hidden layer, i.e., bi-directional LSTM networks have two ways of conveying information: one from the past to the future and the other from the future to the past. This structure allows the output of the prediction network to be used as input data in future prediction steps while having a powerful memory, and allows all useful history and future prediction data to be stored with high accuracy. Thus, a more accurate bi-directional LSTM network is employed in predicting load adjustable potential with high randomness and intermittent behavior.
In a bidirectional LSTM network of time period t, data is input in small batches
The forward and backward hidden layers are assumed to be +.>
And->
Hidden layer->
The first two are integrated:
O fn ＝H t W o +b o (17)
here, the model parameters are defined as
The overall flow chart of the proposed method is shown in fig. 2. After training, the new input data is fed back only to the SVM and the predictive part. During bi-directional LSTM network prediction, the optimal number of clusters and the centroid of each cluster (data tag of each cluster) are fixed. The method avoids the overfitting process for training the proposed deep structure network. Thus, the calculation amount of the prediction process remains within an acceptable range.
Claims (4)
1. A method of predicting an industrial load demand response adjustable potential, the method comprising:
the historical electricity utilization data and the electricity utilization characteristics of main electric equipment are used as model input; and
quantitatively taking an index for regulating the new energy consumption requirement side as a model coefficient; and
and optimizing model parameters under the bidirectional LSTM network prediction method through an iterative algorithm to improve accuracy.
2. The scheme of claim 1, wherein the prediction model input comprises historical electricity utilization data of an industrial typical Buddha factory and electricity utilization characteristics of major electric equipment of a sub-industry owner, and the prediction model input can be expanded into a prediction factor of the load of the same type of industrial user.
3. The scheme of claim 1, wherein the index for regulating the new energy consumption requirement side is mainly described as a constraint condition which is satisfied by combining the actual industrial electricity consumption situation and the new energy consumption requirement under the condition of regulating the maximum capacity based on the history contract requirement response.
4. The scheme of claim 1, wherein the optimization of the bi-directional LSTM network parameters by an iterative algorithm improves accuracy by dividing training data into several equal subsets during a training phase and cross-validating the prediction accuracy parameters until convergence testing is passed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111530350.5A CN116266700A (en) | 2021-12-15 | 2021-12-15 | Prediction method for adjustable potential of industrial load demand response |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111530350.5A CN116266700A (en) | 2021-12-15 | 2021-12-15 | Prediction method for adjustable potential of industrial load demand response |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116266700A true CN116266700A (en) | 2023-06-20 |
Family
ID=86742750
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111530350.5A Pending CN116266700A (en) | 2021-12-15 | 2021-12-15 | Prediction method for adjustable potential of industrial load demand response |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116266700A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116579590A (en) * | 2023-07-13 | 2023-08-11 | 北京圆声能源科技有限公司 | Demand response evaluation method and system in virtual power plant |
-
2021
- 2021-12-15 CN CN202111530350.5A patent/CN116266700A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116579590A (en) * | 2023-07-13 | 2023-08-11 | 北京圆声能源科技有限公司 | Demand response evaluation method and system in virtual power plant |
| CN116579590B (en) * | 2023-07-13 | 2023-11-10 | 北京圆声能源科技有限公司 | Demand response evaluation method and system in virtual power plant |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106651020B (en) | Short-term power load prediction method based on big data reduction | |
| CN111353656B (en) | Steel enterprise oxygen load prediction method based on production plan | |
| Hong | Hybrid evolutionary algorithms in a SVR-based electric load forecasting model | |
| CN105631528B (en) | Multi-target dynamic optimal power flow solving method based on NSGA-II and approximate dynamic programming | |
| CN111460001B (en) | Power distribution network theoretical line loss rate evaluation method and system | |
| CN108876001A (en) | A kind of Short-Term Load Forecasting Method based on twin support vector machines | |
| Al Mamun et al. | A hybrid deep learning model with evolutionary algorithm for short-term load forecasting | |
| Zhang et al. | A CNN and LSTM-based multi-task learning architecture for short and medium-term electricity load forecasting | |
| Liu et al. | Forecasting the occurrence of extreme electricity prices using a multivariate logistic regression model | |
| CN115186803A (en) | Data center computing power load demand combination prediction method and system considering PUE | |
| CN114066071A (en) | Power parameter optimization method based on energy consumption, terminal equipment and storage medium | |
| Zhou et al. | An External Archive‐Based Constrained State Transition Algorithm for Optimal Power Dispatch | |
| CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
| CN116266700A (en) | Prediction method for adjustable potential of industrial load demand response | |
| CN111680939A (en) | Enterprise re-work and re-production degree monitoring method based on artificial intelligence | |
| Wang et al. | Big data analytics for price forecasting in smart grids | |
| Haq et al. | Classification of electricity load profile data and the prediction of load demand variability | |
| CN114819395A (en) | Industry medium and long term load prediction method based on long and short term memory neural network and support vector regression combination model | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| CN109214610A (en) | A kind of saturation Methods of electric load forecasting based on shot and long term Memory Neural Networks | |
| KR102478684B1 (en) | Method for predicting energy consumption for using ensemble learning, and computing apparatus for performing the method | |
| Wibawa et al. | Deep Learning Approaches with Optimum Alpha for Energy Usage Forecasting | |
| CN115689001A (en) | Short-term load prediction method based on pattern matching | |
| Kayakuş | The Estimation of Turkey's Energy Demand Through Artificial Neural Networks and Support Vector Regression Methods | |
| Qu et al. | Probability prediction method of short-term electricity price based on quantile neural network model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |