CN113934158A - Electric arc furnace modeling method based on improved random forest - Google Patents
Electric arc furnace modeling method based on improved random forest Download PDFInfo
- Publication number
- CN113934158A CN113934158A CN202111223042.8A CN202111223042A CN113934158A CN 113934158 A CN113934158 A CN 113934158A CN 202111223042 A CN202111223042 A CN 202111223042A CN 113934158 A CN113934158 A CN 113934158A
- Authority
- CN
- China
- Prior art keywords
- arc furnace
- electric arc
- random forest
- model
- electric
- 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
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B17/00—Systems involving the use of models or simulators of said systems
- G05B17/02—Systems involving the use of models or simulators of said systems electric
Abstract
The invention relates to the field of electrical engineering, and discloses an electric arc furnace modeling method based on an improved random forest. The invention learns the Stacking algorithm and combines through a secondary learner. Compared with a model established by a single decision tree, the electric arc furnace model based on the improved random forest has extremely high accuracy, due to the introduction of randomness, the model is not easy to be over-fitted and has good anti-noise capability, the model can process high-dimensional data and is not used as feature selection, the workload of data preprocessing can be greatly reduced, and the robustness is further improved due to the introduction of a Stacking method in the model. In conclusion, the electric arc furnace modeling method based on the improved random forest has excellent generalization capability, can reflect the influence of various factors in electric arc furnace production application on the energy consumption, and can provide a good reference for modeling other types of industrial loads.
Description
Technical Field
The invention relates to the field of electrical engineering, in particular to an electric arc furnace modeling method based on improved random forest.
Background
With the rapid development of national economy and science and technology, novel loads such as electric vehicles and energy storage are continuously emerging, and great challenges are provided for the supply and demand balance of a power system. Faced with new development problems, the traditional "source-follow-up" regulation mode has not been able to satisfy the new power utilization form which is variable and increasing. In order to solve the problem, the nation emphasizes the enhancement of demand-side management and the improvement of load regulation and control capability. Among a plurality of power loads, the industrial load has large capacity, is distributed intensively and has larger regulation potential. Because a quantitative analysis model is not available for different types of industrial loads, the existing batch load control method is simple and rough and is easy to cause serious influence on the production society. Meanwhile, the lack of planning means for industrial load to participate in power grid interaction is caused, and the regulation and control potential is difficult to fully exploit.
With the development of artificial intelligence, machine learning and deep learning are widely combined in various fields, especially the electrical field, and are combined with the traditional technical method in the electrical field, so that some problems which are difficult to solve in the past are properly treated. The invention provides an electric arc furnace modeling method based on an improved random forest to reflect the relation between electric quantity characteristics in electric arc furnace production application and non-electric quantity production factors Shu electric arc furnace energy consumption by combining an ensemble learning method in machine learning. The modeling method has good generalization capability and can provide a good reference for modeling other types of industrial loads.
Disclosure of Invention
In order to solve the above mentioned disadvantages in the background art, the present invention provides an arc furnace modeling method based on improved random forest
The purpose of the invention can be realized by the following technical scheme:
an electric arc furnace modeling method based on improved random forests, comprising the following steps:
step 1: researching electric and non-electric influence factors of the electric arc furnace and collecting data of the factors;
step 2: cleaning and dividing the collected original data;
and step 3: and (4) building an electric arc furnace model based on the improved random forest.
Further, the electric and non-electric influencing factors of the electric arc furnace in the step 1 comprise: voltage and current of electric arc furnace load, capacity of electric furnace transformer, oxygen blowing amount, energy recovery rate and raw material structure.
Further, the method for cleaning and dividing the collected original data comprises the following steps:
step 2.1: adopting dropna in pandas to carry out deletion value filtration;
step 2.2: for abnormal values in the data, the collected raw data is preprocessed by Local Outlier Factor (LOF)
Further, the LOF method is a high-precision outlier detection method based on density, and each data is assigned with an outlier factor LOF depending on the neighborhood density, and if LOF >1, the data point is an abnormal point, and if LOF >1, the data point is a normal data point.
Further, the electric arc furnace model building method based on the improved random forest comprises the following steps:
step 3.1: randomly dividing an initial training set D into k sets D with the same size1、D2、.....DkLet DjFor the jth test set, the test set,
is a corresponding training set;
step 3.2: by using
Training T primary learners;
step 3.3: testing with jth test set, for test set DjEach sample x in (1)iVector (z) composed of output results of T primary learnersi1,zi2,...,ziT) Forms part of the secondary learner input;
step 3.4: repeating the step 3.2 and the step 3.3 to generate a secondary training set;
step 3.5: training a secondary learner by utilizing a secondary training set;
step 3.6: and (5) testing the model.
Further, the method for testing the model comprises the following steps:
step 3.6.1: testing the trained model on a test set, wherein the test set covers 120 groups of data, and the content comprises voltage and current of an electric arc furnace load, capacity of an electric furnace transformer, oxygen blowing amount, energy recovery rate and a raw material structure;
step 3.6.2: taking the test set as the input of the model obtained by training to obtain the energy consumption estimated value of the electric arc furnace as an output structure;
step 3.6.3: and measuring the difference between the output of the model and an actual value by adopting the root mean square error, and comparing the difference with the test results of a decision tree, a random forest and an extreme random forest.
The invention has the beneficial effects that:
1. the decision tree is used as the primary learner, and data sample disturbance and input attribute disturbance are introduced in the training of the primary learner, so that the diversity of the primary learner is greatly improved, and the integration effect of the model is improved. And due to the introduction of randomness, the model is not easy to over-fit and has good anti-noise capability.
2. The invention adopts an integrated learning method, and compared with a model formed by a single decision tree, the proposed model has higher accuracy.
3. The combination module of the invention is a secondary learner, the learning algorithm is multi-response linear regression, and meanwhile, in order to prevent overfitting, a cross validation mode is adopted in the training stage of the primary learner, and the design enables the robustness of the model to be improved.
4. The electric arc furnace modeling method based on the improved random forest has excellent generalization capability and can provide a good reference for modeling other types of industrial loads.
Drawings
The invention will be further described with reference to the accompanying drawings.
FIG. 1 is a schematic diagram of the electric arc furnace model learning based on the improved random forest according to the present invention;
FIG. 2 is a flow chart of an electric arc furnace modeling method based on improved random forest according to the present invention;
FIG. 3 is an ensemble learning algorithm of the proposed model of the present invention;
FIG. 4 is a test result of the improved random forest model in the test set, where the solid line part is the actual value and the dotted line part is the model result;
FIG. 5 is a test result of a decision tree on a test set.
FIG. 6 is a test result of a random forest on a test set.
FIG. 7 is the test results of an extreme random forest on the test set.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1 to 3, an electric arc furnace modeling method based on improved random forest includes the following steps:
step 1: researching electric and non-electric influence factors of the electric arc furnace and collecting data of the factors;
the electric arc furnace is mainly applied to steel smelting production, and steel and alloy with various components are smelted by melting ores and metals at high temperature generated by electrode arc. The power consumption is huge and generally accounts for more than 50% of the total power consumption of an enterprise. In steel smelting, there are many factors affecting the energy consumption of the electric arc furnace, and the main electrical and non-electrical affecting factors include the voltage and current of the electric arc furnace load, the capacity of the electric furnace transformer, the oxygen blowing amount, the energy recovery rate and the raw material structure without considering human factors and production schemes. The electric arc furnace steel making uses scrap steel as main raw material and pig iron, direct reduced iron and molten iron as auxiliary materials, and the proportion of different raw materials has different influences on the power consumption. In which the ratio of molten iron to pig iron is inversely related to energy consumption, and an increase in the proportion of direct reduced iron leads to an increase in the energy consumption of the electric arc furnace.
Step 2: cleaning and dividing the collected original data;
because the initial data has many attributes and large sample size, including a considerable number of missing values and abnormal values, data cleaning is required before modeling. Aiming at the missing value, due to the large sample amount, dropna is directly adopted for missing value filtering, and abnormal points in the data set are removed by a local outlier detection method. The final training set D contained 1300 sets of data and the test set T contained 120 sets of data.
And step 3: and (4) building an electric arc furnace model based on the improved random forest.
The electric and non-electric influencing factors of the electric arc furnace in the step 1 comprise: voltage and current of electric arc furnace load, capacity of electric furnace transformer, oxygen blowing amount, energy recovery rate and raw material structure.
The method for cleaning and dividing the collected original data comprises the following steps:
step 2.1: adopting dropna in pandas to carry out deletion value filtration;
step 2.2: for abnormal values in the data, the collected raw data is preprocessed by Local Outlier Factor (LOF)
The LOF method is a high-precision outlier detection method based on density, each data is allocated with an outlier LOF depending on neighborhood density, if LOF >1, the data point is an abnormal point, and if LOF is close to 1, the data point is a normal data point.
According to the method, an electric arc furnace model is improved and constructed on the basis of a common random forest, and in order to enhance the diversity of the primary learner, when each node selects the division attribute, a random selection method is adopted, so that compared with the traditional random forest, the diversity of the obtained primary learner is stronger, and the final integration effect is better. In the combination strategy of the model, the invention learns the Stacking algorithm, the output of the primary learner forms a secondary training set, the secondary learner is used for combination, and the secondary learner adopts multi-response linear regression as the learning algorithm. Therefore, the model building flow is as follows:
step 3.1: randomly dividing an initial training set D into k sets D with the same size1、D2、.....DkLet DjFor the jth test set, the test set,
is a corresponding training set;
step 3.2: by using
Training T primary learners;
step 3.3: testing with jth test set, for test set DjEach sample x in (1)iVector (z) composed of output results of T primary learnersi1,zi2,...,ziT) Forms part of the secondary learner input;
step 3.4: repeating the step 3.2 and the step 3.3 to generate a secondary training set;
step 3.5: training a secondary learner by utilizing a secondary training set;
step 3.6: and (5) testing the model.
The model testing method comprises the following steps:
step 3.6.1: testing the trained model on a test set, wherein the test set covers 120 groups of data, and the content comprises voltage and current of an electric arc furnace load, capacity of an electric furnace transformer, oxygen blowing amount, energy recovery rate and a raw material structure;
step 3.6.2: taking the test set as the input of the model obtained by training to obtain the energy consumption estimated value of the electric arc furnace as an output structure;
step 3.6.3: and measuring the difference between the output of the model and an actual value by adopting the root mean square error, and comparing the difference with the test results of a decision tree, a random forest and an extreme random forest.
The test results are shown in FIGS. 4-7, in which the solid line part is the actual value and the dashed line part is the model result. The following table shows the root mean square error of the test results of each model, and the model provided by the invention has the highest accuracy and is superior to a single decision tree, a traditional random forest and an extreme random forest by comparison.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed.
Claims (6)
1. An electric arc furnace modeling method based on improved random forests is characterized by comprising the following steps:
step 1: researching electric and non-electric influence factors of the electric arc furnace and collecting data of the factors;
step 2: cleaning and dividing the collected original data;
and step 3: and (4) building an electric arc furnace model based on the improved random forest.
2. The method as claimed in claim 1, wherein the electric arc furnace modeling method based on the improved random forest is characterized in that the electric and non-electric influence factors of the electric arc furnace in the step 1 comprise: voltage and current of electric arc furnace load, capacity of electric furnace transformer, oxygen blowing amount, energy recovery rate and raw material structure.
3. The electric arc furnace modeling method based on the improved random forest as claimed in claim 1, wherein the method for cleaning and dividing the collected raw data comprises the following steps:
step 2.1: adopting dropna in pandas to carry out deletion value filtration;
step 2.2: for abnormal values in the data, the collected raw data is preprocessed by Local Outlier Factor detection (LOF).
4. The method of claim 3, wherein the LOF method is a density-based high-precision outlier detection method, and each datum is assigned an outlier LOF that depends on the neighborhood density if LOF >1 and a normal datum if LOF is close to 1.
5. The electric arc furnace modeling method based on the improved random forest as claimed in claim 1, wherein the electric arc furnace model building step based on the improved random forest is as follows:
step 3.1: randomly dividing an initial training set D into k sets D with the same size1、D2、.....DkLet DjFor the jth test set, the test set,
is a corresponding training set;
step 3.2: by using
Training T primary learners;
step 3.3: testing with jth test set, for test set DjEach sample x in (1)iVector (z) composed of output results of T primary learnersi1,zi2,...,ziT) Forms part of the secondary learner input;
step 3.4: repeating the step 3.2 and the step 3.3 to generate a secondary training set;
step 3.5: training a secondary learner by utilizing a secondary training set;
step 3.6: and (5) testing the model.
6. The method for electric arc furnace modeling based on improved random forest as claimed in claim 5, wherein the method for model testing comprises the steps of:
step 3.6.1: testing the trained model on a test set, wherein the test set covers 120 groups of data, and the content comprises voltage and current of an electric arc furnace load, capacity of an electric furnace transformer, oxygen blowing amount, energy recovery rate and a raw material structure;
step 3.6.2: taking the test set as the input of the model obtained by training to obtain the energy consumption estimated value of the electric arc furnace as an output structure;
step 3.6.3: and measuring the difference between the output of the model and an actual value by adopting the root mean square error, and comparing the difference with the test results of a decision tree, a random forest and an extreme random forest.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111223042.8A CN113934158A (en) | 2021-10-20 | 2021-10-20 | Electric arc furnace modeling method based on improved random forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111223042.8A CN113934158A (en) | 2021-10-20 | 2021-10-20 | Electric arc furnace modeling method based on improved random forest |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113934158A true CN113934158A (en) | 2022-01-14 |
Family
ID=79280882
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111223042.8A Pending CN113934158A (en) | 2021-10-20 | 2021-10-20 | Electric arc furnace modeling method based on improved random forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113934158A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106019093A (en) * | 2016-05-15 | 2016-10-12 | 北华大学 | Online soft measurement method for three-phase arc furnace arc length |
| CN109446189A (en) * | 2018-10-31 | 2019-03-08 | 成都天衡智造科技有限公司 | A kind of technological parameter outlier detection system and method |
| CN110503251A (en) * | 2019-08-12 | 2019-11-26 | 江苏方天电力技术有限公司 | A kind of non-festivals or holidays load forecasting method based on Stacking algorithm |
| CN111126423A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术（北京）有限公司 | Feature set acquisition method and device, computer equipment and medium |
| CN111157899A (en) * | 2020-01-20 | 2020-05-15 | 南京邮电大学 | Method for estimating SOC of battery based on model fusion idea |
| US20200279025A1 (en) * | 2019-02-28 | 2020-09-03 | Kalibrate Technologies Limited | Machine-learned model selection network planning |
-
2021
- 2021-10-20 CN CN202111223042.8A patent/CN113934158A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106019093A (en) * | 2016-05-15 | 2016-10-12 | 北华大学 | Online soft measurement method for three-phase arc furnace arc length |
| CN109446189A (en) * | 2018-10-31 | 2019-03-08 | 成都天衡智造科技有限公司 | A kind of technological parameter outlier detection system and method |
| CN111126423A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术（北京）有限公司 | Feature set acquisition method and device, computer equipment and medium |
| US20200279025A1 (en) * | 2019-02-28 | 2020-09-03 | Kalibrate Technologies Limited | Machine-learned model selection network planning |
| CN110503251A (en) * | 2019-08-12 | 2019-11-26 | 江苏方天电力技术有限公司 | A kind of non-festivals or holidays load forecasting method based on Stacking algorithm |
| CN111157899A (en) * | 2020-01-20 | 2020-05-15 | 南京邮电大学 | Method for estimating SOC of battery based on model fusion idea |
Non-Patent Citations (1)
| Title |
|---|
| 李正光 等: "基于多维特征量融合的配电网拓扑异常溯源与应用模型研究", 浙江电力, vol. 39, no. 7 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Multi-objective optimal reactive power dispatch of power systems by combining classification-based multi-objective evolutionary algorithm and integrated decision making | |
| Arya et al. | Fuzzy gain scheduling controllers for automatic generation control of two-area interconnected electrical power systems | |
| CN110112767B (en) | Load source optimization control method for peak regulation of wide-area polymorphic demand side load participation system | |
| Wu et al. | Source-network-storage joint planning considering energy storage systems and wind power integration | |
| Zhang et al. | MOEA/D with many-stage dynamical resource allocation strategy to solution of many-objective OPF problems | |
| CN114065483A (en) | Hydrogen production system efficiency optimization method and device, computer equipment and storage medium | |
| Sahoo et al. | Self-adaptive fuzzy-PID controller for AGC study in deregulated Power System | |
| CN109934476B (en) | Micro-grid source-storage joint planning multi-strategy evolution game analysis method based on subject limited rational decision | |
| He et al. | Biobjective optimization-based frequency regulation of power grids with high-participated renewable energy and energy storage systems | |
| Cao et al. | Variable universe fuzzy expert system for aluminum electrolysis | |
| CN113934158A (en) | Electric arc furnace modeling method based on improved random forest | |
| Agrawal et al. | Optimal location of static VAR compensator using evolutionary optimization techniques | |
| CN108734349A (en) | Distributed generation resource addressing constant volume optimization method based on improved adaptive GA-IAGA and system | |
| CN108718084B (en) | Power supply and power grid coordination planning method suitable for electric power market reformation | |
| CN111049151A (en) | NSGA 2-based two-stage optimization algorithm for power distribution network voltage | |
| Ageev et al. | Application of the genetic algorithm during electric network mode optimization | |
| Qin et al. | Coot algorithm for optimal carbon–energy combined flow of power grid with aluminum plants | |
| CN116526496B (en) | Novel auxiliary decision-making method for power system load control | |
| CN116485042B (en) | Method and device for optimizing park energy system operation based on load clustering | |
| Zhu et al. | Many-objective power flow optimization problems based on an improved MOEA/D with dynamical resource allocation strategy | |
| Liu et al. | Enterprise financial risk prevention and control and data analysis method based on blockchain technology | |
| CN116127345B (en) | Converter steelmaking process mode design method based on deep clustering generation countermeasure network | |
| Borujeni et al. | Fuel cell voltage control using neural network based on model predictive control | |
| CN111666666A (en) | Multi-objective optimized scheduling method for thermal load of biomass thermal power plant unit | |
| Sun et al. | A Bcs-Gde Multi-Objective Optimization Algorithm for Cchp Model with Decision Strategies |
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 |