CN116384251A - New energy generated power combination prediction method and system considering risk avoidance - Google Patents
New energy generated power combination prediction method and system considering risk avoidance Download PDFInfo
- Publication number
- CN116384251A CN116384251A CN202310416136.XA CN202310416136A CN116384251A CN 116384251 A CN116384251 A CN 116384251A CN 202310416136 A CN202310416136 A CN 202310416136A CN 116384251 A CN116384251 A CN 116384251A
- Authority
- CN
- China
- Prior art keywords
- model
- prediction
- data
- new energy
- predicted
- 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
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000005457 optimization Methods 0.000 claims abstract description 50
- 238000012360 testing method Methods 0.000 claims abstract description 37
- 238000007781 pre-processing Methods 0.000 claims abstract description 18
- 238000005096 rolling process Methods 0.000 claims abstract description 15
- 238000002790 cross-validation Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims description 47
- 238000012795 verification Methods 0.000 claims description 32
- 238000011156 evaluation Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 7
- 238000010200 validation analysis Methods 0.000 claims description 6
- 238000007637 random forest analysis Methods 0.000 claims description 5
- 238000009499 grossing Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000012706 support-vector machine Methods 0.000 claims description 4
- 238000002759 z-score normalization Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000004590 computer program Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 10
- 238000010248 power generation Methods 0.000 description 9
- 230000006872 improvement Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 8
- 238000000638 solvent extraction Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 4
- 230000002411 adverse Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000000053 physical method Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 241000728173 Sarima Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000002803 fossil fuel Substances 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Game Theory and Decision Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Primary Health Care (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Water Supply & Treatment (AREA)
- Medical Informatics (AREA)
Abstract
The invention discloses a new energy generated power combination prediction method and a system considering risk avoidance, comprising the following steps: carrying out data preprocessing and standardization on the historical output data of the new energy station and the numerical weather forecast of the corresponding time period to obtain a whole data set; adopting M different base models to generate corresponding prediction results; optimizing and solving the data in the selected time window to obtain the optimal weight of the combined model, obtaining weights of different base models, and obtaining an optimized combined model considering the condition risk value by weighting and summing the predicted values of the base models; performing super-parameter optimization on the optimized combination model considering the condition risk value in a time sequence cross-validation mode; and rolling prediction is carried out on the test set by applying an optimized combination model according to the optimal super parameters obtained by super parameter optimizing, and a prediction result corresponding to the time period to be predicted is generated. The invention can improve the prediction precision and reduce the extreme error of prediction at the same time, thereby providing high-quality prediction results for decision makers.
Description
Technical Field
The invention relates to the field of prediction of new energy generated power of an electric power system, in particular to a new energy generated power combination prediction method and a new energy generated power combination prediction system considering risk avoidance.
Background
Energy is an important material basis for human survival. Since the eighteenth century industrial revolution, fossil energy such as coal plays an indispensable role in human society development, but at the same time, massive combustion of fossil fuel also causes a heavy burden on natural environment, and new energy such as wind power, photovoltaics and the like is used for replacing fossil energy for sustainable development of human society and environment, so that the novel energy becomes an important task for building a novel power system. But the new energy power generation has the characteristics of intermittence, randomness and fluctuation, which has adverse effect on the power system. The new energy power generation large-scale grid connection brings higher requirements on means and capability of power system regulation, increases difficulty of economic dispatching of the system, reduces availability of power equipment and increases power supply cost of the system. The advanced new energy prediction method is developed, the power prediction level is improved, adverse effects on the operation of the power system caused by the uncertainty of new energy generation can be reduced, and the power-assisted energy field is comprehensively and environmentally-friendly transformed and upgraded.
The current widely applied new energy prediction methods mainly can be divided into two types, namely a physical method and a statistical method, wherein the physical method focuses on constructing a physical model of weather information and power generation equipment, and meteorological features obtained through numerical weather forecast are input into the constructed physical model to obtain a power prediction result; the statistical method is to build the mapping relation between factors such as meteorological features and historical output and output power, learn from historical data and generalize the mapping relation to unknown data, and common methods include autoregressive moving average models, support vector machines, random forests, artificial neural networks and the like.
Whether physical or statistical, researchers often seek to find the best individual model for new energy predictions, and due to the complexity, contingency, and variability of real data, selecting only one optimal individual model to predict location data may be lacking in stability and accuracy. In recent years, a combination prediction method is widely focused, and a plurality of prediction results with different sources and different precision are combined according to a certain strategy to obtain a more accurate and more robust prediction result.
The combined prediction method widely used at present often focuses on the accuracy of the prediction result. In practical application, besides the accuracy of the prediction result, the extreme error of the prediction result is also focused by the decision maker because the extreme error of the prediction result can adversely affect the scheduling scheme of the decision maker and the stable operation of the power system.
Disclosure of Invention
The invention aims to provide a new energy power generation power combination prediction method and system which are suitable for engineering practice and can realize risk avoidance. The invention aims to provide a new energy power generation combination prediction method considering risk avoidance, wherein the weight of each base model is calculated through an optimization method, so that the robustness and accuracy of a prediction result are improved, and meanwhile, related risk metering is added into a target and constraint to provide the prediction result for reducing extreme errors.
In order to achieve the above objective, the present invention provides a new energy power generation optimal combination prediction method considering a condition risk value (CVaR), which provides a decision maker with a prediction result considering both extreme error (measured by CVaR) and accuracy (measured by root mean square error RMSE). The method specifically comprises the following steps:
a new energy generated power combination prediction method considering risk avoidance comprises the following steps:
carrying out data preprocessing and standardization on the historical output data of the new energy station and the numerical weather forecast of the corresponding time period to obtain a whole data set; dividing the whole data set into a training set, a verification set and a test set;
m different base models are adopted, and corresponding prediction results are generated according to training of the training set;
optimizing and solving the data in the selected time window to obtain the optimal weight of the combined model, obtaining weights of different base models, and obtaining an optimized combined model considering the condition risk value by weighting and summing the predicted values of the base models;
performing super-parameter optimization on the optimized combination model considering the condition risk value in a time sequence cross-validation mode;
and rolling prediction is carried out on the test set by applying an optimized combination model according to the optimal super parameters obtained by super parameter optimizing, and a prediction result corresponding to the time period to be predicted is generated.
As a further improvement of the invention, the training set is used for training a base model, the verification set is used for carrying out super-parameter optimization on the combined prediction model, and the test set is used for testing the performance of the combined prediction model.
As a further improvement of the present invention, the base model includes an autoregressive moving average model, an exponential smoothing model, etc., and a machine learning method such as a random forest, a recurrent neural network, a support vector machine, etc.
As a further improvement of the present invention, the normalization is performed by a z-score normalization method, which comprises:
wherein x is * For standardized data, x is the original data, mu is the mean value of the data, and sigma is the standard deviation of the data.
As a further improvement of the invention, the optimal weight of the combined model is obtained by carrying out optimal solution on the data in the selected time window, and the weights omega of different base models are obtained j ，j＝1,2,...,N m Obtaining an optimized combination model considering the condition risk value by weighting and summing the predicted values of the base model; comprising the following steps:
the constraint conditions of the optimization combination model considering the conditional risk value are shown in formulas (2) to (6), the objective function comprises three items of an extreme error item, an accuracy item and a regular item, and the constraint conditions (5) and (6) are related constraints of the conditional risk value; the extreme error adopts a conditional risk value method, and the accuracy adopts a root mean square error method;
obtaining the weight omega of the basic model j ，j＝1,2,...,N m The predicted value of the optimized combination model at time t is represented by equation (7), where y j,t The predicted value of the base model j at the time t;
wherein omega is j For the corresponding weight of the base model j, E is the base model error matrix in the time window, N t For the time window length, lambda is the relative proportion of the extreme error term and the precision term, mu is the weight of the regular term, beta is the preset condition risk value confidence level, and alpha is the auxiliary variable.
As a further improvement of the invention, the method for optimizing the super-parameters of the optimized combination model considering the conditional risk value by the time sequence cross-validation comprises the following steps:
performing super-parameter optimization on the optimized combination model considering the condition risk value in a time sequence cross-validation mode on the validation set;
the time series cross verification adopts a rolling mode, and a certain length N is selected from a certain moment t Based on which the data subset of the data points is predicted and the predicted result information of the predicted data points is counted; then, the starting point of the data subset is moved backwards by the step number which is the same as the number of the predicted data points to form a new data subset, and the subsequent data points are predicted; and so on until the prediction of the whole verification set is realized and the result is recorded.
As a further improvement of the present invention, the super-parameters include a time window length N t And the regular term coefficient mu, wherein the evaluation index is a comprehensive index which simultaneously considers root mean square error and conditional risk value, and the comprehensive evaluation index OWA is selected.
As a further improvement of the present invention, the evaluation index is a comprehensive index which considers both root mean square error and conditional risk value, and includes the following contents:
the comprehensive evaluation index OWA is used for calculating the comprehensive score of the prediction result in the two aspects of RMSE and CVaR, and the specific expression is shown in (8);
wherein RMSE Ave And CVaR (CVaR) Ave The performance indexes corresponding to the simple average combination method are respectively shown.
As a further improvement of the present invention, the generating the prediction result corresponding to the period to be predicted further includes:
and reporting the prediction result for further application or comparison analysis with other prediction models.
A new energy generated power combination prediction system considering risk avoidance comprises the following steps:
the data preprocessing and set dividing module is used for preprocessing and standardizing the historical output data of the new energy station and the numerical weather forecast of the corresponding time period to obtain a whole data set; dividing the whole data set into a training set, a verification set and a test set;
the base model training and predicting module is used for generating corresponding predicting results by adopting M different base models according to training set training;
the optimization combination model construction module is used for obtaining the optimal weight of the combination model by carrying out optimization solution on the data in the selected time window, obtaining weights of different base models, and obtaining an optimization combination model considering the conditional risk value by weighting and summing the predicted values of the base models;
the optimizing module is used for optimizing the super parameters of the optimizing combination model which takes the condition risk value into account in a time sequence cross verification mode;
and the optimal combination model prediction result generation module is used for carrying out rolling prediction on the test set by applying the optimal combination model according to the optimal super parameter obtained by super parameter optimization so as to generate a prediction result corresponding to the time period to be predicted.
Compared with the prior art, the invention has at least the following beneficial effects:
the method comprises data preprocessing and set partitioning, base model training and prediction result recording, construction of an optimization combination model considering the condition risk value, super-parameter optimization of the optimization combination model and generation of the prediction result of the optimization combination model. The method can effectively measure the performance of the new energy prediction extreme error, namely, the index of introducing the conditional risk value in the new energy prediction aspect is used for measuring the extreme error between the prediction result and the actual value. The method can effectively reduce the extreme error of the prediction result and has outstanding performance in the aspect of accuracy. The method has the advantages that the condition risk value related items are introduced into the objective function and the constraint condition of the optimized combination model, weights of different base models are obtained through solving in a time window, and the risk avoidance effect is achieved. The invention has excellent generalization capability. By introducing a regular term related to the weight, the extreme of weight distribution is avoided; through cross-validation-based super-parameter optimization on the validation set, the optimal super-parameters including time window length and regular term coefficients are obtained, so that the model provided by the invention has excellent generalization capability. The method considers that the extreme errors of the prediction result are reduced by a combined prediction method, and provides high-quality prediction results for a decision maker.
Drawings
FIG. 1 is a block diagram of a technical scheme adopted by the invention;
FIG. 2 is a schematic diagram of cross-validation of the present invention during a super-parameter optimization process;
FIG. 3 is a schematic diagram of the rolling optimization and combination prediction in step 5 according to the present invention;
FIG. 4 is a schematic diagram of set partitioning in an embodiment;
FIG. 5 is a graph showing test set prediction results versus the example;
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The first object of the present invention is to provide a new energy generated power combination prediction method considering risk avoidance, comprising the following steps:
carrying out data preprocessing and standardization on the historical output data of the new energy station and the numerical weather forecast of the corresponding time period to obtain a whole data set; dividing the whole data set into a training set, a verification set and a test set;
m different base models are adopted, and corresponding prediction results are generated according to training of the training set;
optimizing and solving the data in the selected time window to obtain the optimal weight of the combined model, obtaining weights of different base models, and obtaining an optimized combined model considering the condition risk value by weighting and summing the predicted values of the base models;
performing super-parameter optimization on the optimized combination model considering the condition risk value in a time sequence cross-validation mode;
and rolling prediction is carried out on the test set by applying an optimized combination model according to the optimal super parameters obtained by super parameter optimizing, and a prediction result corresponding to the time period to be predicted is generated.
The prediction method mainly comprises data preprocessing and set division, basic model training and prediction result recording, construction of an optimization combination model considering the condition risk value, super-parameter optimization of the optimization combination model and generation of the prediction result of the optimization combination model.
Wherein: the data preprocessing and the set dividing are used for performing z-score standardization processing on the original data, dividing the set into a training set, a verification set and a test set, and respectively used for basic model training, optimizing the super-parameter of the combined model and comparing the prediction results; the base model training and predicting result record takes a plurality of predicting methods as a base model, and generates the base model predicting results of a verification set and a test set after training by a training set; the optimized combination model considering the condition risk value is constructed in an objective function and constraint conditions of the optimized combination model, meanwhile, root mean square error, condition risk value, regular term and constraint with nonnegative sum of weights being 1 are considered, combination weights of all base models can be obtained by optimizing and solving data in a time window, and combination prediction results can be obtained by weighting and summing prediction results of all base models in a prediction time period; the super-parameter optimizing of the optimized combination model is to optimize the time window length and the regular term coefficient related in the proposed optimized combination model in a time sequence cross-validation mode on a validation set, and takes a comprehensive index which simultaneously considers root mean square error and conditional risk value as an evaluation standard; the generation of the optimal combination model prediction result is to generate a combination prediction result on a test set in a rolling optimization and prediction mode on the premise of applying optimal super parameters, and compare the combination prediction result with a base model and other combination prediction models or report the result to related departments for further reference and application.
The invention realizes the new energy power generation combined prediction method capable of reducing extreme errors, can reduce the predicted extreme errors while improving the prediction precision, and provides high-quality prediction results for decision makers.
The following examples are given to illustrate the steps of the present invention in detail. The method specifically comprises the following steps:
step 1: data preprocessing and set partitioning
Firstly, carrying out data preprocessing and standardization processing on historical output data of a new energy station and numerical weather forecast of a corresponding time period. And then dividing the whole data set into a training set, a verification set and a test set, wherein the training set is used for training the base model, the verification set is used for carrying out super-parameter optimization on the combined prediction model, and the test set is used for testing the performance of the combined prediction model. In actual production, the test set is the period to be predicted.
Step 2: base model training and predictive outcome logging
M different basic prediction methods are adopted, and prediction results of a verification set and a test set are generated after training according to training set data. The base model comprises an autoregressive moving average model, an exponential smoothing model and the like, and a random forest, a cyclic neural network, a support vector machine and other machine learning methods.
Step 3: optimization combination model construction considering condition risk value
The invention obtains the optimal weight of the combined model by carrying out optimization solution on the data in the selected time window, and obtains the weights omega of different base models j (j＝1,2,...,N m ) The predicted value of the optimized combination model at the time t can be obtained by weighted summation of the predicted values of the base model.
Step 4: super-parameter optimizing method for optimizing combined model
And performing super-parameter optimization on the optimized combination model considering the conditional risk value on the verification set by a time sequence cross-verification mode, wherein the cross-verification process is schematically shown in figure 2. Super-parameters include time window length N t And the regular term coefficient mu, wherein the evaluation index is a comprehensive index which simultaneously considers root mean square error and conditional risk value, and the comprehensive evaluation index OWA is selected.
The time series cross verification adopts a rolling mode, and a certain length N is selected from a certain moment t Based on which the following data points are predicted and the predicted outcome information of the predicted data points is counted. And then, the starting point of the data subset is moved backwards by the same steps as the number of the predicted data points to form a new data subset, and the subsequent data points are predicted. And so on until the prediction of the whole verification set is realized and the result is recorded.
Step 5: optimizing combination model prediction result generation
According to the optimal super-parameters obtained by super-parameter optimization in the step 4, rolling prediction is carried out on a test set by applying the optimal combination model provided by the invention, and the prediction process is schematically shown in figure 3. And generating a prediction result corresponding to the time period to be predicted, and reporting the result for further application or comparison analysis with other prediction models.
Specifically, the data normalization processing mode in the step 1 is a z-score normalization method, and a specific formula can be expressed as formula (1), wherein x is * For standardized data, x is the original data, mu is the mean value of the data, and sigma is the standard deviation of the data. After z-score normalization, each feature had a mean value of 0 and standard deviation of 1.
Specifically, in step 3, the objective function and constraint conditions of the optimized combination model provided by the invention are shown in formulas (2) to (6), the objective function comprises three items of extreme error (conditional risk value CVaR), precision (root mean square error RMSE) and regular term, constraint conditions (5) and (6) are related constraints of conditional risk value, and constraints (3) and (4) ensure that the combination weights are non-negative and sum to be 1. Omega in j For the corresponding weight of the base model j, E is the base model error matrix in the time window, N t For the time window length, lambda is the relative proportion of the extreme error term and the precision term, mu is the weight of the regular term, beta is the preset condition risk value confidence level, and alpha is the auxiliary variable.
Specifically, in step 3, the predicted value of the optimal combination model at time t may be expressed as formula (7), where y j,t Is the predicted value of the base model j at time t.
Specifically, in step 4, the comprehensive evaluation index calculates the comprehensive score of RMSE and CVaR, and the specific expression is shown in (8). Wherein RMSE Ave And CVaR (CVaR) Ave The performance indexes corresponding to the simple average combination method are respectively shown.
The invention will be further described with reference to the drawings and preferred embodiments. The present disclosure is not limited in this regard.
Step 1: data preprocessing and set partitioning
Firstly, carrying out data preprocessing and standardization processing on historical output data of a new energy station and numerical weather forecast of a corresponding time period. And then dividing the whole data set into a training set, a verification set and a test set, wherein the training set is used for training the base model, the verification set is used for carrying out super-parameter optimization on the combined prediction model, and the test set is used for testing the performance of the combined prediction model. In actual production, the test set is the period to be predicted.
The data used in this embodiment is from 2014 global electric power energy prediction competition, which is 2012, 4 to 2014, 6, and the photovoltaic output data of a station and the corresponding numerical weather forecast data. After data preprocessing and normalization, the mean value of each feature was 0 and the standard deviation was 1. The set is divided into three parts, a training set, a validation set and a test set, as shown in fig. 4.
Step 2: base model training and prediction
M different basic prediction methods are adopted, and prediction results of a verification set and a test set are generated after training according to training set data.
The embodiment selects twelve prediction methods as base models, including seasonal autoregressive moving average SARIMA, exponential smoothing ETS, support vector regression SVR, random forest RF, extreme gradient lifting XGBoost, naive prediction
Multiple linear regression MLR, harmonic dynamic regression DHR, back propagation neural network BPNN, k neighbor KNN, cyclic neural network RNN, the respective input characteristic types are shown in Table 1
TABLE 1 basic model input features
And after the training of the base model in the training set is completed, generating the prediction results of the base model in the verification set and the test set.
Step 3: optimization combination model construction considering condition risk value
And (3) constructing an optimized combination model considering the conditional risk value according to the formulas (2) - (5). The invention obtains the optimal weight of the combined model by carrying out optimization solution on the data in the selected time window, and obtains the weights omega of different base models j (j＝1,2,...,N m ) The predicted value of the optimized combination model at the time t can be obtained by weighted summation of the predicted values of the base model, namely
Step 4: super-parameter optimizing method for optimizing combined model
In this embodiment, the selected comprehensive evaluation index OWA is represented by formula (8). Wherein RMSE Ave And CVaR (CVaR) Ave The performance indexes corresponding to the simple average combination method are respectively shown.
Performing super-parameter optimization on the optimized combination model considering the conditional risk value in a time sequence cross-validation mode on a validation set, wherein the super-parameter comprises a time window length N t Regular term coefficient μ.
The time series cross verification adopts a rolling mode, and a certain length N is selected from a certain moment t Based on which the following data points are predicted and the predicted outcome information of the predicted data points is counted. And then, the starting point of the data subset is moved backwards by the same steps as the number of the predicted data points to form a new data subset, and the subsequent data points are predicted. And so on until the prediction of the whole verification set is realized and the result is recorded. A schematic of the cross-validation process is shown in fig. 2.
The optimization results of this example are shown in tables 2 and 3. The optimal parameters are selected as
μ * ＝0.02。
Table 2 optimization of combined model time window length optimization results
Table 3 optimization combination model regular term coefficient optimization result
Step 5: optimizing combination model prediction result generation
And (3) applying the optimal super parameters obtained in the step (4) on the test set, and generating a prediction result of the optimal combination model on the whole test set by a rolling prediction mode, wherein the prediction process is schematically shown in figure 3. The prediction results may be reported to relevant departments for further reference and use.
In this embodiment, the optimized combination model proposed by the present invention is compared with twelve base models and five other combination prediction models (simple average SA, truncated average TA, tail-end average WA, inverse RMSE average IRMSEA, and weighted average NNA based on neural network), and the results are shown in table 4. In order to intuitively embody the superiority of the optimized combination model provided by the invention, the comparison of the results is shown in a horizontal bar chart form as shown in fig. 5. As can be seen from comparison of results, the optimized combination model provided by the invention is stronger than the base model and other combination models in the aspects of precision, extreme error and comprehensive scoring.
Table 4 comparison of test set results
The result shows that compared with the base model and other combined models, the method can obtain the probability prediction result which has the advantages of both precision and extreme error and has higher quality.
The invention also provides a new energy generated power combination prediction system considering risk avoidance, which comprises the following steps:
the data preprocessing and set dividing module is used for preprocessing and standardizing the historical output data of the new energy station and the numerical weather forecast of the corresponding time period to obtain a whole data set; dividing the whole data set into a training set, a verification set and a test set;
the base model training and predicting module is used for generating corresponding predicting results by adopting M different base models according to training set training;
the optimization combination model construction module is used for obtaining the optimal weight of the combination model by carrying out optimization solution on the data in the selected time window, obtaining weights of different base models, and obtaining an optimization combination model considering the conditional risk value by weighting and summing the predicted values of the base models;
the optimizing module is used for optimizing the super parameters of the optimizing combination model which takes the condition risk value into account in a time sequence cross verification mode;
and the optimal combination model prediction result generation module is used for carrying out rolling prediction on the test set by applying the optimal combination model according to the optimal super parameter obtained by super parameter optimization so as to generate a prediction result corresponding to the time period to be predicted.
The invention provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the steps of the new energy power generation combination prediction method considering risk avoidance and considering multi-interaction function time delay characteristics when executing the computer program.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor realizes the steps of the new energy power generation combination prediction method considering risk avoidance considering multi-interaction function delay characteristics.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical aspects of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.
Claims (10)
1. The new energy generated power combination prediction method considering risk avoidance is characterized by comprising the following steps of:
carrying out data preprocessing and standardization on the historical output data of the new energy station and the numerical weather forecast of the corresponding time period to obtain a whole data set; dividing the whole data set into a training set, a verification set and a test set;
m different base models are adopted, and corresponding prediction results are generated according to training of the training set;
optimizing and solving the data in the selected time window to obtain the optimal weight of the combined model, obtaining weights of different base models, and obtaining an optimized combined model considering the condition risk value by weighting and summing the predicted values of the base models;
performing super-parameter optimization on the optimized combination model considering the condition risk value in a time sequence cross-validation mode;
and rolling prediction is carried out on the test set by applying an optimized combination model according to the optimal super parameters obtained by super parameter optimizing, and a prediction result corresponding to the time period to be predicted is generated.
2. The risk avoidance considered new energy generation power combination prediction method of claim 1, characterized by: the training set is used for training the base model, the verification set is used for carrying out super-parameter optimization on the combined prediction model, and the test set is used for testing the performance of the combined prediction model.
3. The risk avoidance considered new energy generation power combination prediction method of claim 1, characterized by: the base model comprises an autoregressive moving average model, an exponential smoothing model and the like, and a random forest, a cyclic neural network, a support vector machine and other machine learning methods.
4. The risk avoidance considered new energy generation power combination prediction method of claim 1, characterized by: the normalization is to adopt a z-score normalization method, which comprises the following steps:
wherein x is * For standardized data, x is the original data, mu is the mean value of the data, and sigma is the standard deviation of the data.
5. The new energy generation power combination prediction method considering risk avoidance as claimed in claim 1, wherein: the optimal weight of the combined model is obtained by carrying out optimization solution on the data in the selected time window, and the weights omega of different base models are obtained j ，j＝1,2,...,N m Obtaining an optimized combination model considering the condition risk value by weighting and summing the predicted values of the base model; comprising the following steps:
the constraint conditions of the optimization combination model considering the conditional risk value are shown in formulas (2) to (6), the objective function comprises three items of an extreme error item, an accuracy item and a regular item, and the constraint conditions (5) and (6) are related constraints of the conditional risk value; the extreme error adopts a conditional risk value method, and the accuracy adopts a root mean square error method;
obtaining the weight omega of the basic model j ，j＝1,2,...,N m The predicted value of the optimized combination model at time t is represented by equation (7), where y j,t The predicted value of the base model j at the time t;
wherein omega is j For the corresponding weight of the base model j, E is the base model error matrix in the time window, N t For the time window length, lambda is the relative proportion of the extreme error term and the precision term, mu is the weight of the regular term, beta is the preset condition risk value confidence level, and alpha is the auxiliary variable.
6. The new energy generation power combination prediction method considering risk avoidance as claimed in claim 1, wherein: the method for optimizing the super parameters of the optimized combination model considering the condition risk value by the time sequence cross verification comprises the following steps:
performing super-parameter optimization on the optimized combination model considering the condition risk value in a time sequence cross-validation mode on the validation set;
the time series cross verification adopts a rolling mode, and a certain length N is selected from a certain moment t Based on which the data subset of the data points is predicted and the predicted result information of the predicted data points is counted; then, the starting point of the data subset is moved backwards by the step number which is the same as the number of the predicted data points to form a new data subset, and the subsequent data points are predicted; and so on until the prediction of the whole verification set is realized and the result is recorded.
7. The risk avoidance considered new energy generation power combination prediction method of claim 6, characterized by: the super-parameters include a time window length N t And the regular term coefficient mu, wherein the evaluation index is a comprehensive index which simultaneously considers root mean square error and conditional risk value, and the comprehensive evaluation index OWA is selected.
8. The risk avoidance considered new energy generation power combination prediction method of claim 7, wherein: the evaluation index is a comprehensive index which simultaneously considers root mean square error and conditional risk value, and comprises the following contents:
the comprehensive evaluation index OWA is used for calculating the comprehensive score of the prediction result in the two aspects of RMSE and CVaR, and the specific expression is shown in (8);
wherein RMSE Ave And CVaR (CVaR) Ave The performance indexes corresponding to the simple average combination method are respectively shown.
9. The new energy generation power combination prediction method considering risk avoidance as claimed in claim 1, wherein: the generating of the prediction result corresponding to the time period to be predicted further comprises:
and reporting the prediction result for further application or comparison analysis with other prediction models.
10. The new energy generated power combination prediction system considering risk avoidance is characterized by comprising the following steps:
the data preprocessing and set dividing module is used for preprocessing and standardizing the historical output data of the new energy station and the numerical weather forecast of the corresponding time period to obtain a whole data set; dividing the whole data set into a training set, a verification set and a test set;
the base model training and predicting module is used for generating corresponding predicting results by adopting M different base models according to training set training;
the optimization combination model construction module is used for obtaining the optimal weight of the combination model by carrying out optimization solution on the data in the selected time window, obtaining weights of different base models, and obtaining an optimization combination model considering the conditional risk value by weighting and summing the predicted values of the base models;
the optimizing module is used for optimizing the super parameters of the optimizing combination model which takes the condition risk value into account in a time sequence cross verification mode;
and the optimal combination model prediction result generation module is used for carrying out rolling prediction on the test set by applying the optimal combination model according to the optimal super parameter obtained by super parameter optimization so as to generate a prediction result corresponding to the time period to be predicted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310416136.XA CN116384251A (en) | 2023-04-18 | 2023-04-18 | New energy generated power combination prediction method and system considering risk avoidance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310416136.XA CN116384251A (en) | 2023-04-18 | 2023-04-18 | New energy generated power combination prediction method and system considering risk avoidance |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116384251A true CN116384251A (en) | 2023-07-04 |
Family
ID=86963267
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310416136.XA Pending CN116384251A (en) | 2023-04-18 | 2023-04-18 | New energy generated power combination prediction method and system considering risk avoidance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116384251A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117113267A (en) * | 2023-10-25 | 2023-11-24 | 杭州海兴泽科信息技术有限公司 | Prediction model training method based on big data and photovoltaic power generation performance detection method |
-
2023
- 2023-04-18 CN CN202310416136.XA patent/CN116384251A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117113267A (en) * | 2023-10-25 | 2023-11-24 | 杭州海兴泽科信息技术有限公司 | Prediction model training method based on big data and photovoltaic power generation performance detection method |
| CN117113267B (en) * | 2023-10-25 | 2024-02-09 | 杭州海兴泽科信息技术有限公司 | Prediction model training method based on big data and photovoltaic power generation performance detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110969290B (en) | Runoff probability prediction method and system based on deep learning | |
| CN105930916A (en) | Parallel modular neural network-based byproduct gas real-time prediction method | |
| CN108596242A (en) | Power grid meteorology load forecasting method based on wavelet neural network and support vector machines | |
| CN111967183A (en) | Method and system for calculating line loss of distribution network area | |
| CN116384251A (en) | New energy generated power combination prediction method and system considering risk avoidance | |
| CN110751327A (en) | Long-term load combination prediction method based on multiple linear regression and gray Verhulst model | |
| CN107657349B (en) | Method for extracting scheduling rules of staged power generation of reservoir | |
| CN112686693A (en) | Method, system, equipment and storage medium for predicting marginal electricity price of electric power spot market | |
| CN111008790A (en) | Hydropower station group power generation electric scheduling rule extraction method | |
| CN115470862A (en) | Dynamic self-adaptive load prediction model combination method | |
| CN114169434A (en) | Load prediction method | |
| CN111882363A (en) | Sales prediction method, system and terminal | |
| CN111563615A (en) | Load prediction method based on feature analysis and combination learning | |
| CN108537581B (en) | Energy consumption time series prediction method and device based on GMDH selective combination | |
| CN111832785A (en) | Method and system for predicting electric energy substitution potential | |
| CN111697560B (en) | Method and system for predicting load of power system based on LSTM | |
| CN112836885A (en) | Combined load prediction method, combined load prediction device, electronic equipment and storage medium | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN107748931A (en) | A kind of income of electricity charge Forecasting Methodology based on least square method | |
| CN111177881A (en) | Random production simulation method for power system containing photo-thermal-photovoltaic power generation | |
| CN110298456A (en) | Plant maintenance scheduling method and device in group system | |
| CN112581311B (en) | Method and system for predicting long-term output fluctuation characteristics of aggregated multiple wind power plants | |
| CN113723717B (en) | Method, device, equipment and readable storage medium for predicting short-term load before system day | |
| CN115759455A (en) | Load probability density prediction method based on time sequence Gaussian mixture density network | |
| CN112508280B (en) | Method and system for evaluating and optimizing real-time performance of comprehensive energy system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |