WO2020134801A1 - Energy-saving optimization method of cement raw material vertical mill system - Google Patents

Energy-saving optimization method of cement raw material vertical mill system Download PDF

Info

Publication number
WO2020134801A1
WO2020134801A1 PCT/CN2019/120980 CN2019120980W WO2020134801A1 WO 2020134801 A1 WO2020134801 A1 WO 2020134801A1 CN 2019120980 W CN2019120980 W CN 2019120980W WO 2020134801 A1 WO2020134801 A1 WO 2020134801A1
Authority
WO
WIPO (PCT)
Prior art keywords
variable
data
optimal
historical operation
code
Prior art date
Application number
PCT/CN2019/120980
Other languages
French (fr)
Chinese (zh)
Inventor
刘煜
孙再连
钟骥华
Original Assignee
厦门邑通软件科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 厦门邑通软件科技有限公司 filed Critical 厦门邑通软件科技有限公司
Publication of WO2020134801A1 publication Critical patent/WO2020134801A1/en

Links

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B02—CRUSHING, PULVERISING, OR DISINTEGRATING; PREPARATORY TREATMENT OF GRAIN FOR MILLING
    • B02C—CRUSHING, PULVERISING, OR DISINTEGRATING IN GENERAL; MILLING GRAIN
    • B02C23/00—Auxiliary methods or auxiliary devices or accessories specially adapted for crushing or disintegrating not provided for in preceding groups or not specially adapted to apparatus covered by a single preceding group
    • B02C23/02—Feeding devices
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B02—CRUSHING, PULVERISING, OR DISINTEGRATING; PREPARATORY TREATMENT OF GRAIN FOR MILLING
    • B02C—CRUSHING, PULVERISING, OR DISINTEGRATING IN GENERAL; MILLING GRAIN
    • B02C25/00—Control arrangements specially adapted for crushing or disintegrating
    • 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
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]

Definitions

  • the invention relates to the technical field of energy saving and consumption reduction, in particular to an energy saving optimization method of a cement raw material vertical mill system.
  • the second difficulty is that it requires workers to operate according to the recommendations of experts, that is, according to the approximate operation setting values given by man. Different experts have different operation setting values, and they cannot achieve energy saving and consumption reduction.
  • the present invention provides an energy-saving optimization method for a cement raw material vertical mill system. Without changing any structure and principle of the production equipment, without adding additional measuring points, or without affecting normal production, the machine learning Method, intelligently provide safe, convenient and reasonable auxiliary decision-making.
  • the method includes: collecting historical operation data to form multiple historical operation models, the historical operation data including controllable variables, the controllable variables including mill grinding pressure, mill inlet negative pressure, mill outlet negative pressure, Hot air valve opening, cold air valve opening, mill inlet temperature, mill outlet temperature, fan valve opening;
  • the feeding amount of the mill is divided into zones;
  • the historical operations of the same feed volume partition are sorted to obtain the optimal historical operation records of a feed volume partition, and the optimal historical operation records of all feed volume partitions are merged to form the most Excellent recommendation form;
  • the optimal operation recommendation of the device is obtained according to the real-time operating conditions and the optimal recommendation table.
  • the historical operation data is subjected to box-line graph distribution statistics through machine learning to obtain an upper limit and a lower limit, and excluding abnormal data other than the upper limit and the lower limit.
  • the specific logic is as follows:
  • VMIN> ⁇ -3 ⁇ and VMAX> ⁇ +3 ⁇ the range of (VMIN, ⁇ +3 ⁇ ) is taken as the data screening rule
  • VMIN ⁇ -3 ⁇ and VMAX ⁇ +3 ⁇ the range of ( ⁇ -3 ⁇ , VMAX) is taken as the data screening rule
  • VMIN> ⁇ -3 ⁇ and VMAX ⁇ +3 ⁇ the range of (VMIN, VMAX) is taken as the data screening rule
  • VMIN is the minimum value of the variable, the maximum value of the VMAX variable, the mean of the ⁇ variable, and the standard deviation of the ⁇ variable.
  • the historical operation data is discretized, and then the historical operation model is adaptively encoded with high quality.
  • the frequency of each code in the same feed volume partition is counted and sorted in descending order according to the frequency; the single-electromechanical consumption corresponding to the same code is counted, the same code is integrated and deduplicated to form an unique code, and the corresponding code
  • the single electromechanical consumption is equal to the average of all the single electromechanical consumptions with the same encoding. Obtain the single electromechanical consumption corresponding to the top ten unique codes in descending order, and select the lowest single electromechanical consumption as the best historical operation record of the feeding volume partition where it is obtained. The corresponding optimal recommendation table.
  • the coding corresponds to the historical operation model one by one, and the coding is a mapping of each variable to the model, which can locate the model, that is, the corresponding model can be quickly found according to the variable, and the coding method can not only greatly reduce the memory of the sample Space, improve model training speed, and can greatly improve learning accuracy.
  • the code represents the grinding pressure of the mill, the negative pressure of the mill inlet, the negative pressure of the mill outlet, the opening of the hot air valve, the opening of the cold air valve, the inlet temperature of the mill, the outlet temperature of the mill, the opening of the fan valve and other variables. data range.
  • coding rounding function ((variable-minimum value of variable) / variable step size),
  • This application uses adaptive high-quality coding, and its minimum value and variable step size will not incorporate any subjective experience, which depends on the data itself and deviates from artificial fixed subjective experience.
  • variable step size depends on the amount of change in data accuracy:
  • the recommended optimal historical operation record is decoded, and the decoding calculation formula is as follows:
  • Variable variable code * variable step + minimum value of variable.
  • the operation model generated by the operation forms a real-time encoding, and matches the real-time encoding with the unique encoding .
  • the matching distance is zero and the single-electromechanical consumption of the real-time encoding is lower than that of the single-coded single-electromechanical consumption, it is judged as a better operation record, the operation record is stored, and the average value of the single-electromechanical consumption is recalculated to obtain the unique code
  • the new single electromechanical consumption and then compare the frequency and single electromechanical consumption with other unique codes of the feed volume partition, so as to update the optimal historical operation record of the feed volume partition; when the match does not match the distance of zero alone
  • the real-time code is listed as a new unique code, and then compared with the frequency of the other unique codes of the feed volume partition and the single electro
  • the present invention has the following advantages:
  • Adopting adaptive filtering rules and adaptive high-quality coding which deviates from the artificial fixed subjective experience, does not incorporate any subjective viewing methods or experience after observing data statistics, first adaptively adjust the filtering rules to extract safe and normal Value data, and then adjust the coding formula parameters through adaptive learning, so as to achieve the potential value of historical data, and the recommended optimization scheme is objective, reasonable, reliable and safe;
  • the online push speed of the present invention is fast, and it is only necessary to determine which feed volume partition it belongs to.
  • the present invention has fast learning efficiency and includes an online update function.
  • iterating the optimal recommendation table there is no need to learn for all the feed volume partition data, to obtain the optimal operation record of each feed volume partition, only to determine the corresponding feed volume partition according to the feed volume, and then to the corresponding Re-learn the partition data, and then just update the optimal historical operation record of the partition in the optimal recommendation table.
  • Embodiment 1 is a schematic flowchart of Embodiment 1 of an energy-saving optimization method of a cement raw material vertical mill system of the present invention
  • Embodiment 1 An energy-saving optimization method of a cement raw material vertical mill system, which does not change any structure and principle of production equipment, does not add additional measurement points, and does not affect normal production. It provides intelligently through machine learning methods. Safe, convenient and reasonable auxiliary decision-making.
  • the method includes: collecting historical operation data to form a plurality of historical operation models, the historical operation data including controllable variables, regular filtering variables and variables with less correlation, the controllable variables including mill grinding pressure, grinding Machine inlet negative pressure, mill outlet negative pressure, hot air valve opening, cold air valve opening, mill inlet temperature, mill outlet temperature, fan valve opening, the rule filtering variables include mill vibration value and powder separator
  • this embodiment only considers how to optimize related and controllable variables as optimization suggestions for energy saving and consumption reduction, so as to reduce energy consumption of the mill and circulating fan without affecting normal production.
  • the feeding amount of the mill is divided into zones.
  • the historical operations of the same feed volume partition are sorted to obtain the optimal historical operation records of a feed volume partition, and the optimal historical operation records of all feed volume partitions are merged to form the most Excellent recommendation form.
  • Embodiment 2 Based on the first embodiment, the historical operation data is analyzed and processed, specifically: before obtaining the optimal historical operation record, the box operation graph distribution statistics are performed on the historical operation data by machine learning, Obtain the upper and lower limits, remove the abnormal data except the upper and lower limits, and then, through the normal distribution of self-learning data, adaptively determine different data screening rules, the specific logic is as follows:
  • VMIN> ⁇ -3 ⁇ and VMAX> ⁇ +3 ⁇ the range of (VMIN, ⁇ +3 ⁇ ) is taken as the data screening rule
  • VMIN ⁇ -3 ⁇ and VMAX ⁇ +3 ⁇ the range of ( ⁇ -3 ⁇ , VMAX) is taken as the data screening rule
  • VMIN> ⁇ -3 ⁇ and VMAX ⁇ +3 ⁇ the range of (VMIN, VMAX) is taken as the data screening rule
  • VMIN is the minimum value of the variable, the maximum value of the VMAX variable, the mean of the ⁇ variable, the standard deviation of the ⁇ variable, and the rigidity of the initial target mill current and the high-voltage current of the circulating fan meet the following conditions:
  • the historical operation data is discretized, and then the historical operation model is adaptively encoded with high quality, such as: X11, X12, X13,..., X1N, see Figure 1.
  • the frequency of each code in the same feed volume partition is counted and sorted in descending order according to the frequency; the single electromechanical consumption Y1 corresponding to the same code is counted, and the same code is integrated and deduplicated to form an unique code, and the unique code corresponds to
  • the single electromechanical consumption Y2 is equal to the average of all the single electromechanical consumption Y1 of the same code, and the single electromechanical consumption Y12 corresponding to the top ten unique codes in descending order is obtained.
  • the lowest single electromechanical consumption is selected as the optimal historical operation of the feed volume partition Record Ybest (Y optimal) and obtain the corresponding optimal recommendation table.
  • the coding corresponds to the historical operation model one by one, and the coding is a mapping of each variable to the model, which can locate the model, that is, the corresponding model can be quickly found according to the variable, and the coding method can not only greatly reduce the memory of the sample Space, improve model training speed, and can greatly improve learning accuracy.
  • the code represents the grinding pressure of the mill, the negative pressure of the mill inlet, the negative pressure of the mill outlet, the opening of the hot air valve, the opening of the cold air valve, the inlet temperature of the mill, the outlet temperature of the mill, the opening of the fan valve and other variables. data range.
  • coding rounding function ((variable-minimum value of variable) / variable step size),
  • This application uses adaptive high-quality coding, and its minimum value and variable step size will not incorporate any subjective experience, which depends on the data itself and deviates from artificial fixed subjective experience.
  • variable step size depends on the amount of change in data accuracy:
  • the recommended optimal historical operation record is decoded, and the decoding calculation formula is as follows:
  • Variable variable code * variable step + minimum value of variable.
  • the operation model generated by the operation forms a real-time code, and the real-time code is matched with the unique code.
  • the matching distance is zero and the single-electromechanical consumption of the real-time code is lower than that of the single-coded single electromechanical, it is judged as a better operation record.
  • the operation record is stored, and the average value of the single electromechanical consumption is recalculated, that is, the new single electromechanical consumption of the unique code is obtained, and then the frequency and the single electromechanical consumption are compared with the other unique codes of the feed volume partition to update the feed.
  • the best historical operation record of the volume partition
  • the real-time code is listed as a new unique code, and then compared with the other unique codes in the feed volume partition where the frequency and single electrical consumption are compared to update the feed volume.
  • the best historical operation record of the partition is listed as a new unique code, and then compared with the other unique codes in the feed volume partition where the frequency and single electrical consumption are compared to update the feed volume. The best historical operation record of the partition.
  • Adopting adaptive filtering rules and adaptive high-quality coding which deviates from the artificial fixed subjective experience, does not incorporate any subjective viewing methods or experience after observing data statistics, first adaptively adjust the filtering rules to extract safe and normal Value data, and then adjust the coding formula parameters through adaptive learning, so as to achieve the potential value of historical data, and the recommended optimization scheme is objective, reasonable, reliable and safe;
  • the online push speed of the present invention is fast. Only by judging which feed volume partition it belongs to, the corresponding recommendation recommendations can be quickly and efficiently extracted from the stored optimal recommendation table. It has fast efficiency and recommendation speed, which is far from satisfactory The need for a recommendation every 3s;
  • the present invention has fast learning efficiency and includes an online update function.
  • iterating the optimal recommendation table there is no need to learn for all the feed volume partition data, to obtain the optimal operation record of each feed volume partition, only to determine the corresponding feed volume partition according to the feed volume, and then to the corresponding Re-learn the partition data, and then just update the optimal historical operation record of the partition in the optimal recommendation table.

Landscapes

  • Engineering & Computer Science (AREA)
  • Food Science & Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Quality & Reliability (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Feedback Control In General (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

An energy-saving optimization method of a cement raw material vertical mill system. Safe, convenient and reasonable auxiliary decision-making is smartly provided by a machine learning method without changing any structure and principle of a production equipment, adding additional measuring spots and affecting normal production. The method comprises: acquiring historical operational data and form multiple historical operational models, the historical operational data comprising controllable variables, the controllable variables comprising a mill milling pressure, a mill inlet vacuum, a mill outlet vacuum, a hot blast valve opening, a cold blast valve opening, a mill inlet temperature, a mill outlet temperature and a fan valve opening; dividing feeding amount partitions; sorting the historical operations of the same feeding amount partition according to the power consumption per unit, to obtain the optimal historical operational record of one feeding amount partition, and merging the optimal historical operational records of all the feeding amount partitions to form an optimal recommendation form; and acquiring optimal operational suggestion of the device according to a real-time operational condition and the optimal recommendation form.

Description

一种水泥原料立磨***的节能优化方法Energy-saving optimization method of cement raw material vertical mill system 技术领域Technical field
本发明涉及节能降耗技术领域，尤其涉及一种水泥原料立磨***的节能优化方法。The invention relates to the technical field of energy saving and consumption reduction, in particular to an energy saving optimization method of a cement raw material vertical mill system.
背景技术Background technique
水泥原料立磨***生产合格且稳定产量下的节能优化是水泥厂关注的重要课题。Cement raw material vertical mill system production is qualified and energy-saving optimization under stable output is an important topic of concern for cement plants.
技术问题technical problem
其难点一，在于实时喂料量波动大，工人需要实时的进行人工微调，根据传统管理及操作方式，只能做到产量正常，无法做到耗能的最优；One of the difficulties is that the real-time feeding amount fluctuates greatly, and workers need to make manual fine-tuning in real time. According to traditional management and operation methods, only normal output can be achieved, and energy consumption cannot be optimized;
其难点二，需要工人根据专家建议，即根据人为给出的大概的操作设定值操作，不同的专家，其给出的操作设定值不同，同样无法做到节能降耗。The second difficulty is that it requires workers to operate according to the recommendations of experts, that is, according to the approximate operation setting values given by man. Different experts have different operation setting values, and they cannot achieve energy saving and consumption reduction.
因此有必要提出一种自适应的低成本、安全、便捷的智能化辅助决策方案，帮助水泥厂提供一个较准确可靠的优化建议，从而达到智能化操作，且稳定产量下的节能甚至优产的效果。Therefore, it is necessary to propose an adaptive low-cost, safe, and convenient intelligent assistant decision-making plan to help the cement plant provide a more accurate and reliable optimization proposal, so as to achieve intelligent operation, and energy-saving or even optimal production under stable production. effect.
技术解决方案Technical solution
本发明为解决上述技术问题，提供了一种水泥原料立磨***的节能优化方法，在不改变生产设备任何结构和原理、不增加额外测点、不影响正常生产的前提下，通过机器学习的方法，智能化的提供安全、便捷、合理的辅助决策。In order to solve the above technical problems, the present invention provides an energy-saving optimization method for a cement raw material vertical mill system. Without changing any structure and principle of the production equipment, without adding additional measuring points, or without affecting normal production, the machine learning Method, intelligently provide safe, convenient and reasonable auxiliary decision-making.
所述方法包括：采集历史操作数据，形成多个历史操作模型，所述历史操作数据包括可控变量，所述可控变量包括磨机研磨压力、磨机进口负压、磨机出口负压、热风阀门开度、冷风阀门开度、磨机入口温度、磨机出口温度、风机阀门开度；The method includes: collecting historical operation data to form multiple historical operation models, the historical operation data including controllable variables, the controllable variables including mill grinding pressure, mill inlet negative pressure, mill outlet negative pressure, Hot air valve opening, cold air valve opening, mill inlet temperature, mill outlet temperature, fan valve opening;
根据磨机的喂料量划分出喂料量分区；According to the feeding amount of the mill, the feeding amount is divided into zones;
根据单机电耗量的高低，对同一个喂料量分区的历史操作进行排序，获得一个喂料量分区的最优历史操作记录，将所有喂料量分区的最优历史操作记录合并，形成最优推荐表；According to the level of single electromechanical consumption, the historical operations of the same feed volume partition are sorted to obtain the optimal historical operation records of a feed volume partition, and the optimal historical operation records of all feed volume partitions are merged to form the most Excellent recommendation form;
进一步的，根据实时工况和所述最优推荐表获得设备的最优操作建议。Further, the optimal operation recommendation of the device is obtained according to the real-time operating conditions and the optimal recommendation table.
进一步的，在获得最优历史操作记录前，通过机器学习的方式对所述历史操作数据进行箱线图分布统计，得到上限和下限，剔除上限和下限之外的异常数据。Further, before obtaining the optimal historical operation record, the historical operation data is subjected to box-line graph distribution statistics through machine learning to obtain an upper limit and a lower limit, and excluding abnormal data other than the upper limit and the lower limit.
进一步的，剔除历史操作数据中的异常数据后，通过自主学习数据的正态分布情况,自适应确定不同数据筛选规则，具体逻辑如下：Further, after removing the abnormal data in the historical operation data, through the normal distribution of the self-learning data, adaptively determine different data screening rules, the specific logic is as follows:
若VMIN<μ-3σ且VMAX>μ+3σ，则取(μ-3σ,μ+3σ)范围作为数据筛选规则；If VMIN<μ-3σ and VMAX>μ+3σ, the range of (μ-3σ, μ+3σ) is taken as the data screening rule;
若VMIN>μ-3σ且VMAX>μ+3σ，则取(VMIN ,μ+3σ)范围作为数据筛选规则；If VMIN>μ-3σ and VMAX>μ+3σ, the range of (VMIN, μ+3σ) is taken as the data screening rule;
若VMIN<μ-3σ且VMAX<μ+3σ，则取(μ-3σ,VMAX)范围作为数据筛选规则；If VMIN<μ-3σ and VMAX<μ+3σ, the range of (μ-3σ, VMAX) is taken as the data screening rule;
若VMIN>μ-3σ且VMAX<μ+3σ，则取(VMIN ,VMAX)范围作为数据筛选规则；If VMIN>μ-3σ and VMAX<μ+3σ, the range of (VMIN, VMAX) is taken as the data screening rule;
其中，VMIN为变量最小值，VMAX变量最大值，μ变量均值，σ变量标准差。Among them, VMIN is the minimum value of the variable, the maximum value of the VMAX variable, the mean of the μ variable, and the standard deviation of the σ variable.
进一步的，剔除异常数据且根据数据的正态分布情况确定不同数据筛选规则后，对历史操作数据进行离散化处理，再对历史操作模型进行自适应优质编码。Further, after removing abnormal data and determining different data filtering rules according to the normal distribution of the data, the historical operation data is discretized, and then the historical operation model is adaptively encoded with high quality.
编码后，统计同一个喂料量分区中各个编码出现的频率，并根据频率进行降序排序；统计相同编码对应的单机电耗，将相同编码进行整合去重，形成独自编码，且独自编码对应的单机电耗等于所有相同编码的单机电耗的均值，获得降序排序中前十位独自编码对应的单机电耗，选择单机电耗最低的作为所在喂料量分区的最优历史操作记录，同时获得相应的最优推荐表。After coding, the frequency of each code in the same feed volume partition is counted and sorted in descending order according to the frequency; the single-electromechanical consumption corresponding to the same code is counted, the same code is integrated and deduplicated to form an unique code, and the corresponding code The single electromechanical consumption is equal to the average of all the single electromechanical consumptions with the same encoding. Obtain the single electromechanical consumption corresponding to the top ten unique codes in descending order, and select the lowest single electromechanical consumption as the best historical operation record of the feeding volume partition where it is obtained. The corresponding optimal recommendation table.
其中，所述编码与所述历史操作模型一一对应，编码是各变量对模型的映射，能够对模型定位，即根据变量可以快速找到相应的模型，通过编码方式，不仅能大大降低样本的内存空间，提高模型训练速度，且能大大提高学习精确度。Among them, the coding corresponds to the historical operation model one by one, and the coding is a mapping of each variable to the model, which can locate the model, that is, the corresponding model can be quickly found according to the variable, and the coding method can not only greatly reduce the memory of the sample Space, improve model training speed, and can greatly improve learning accuracy.
编码代表了磨机研磨压力、磨机进口负压、磨机出口负压、热风阀门开度、冷风阀门开度、磨机入口温度、磨机出口温度、风机阀门开度等变量，不同变量根据数据范围。The code represents the grinding pressure of the mill, the negative pressure of the mill inlet, the negative pressure of the mill outlet, the opening of the hot air valve, the opening of the cold air valve, the inlet temperature of the mill, the outlet temperature of the mill, the opening of the fan valve and other variables. data range.
具体编码计算公式为：编码＝取整函数((变量–变量最低值)/变量步长)，The specific coding calculation formula is: coding = rounding function ((variable-minimum value of variable) / variable step size),
本申请采用自适应优质编码，其变量最低值以及变量步长将不掺和任何主观经验，取决于数据本身，脱离了人为的固定主观经验。This application uses adaptive high-quality coding, and its minimum value and variable step size will not incorporate any subjective experience, which depends on the data itself and deviates from artificial fixed subjective experience.
其中，所述变量最低值取决于正太分布情况：Among them, the lowest value of the variable depends on the distribution of positive too:
若是标准正态分布数据，则变量最低值＝均值-3*标准差；In the case of standard normal distribution data, the lowest value of the variable = mean-3 * standard deviation;
若是左偏态(偏态系数SK<-0 .1)分布数据(分布的左侧有长尾)，当数据集中(峰态系数 KT>0)，则变量最低值＝均值-3*标准差，当数据分散(峰态系数KT<0)，则变量最低值＝均值- 标准差；If it is left skew (skew coefficient SK<-0 .1) Distribution data (the left side of the distribution has a long tail), when the data is concentrated (kurtosis coefficient KT>0), then the lowest value of the variable=mean-3*standard deviation, when the data is dispersed (kurtosis coefficient KT<0), then the lowest value of the variable=mean-standard deviation;
若是右偏态(偏态系数SK>0 .1)分布数据(分布的右侧有长尾)，当数据集中(峰态系数 KT>0)，则变量最低值＝均值-3*标准差，当数据分散(峰态系KT<0)，则变量最低值＝均值- 标准差；If it is right skew (skew coefficient SK>0 .1) Distribution data (long tail on the right side of the distribution), when the data is concentrated (kurtosis coefficient KT>0), the lowest value of the variable=mean-3*standard deviation, when the data are scattered (peak state KT<0), then the lowest value of the variable=mean-standard deviation;
所述变量步长取决于数据精度变化量：The variable step size depends on the amount of change in data accuracy:
DIFF＝(均值+3*标准差)-(均值-3*标准差)DIFF=(mean+3*standard deviation)-(mean-3*standard deviation)
IF DIFF<＝5AND DIFF>＝0 .5：IF DIFF<=5AND DIFF>=0.5:
变量步长＝0 .1；Variable step size = 0.1;
ELIF DIFF>5AND DIFF<＝50:ELIF DIFF>5AND DIFF<=50:
变量步长＝1；Variable step size=1;
ELIF DIFF<0 .5：ELIF DIFF<0 .5:
变量步长＝0 .01；Variable step size = 0.01;
ELIF DIFF>50：ELIF DIFF>50:
变量步长＝10。Variable step size=10.
进一步的，根据最优推荐表进行最优操作建议时，对推荐的最优历史操作记录的编码进行解码，解码计算公式如下：Further, when performing the optimal operation recommendation according to the optimal recommendation table, the recommended optimal historical operation record is decoded, and the decoding calculation formula is as follows:
变量＝变量编码*变量步长+变量最低值。Variable = variable code * variable step + minimum value of variable.
进一步的，根据实时工况，对最优推荐表对应的工况的最优操作建议进行迭代优化，具体迭代优化过程为，操作产生的操作模型形成实时编码，将实时编码与所述独自编码匹配，当匹配距离为零且实时编码的单机电耗低于独自编码的单机电耗时，判断为较优的操作记录，将该操作记录存储，且重新计算单机电耗均值，即获得该独自编码的新的单机电耗，再与所在喂料量分区的其它独自编码比较出现的频率和单机电耗，从而更新所在喂料量分区的最优历史操作记录；当匹配不到距离为零的独自编码时，将所述实时编码列为新的独自编码，再与所在喂料量分区的其它独自编码比较出现的频率和单机电耗，从而更新所在喂料量分区的最优历史操作记录。Further, according to the real-time operating conditions, iterative optimization is performed on the optimal operation recommendations of the operating conditions corresponding to the optimal recommendation table. The specific iterative optimization process is that the operation model generated by the operation forms a real-time encoding, and matches the real-time encoding with the unique encoding , When the matching distance is zero and the single-electromechanical consumption of the real-time encoding is lower than that of the single-coded single-electromechanical consumption, it is judged as a better operation record, the operation record is stored, and the average value of the single-electromechanical consumption is recalculated to obtain the unique code The new single electromechanical consumption, and then compare the frequency and single electromechanical consumption with other unique codes of the feed volume partition, so as to update the optimal historical operation record of the feed volume partition; when the match does not match the distance of zero alone When encoding, the real-time code is listed as a new unique code, and then compared with the frequency of the other unique codes of the feed volume partition and the single electromechanical consumption, thereby updating the optimal historical operation record of the feed volume partition.
有益效果Beneficial effect
由上述对本发明的描述可知，和现有技术相比，本发明具有如下优点：As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following advantages:
1、通过记录历史操作记录，根据实时工况和所述最优推荐表获得设备的最优操作建议，达到节能降耗；1. Obtain the optimal operation recommendations of the equipment based on the real-time operating conditions and the optimal recommendation table by recording historical operation records to achieve energy saving and consumption reduction;
2、采用自适应的筛选规则和自适应的优质编码，脱离了人为的固定主观经验，不掺和任何主观看法或观察数据统计后的经验，先自适应学习调整筛选规则，提取安全正常的有价值数据，再通过自适应学习调整编码公式参数，从而达到历史数据中挖掘出潜在价值，其推荐的优化方案客观、合理、可靠、安全；2. Adopting adaptive filtering rules and adaptive high-quality coding, which deviates from the artificial fixed subjective experience, does not incorporate any subjective viewing methods or experience after observing data statistics, first adaptively adjust the filtering rules to extract safe and normal Value data, and then adjust the coding formula parameters through adaptive learning, so as to achieve the potential value of historical data, and the recommended optimization scheme is objective, reasonable, reliable and safe;
3、本发明的在线推送速度快，只需判断是属于哪个喂料量分区，即可从存储的最3. The online push speed of the present invention is fast, and it is only necessary to determine which feed volume partition it belongs to.
优推荐表中快速高效的提取相应的推荐建议，拥有快效率推荐速度，远远满足了每3s一个推荐建议的需求；Quickly and efficiently extract the corresponding recommendation suggestions in the excellent recommendation table, with fast efficiency recommendation speed, which far meets the requirement of one recommendation every 3s;
4、本发明拥有快速学习效率，包括了在线更新功能。在迭代最优推荐表时，无需针对所有喂料量分区数据再进行学习，去得到各喂料量分区的最优操作记录，只需根据喂料量判断相应的喂料量分区，再对对应分区数据进行再学习，接着只需将最优推荐表中该分区的最优历史操作记录进行更新即可。4. The present invention has fast learning efficiency and includes an online update function. When iterating the optimal recommendation table, there is no need to learn for all the feed volume partition data, to obtain the optimal operation record of each feed volume partition, only to determine the corresponding feed volume partition according to the feed volume, and then to the corresponding Re-learn the partition data, and then just update the optimal historical operation record of the partition in the optimal recommendation table.
附图说明BRIEF DESCRIPTION
此处所说明的附图用来提供对本发明的进一步理解，构成本发明的一部分，本发明的示意性实施例及其说明用于解释本发明，并不构成对本发明的不当限定。The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an undue limitation on the present invention.
其中：among them:
图1是本发明一种水泥原料立磨***的节能优化方法实施例一的流程示意图；1 is a schematic flowchart of Embodiment 1 of an energy-saving optimization method of a cement raw material vertical mill system of the present invention;
本发明的实施方式Embodiments of the invention
为了使本发明所要解决的技术问题、技术方案及有益效果更加清楚、明白，以下结合附图和实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本发明，并不用于限定本发明。In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention clearer and more comprehensible, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
实施例一：一种水泥原料立磨***的节能优化方法，在不改变生产设备任何结构和原理、不增加额外测点、不影响正常生产的前提下，通过机器学习的方法，智能化的提供安全、便捷、合理的辅助决策。Embodiment 1: An energy-saving optimization method of a cement raw material vertical mill system, which does not change any structure and principle of production equipment, does not add additional measurement points, and does not affect normal production. It provides intelligently through machine learning methods. Safe, convenient and reasonable auxiliary decision-making.
所述方法包括：采集历史操作数据，形成多个历史操作模型，所述历史操作数据包括可控变量、规则过滤变量和相关性较小的变量，所述可控变量包括磨机研磨压力、磨机进口负压、磨机出口负压、热风阀门开度、冷风阀门开度、磨机入口温度、磨机出口温度、风机阀门开度，所述规则过滤变量包括磨机振动值和选粉机电流，本实施例只考虑如何优化相关且可控变量来作为节能降耗的优化建议，从而实现在不影响正常生产的前提下，降低磨机和循环风机能耗。The method includes: collecting historical operation data to form a plurality of historical operation models, the historical operation data including controllable variables, regular filtering variables and variables with less correlation, the controllable variables including mill grinding pressure, grinding Machine inlet negative pressure, mill outlet negative pressure, hot air valve opening, cold air valve opening, mill inlet temperature, mill outlet temperature, fan valve opening, the rule filtering variables include mill vibration value and powder separator For current, this embodiment only considers how to optimize related and controllable variables as optimization suggestions for energy saving and consumption reduction, so as to reduce energy consumption of the mill and circulating fan without affecting normal production.
根据磨机的喂料量划分出喂料量分区。According to the feeding amount of the mill, the feeding amount is divided into zones.
根据单机电耗量的高低，对同一个喂料量分区的历史操作进行排序，获得一个喂料量分区的最优历史操作记录，将所有喂料量分区的最优历史操作记录合并，形成最优推荐表。According to the level of single electromechanical consumption, the historical operations of the same feed volume partition are sorted to obtain the optimal historical operation records of a feed volume partition, and the optimal historical operation records of all feed volume partitions are merged to form the most Excellent recommendation form.
根据实时工况和所述最优推荐表获得设备的最优操作建议。Obtain the optimal operation recommendation of the device according to the real-time operating conditions and the optimal recommendation table.
实施例二：在实施例一的基础上，对历史操作数据进行分析处理，具体的：在获得最优历史操作记录前，通过机器学习的方式对所述历史操作数据进行箱线图分布统计，得到上限和下限，剔除上限和下限之外的异常数据，之后，通过自主学习数据的正态分布情况,自适应确定不同数据筛选规则，具体逻辑如下：Embodiment 2: Based on the first embodiment, the historical operation data is analyzed and processed, specifically: before obtaining the optimal historical operation record, the box operation graph distribution statistics are performed on the historical operation data by machine learning, Obtain the upper and lower limits, remove the abnormal data except the upper and lower limits, and then, through the normal distribution of self-learning data, adaptively determine different data screening rules, the specific logic is as follows:
若VMIN<μ-3σ且VMAX>μ+3σ，则取(μ-3σ,μ+3σ)范围作为数据筛选规则；If VMIN<μ-3σ and VMAX>μ+3σ, the range of (μ-3σ, μ+3σ) is taken as the data screening rule;
若VMIN>μ-3σ且VMAX>μ+3σ，则取(VMIN ,μ+3σ)范围作为数据筛选规则；If VMIN>μ-3σ and VMAX>μ+3σ, the range of (VMIN, μ+3σ) is taken as the data screening rule;
若VMIN<μ-3σ且VMAX<μ+3σ，则取(μ-3σ,VMAX)范围作为数据筛选规则；If VMIN<μ-3σ and VMAX<μ+3σ, the range of (μ-3σ, VMAX) is taken as the data screening rule;
若VMIN>μ-3σ且VMAX<μ+3σ，则取(VMIN ,VMAX)范围作为数据筛选规则；If VMIN>μ-3σ and VMAX<μ+3σ, the range of (VMIN, VMAX) is taken as the data screening rule;
其中，VMIN为变量最小值，VMAX变量最大值，μ变量均值，σ变量标准差，且初始目标磨机主机电流和循环风机高压电流的硬性满足条件如下：Among them, VMIN is the minimum value of the variable, the maximum value of the VMAX variable, the mean of the μ variable, the standard deviation of the σ variable, and the rigidity of the initial target mill current and the high-voltage current of the circulating fan meet the following conditions:
磨机主机电流：[90 ,125]Mill main current: [90,125]
循环风机高压电流：[60 ,70]。High voltage current of circulating fan: [60,70].
接着对历史操作数据进行离散化处理，再对历史操作模型进行自适应优质编码，比如： X11 ,X12 ,X13 ,...,X1N，可参阅图1。Next, the historical operation data is discretized, and then the historical operation model is adaptively encoded with high quality, such as: X11, X12, X13,..., X1N, see Figure 1.
[根据细则91更正 13.03.2020]　
编码后，统计同一个喂料量分区中各个编码出现的频率，并根据频率进行降序排序；统计相同编码对应的单机电耗Y1，将相同编码进行整合去重，形成独自编码，且独自编码对应的单机电耗Y2等于所有相同编码的单机电耗Y1的均值，获得降序排序中前十位独自编码对应的单机电耗Y12，选择最低的单机电耗作为所在喂料量分区的最优历史操作记录Ybest（Y最优），同时获得相应的最优推荐表。[Correction 13.03.2020 in accordance with Rule 91]
After coding, the frequency of each code in the same feed volume partition is counted and sorted in descending order according to the frequency; the single electromechanical consumption Y1 corresponding to the same code is counted, and the same code is integrated and deduplicated to form an unique code, and the unique code corresponds to The single electromechanical consumption Y2 is equal to the average of all the single electromechanical consumption Y1 of the same code, and the single electromechanical consumption Y12 corresponding to the top ten unique codes in descending order is obtained. The lowest single electromechanical consumption is selected as the optimal historical operation of the feed volume partition Record Ybest (Y optimal) and obtain the corresponding optimal recommendation table.
其中，所述编码与所述历史操作模型一一对应，编码是各变量对模型的映射，能够对模型定位，即根据变量可以快速找到相应的模型，通过编码方式，不仅能大大降低样本的内存空间，提高模型训练速度，且能大大提高学习精确度。Among them, the coding corresponds to the historical operation model one by one, and the coding is a mapping of each variable to the model, which can locate the model, that is, the corresponding model can be quickly found according to the variable, and the coding method can not only greatly reduce the memory of the sample Space, improve model training speed, and can greatly improve learning accuracy.
编码代表了磨机研磨压力、磨机进口负压、磨机出口负压、热风阀门开度、冷风阀门开度、磨机入口温度、磨机出口温度、风机阀门开度等变量，不同变量根据数据范围。The code represents the grinding pressure of the mill, the negative pressure of the mill inlet, the negative pressure of the mill outlet, the opening of the hot air valve, the opening of the cold air valve, the inlet temperature of the mill, the outlet temperature of the mill, the opening of the fan valve and other variables. data range.
具体编码计算公式为：编码＝取整函数((变量–变量最低值)/变量步长)，The specific coding calculation formula is: coding = rounding function ((variable-minimum value of variable) / variable step size),
本申请采用自适应优质编码，其变量最低值以及变量步长将不掺和任何主观经验，取决于数据本身，脱离了人为的固定主观经验。This application uses adaptive high-quality coding, and its minimum value and variable step size will not incorporate any subjective experience, which depends on the data itself and deviates from artificial fixed subjective experience.
其中，所述变量最低值取决于正太分布情况：Among them, the lowest value of the variable depends on the distribution of positive too:
若是标准正态分布数据，则变量最低值＝均值-3*标准差；In the case of standard normal distribution data, the lowest value of the variable = mean-3 * standard deviation;
若是左偏态(偏态系数SK<-0 .1)分布数据(分布的左侧有长尾)，当数据集中(峰态系数 KT>0)，则变量最低值＝均值-3*标准差，当数据分散(峰态系数KT<0)，则变量最低值＝均值- 标准差；If it is left skew (skew coefficient SK<-0 .1) Distribution data (the left side of the distribution has a long tail), when the data is concentrated (kurtosis coefficient KT>0), then the lowest value of the variable=mean-3*standard deviation, when the data is dispersed (kurtosis coefficient KT<0), then the lowest value of the variable=mean-standard deviation;
若是右偏态(偏态系数SK>0 .1)分布数据(分布的右侧有长尾)，当数据集中(峰态系数 KT>0)，则变量最低值＝均值-3*标准差，当数据分散(峰态系数KT<0)，则变量最低值＝均值- 标准差；If it is right skew (skew coefficient SK>0 .1) Distribution data (long tail on the right side of the distribution), when the data is concentrated (kurtosis coefficient KT>0), then the lowest value of the variable=mean-3*standard deviation, when the data is dispersed (kurtosis coefficient KT<0), then the lowest value of the variable=mean-standard deviation;
所述变量步长取决于数据精度变化量：The variable step size depends on the amount of change in data accuracy:
DIFF＝(均值+3*标准差)-(均值-3*标准差)DIFF=(mean+3*standard deviation)-(mean-3*standard deviation)
IF DIFF<＝5AND DIFF>＝0 .5：IF DIFF<=5AND DIFF>=0.5:
变量步长＝0 .1；Variable step size = 0.1;
ELIF DIFF>5AND DIFF<＝50:ELIF DIFF>5AND DIFF<=50:
变量步长＝1；Variable step size=1;
ELIF DIFF<0 .5：ELIF DIFF<0 .5:
变量步长＝0 .01；Variable step size = 0.01;
ELIF DIFF>50：ELIF DIFF>50:
变量步长＝10。Variable step size=10.
根据最优推荐表进行最优操作建议时，对推荐的最优历史操作记录的编码进行解码，解码计算公式如下：When performing the optimal operation recommendation according to the optimal recommendation table, the recommended optimal historical operation record is decoded, and the decoding calculation formula is as follows:
变量＝变量编码*变量步长+变量最低值。Variable = variable code * variable step + minimum value of variable.
实施例三，在实施例二的基础上，根据实时工况，对最优推荐表对应的工况的最优操作建议进行迭代优化，具体迭代优化过程为：In the third embodiment, on the basis of the second embodiment, according to the real-time operating conditions, iterative optimization is performed on the optimal operation recommendations for the operating conditions corresponding to the optimal recommendation table. The specific iterative optimization process is:
操作产生的操作模型形成实时编码，将实时编码与所述独自编码匹配，当匹配距离为零且实时编码的单机电耗低于独自编码的单机电耗时，判断为较优的操作记录，将该操作记录存储，且重新计算单机电耗均值，即获得该独自编码的新的单机电耗，再与所在喂料量分区的其它独自编码比较出现的频率和单机电耗，从而更新所在喂料量分区的最优历史操作记录；The operation model generated by the operation forms a real-time code, and the real-time code is matched with the unique code. When the matching distance is zero and the single-electromechanical consumption of the real-time code is lower than that of the single-coded single electromechanical, it is judged as a better operation record. The operation record is stored, and the average value of the single electromechanical consumption is recalculated, that is, the new single electromechanical consumption of the unique code is obtained, and then the frequency and the single electromechanical consumption are compared with the other unique codes of the feed volume partition to update the feed. The best historical operation record of the volume partition;
当匹配不到距离为零的独自编码时，将所述实时编码列为新的独自编码，再与所在喂料量分区的其它独自编码比较出现的频率和单机电耗，从而更新所在喂料量分区的最优历史操作记录。When no unique code with a distance of zero can be matched, the real-time code is listed as a new unique code, and then compared with the other unique codes in the feed volume partition where the frequency and single electrical consumption are compared to update the feed volume. The best historical operation record of the partition.
综上所述，和现有技术相比，本申请提出的一种水泥原料立磨***的节能优化方法，具有以下优点：In summary, compared with the prior art, the energy-saving optimization method of the cement raw material vertical mill system proposed in this application has the following advantages:
1、通过记录历史操作记录，根据实时工况和所述最优推荐表获得设备的最优操作建议，达到节能降耗；1. Obtain the optimal operation recommendations of the equipment based on the real-time operating conditions and the optimal recommendation table by recording historical operation records to achieve energy saving and consumption reduction;
2、采用自适应的筛选规则和自适应的优质编码，脱离了人为的固定主观经验，不掺和任何主观看法或观察数据统计后的经验，先自适应学习调整筛选规则，提取安全正常的有价值数据，再通过自适应学习调整编码公式参数，从而达到历史数据中挖掘出潜在价值，其推荐的优化方案客观、合理、可靠、安全；2. Adopting adaptive filtering rules and adaptive high-quality coding, which deviates from the artificial fixed subjective experience, does not incorporate any subjective viewing methods or experience after observing data statistics, first adaptively adjust the filtering rules to extract safe and normal Value data, and then adjust the coding formula parameters through adaptive learning, so as to achieve the potential value of historical data, and the recommended optimization scheme is objective, reasonable, reliable and safe;
3、本发明的在线推送速度快，只需判断是属于哪个喂料量分区，即可从存储的最优推荐表中快速高效的提取相应的推荐建议，拥有快效率和推荐速度，远远满足了每3s一个推荐建议的需求；3. The online push speed of the present invention is fast. Only by judging which feed volume partition it belongs to, the corresponding recommendation recommendations can be quickly and efficiently extracted from the stored optimal recommendation table. It has fast efficiency and recommendation speed, which is far from satisfactory The need for a recommendation every 3s;
4、本发明拥有快速学习效率，包括了在线更新功能。在迭代最优推荐表时，无需针对所有喂料量分区数据再进行学习，去得到各喂料量分区的最优操作记录，只需根据喂料量判断相应的喂料量分区，再对对应分区数据进行再学习，接着只需将最优推荐表中该分区的最优历史操作记录进行更新即可。4. The present invention has fast learning efficiency and includes an online update function. When iterating the optimal recommendation table, there is no need to learn for all the feed volume partition data, to obtain the optimal operation record of each feed volume partition, only to determine the corresponding feed volume partition according to the feed volume, and then to the corresponding Re-learn the partition data, and then just update the optimal historical operation record of the partition in the optimal recommendation table.
上面结合附图对本发明进行了示例性描述，显然本发明具体实现并不受上述方式的限制，只要采用了本发明的方法构思和技术方案进行的各种非实质性的改进，或未经改进将本发明的构思和技术方案直接应用于其它场合的，均在本发明的保护范围之内。The present invention has been exemplarily described above with reference to the drawings. Obviously, the specific implementation of the present invention is not limited by the above-mentioned methods, as long as various non-substantial improvements made by the method concept and technical solution of the present invention are adopted, or no improvement It is within the protection scope of the present invention to directly apply the concepts and technical solutions of the present invention to other occasions.
 A

Claims (8)

  1. 一种水泥原料立磨***的节能优化方法，其特征在于，An energy-saving optimization method for a cement raw material vertical mill system is characterized by:
        采集历史操作数据，形成多个历史操作模型，所述历史操作数据包括可控变量，所述可控变量包括磨机研磨压力、磨机进口负压、磨机出口负压、热风阀门开度、冷风阀门开度、磨机入口温度、磨机出口温度、风机阀门开度；Collect historical operation data to form multiple historical operation models. The historical operation data includes controllable variables including mill grinding pressure, mill inlet negative pressure, mill outlet negative pressure, hot air valve opening, Cold air valve opening, mill inlet temperature, mill outlet temperature, fan valve opening;
        根据喂料量获得喂料量分区；According to the feed volume to obtain the feed volume partition;
        根据单机电耗量的高低，对同一个喂料量分区的历史操作进行排序，获得一个喂料量分区的最优历史操作记录，将所有喂料量分区的最优历史操作记录合并，形成最优推荐表。According to the level of single electromechanical consumption, the historical operations of the same feed volume partition are sorted to obtain the optimal historical operation records of a feed volume partition, and the optimal historical operation records of all feed volume partitions are merged to form the most Excellent recommendation form.
  2. 根据权利要求1所述的一种水泥原料立磨***的节能优化方法，其特征在于，根据实时工况和所述最优推荐表获得设备的最优操作建议。The energy-saving optimization method of a cement raw material vertical mill system according to claim 1, wherein the optimal operation recommendation of the equipment is obtained according to the real-time working conditions and the optimal recommendation table.
     A
  3. 根据权利要求1所述的一种水泥原料立磨***的节能优化方法，其特征在于，在获得最优历史操作记录前，通过机器学习的方式对所述历史操作数据进行箱线图分布统计，得到上限和下限，剔除上限和下限之外的异常数据。An energy-saving optimization method for a cement raw material vertical mill system according to claim 1, characterized in that, before obtaining the optimal historical operation record, the box-line graph distribution statistics are performed on the historical operation data by machine learning, Obtain upper and lower limits, and remove abnormal data except the upper and lower limits.
  4. 根据权利要求3所述的一种水泥原料立磨***的节能优化方法，其特征在于，剔除历史操作数据中的异常数据后,自适应确定不同数据筛选规则，通过自主学习数据的正态分布情况,具体逻辑如下：An energy-saving optimization method for a cement raw material vertical grinding system according to claim 3, characterized in that after removing abnormal data in the historical operation data, adaptively determine different data screening rules and learn the normal distribution of the data by autonomous learning The specific logic is as follows:
        若VMIN<μ-3σ且VMAX>μ+3σ，则取(μ-3σ,μ+3σ)范围作为数据筛选规则；If VMIN<μ-3σ and VMAX>μ+3σ, the range of (μ-3σ, μ+3σ) is taken as the data screening rule;
        若VMIN>μ-3σ且VMAX>μ+3σ，则取(VMIN ,μ+3σ)范围作为数据筛选规则；If VMIN>μ-3σ and VMAX>μ+3σ, the range of (VMIN, μ+3σ) is taken as the data screening rule;
        若VMIN<μ-3σ且VMAX<μ+3σ，则取(μ-3σ,VMAX)范围作为数据筛选规则；If VMIN<μ-3σ and VMAX<μ+3σ, the range of (μ-3σ, VMAX) is taken as the data screening rule;
        若VMIN>μ-3σ且VMAX<μ+3σ，则取(VMIN ,VMAX)范围作为数据筛选规则；If VMIN>μ-3σ and VMAX<μ+3σ, the range of (VMIN, VMAX) is taken as the data screening rule;
        其中，VMIN为变量最小值，VMAX变量最大值，μ变量均值，σ变量标准差。Among them, VMIN is the minimum value of the variable, the maximum value of the VMAX variable, the mean of the μ variable, and the standard deviation of the σ variable.
     A
  5. 根据权利要求4所述的一种水泥原料立磨***的节能优化方法，其特征在于，对历史操作数据进行离散化处理且对历史操作模型进行自适应优质编码；统计同一个喂料量分区中各个编码出现的频率，并根据频率进行降序排序；统计相同编码对应的单机电耗，将相同编码进行整合去重，形成独自编码，且独自编码对应的单机电耗等于所有相同编码的单机电耗的均值，获得降序排序中前十位独自编码对应的单机电耗，选择单机电耗最低的作为所在喂料量分区的最优历史操作记录，同时获得相应的最优推荐表。An energy-saving optimization method for a cement raw material vertical mill system according to claim 4, characterized in that the historical operation data is discretized and the historical operation model is adaptively coded with high quality; statistics are in the same feed volume partition The frequency of occurrence of each code, and sort them in descending order according to the frequency; count the single electromechanical consumption corresponding to the same code, integrate and deduplicate the same code to form an unique code, and the single electromechanical consumption corresponding to the unique code is equal to all the single electromechanical consumption of the same code The average value of the first ten digits in the descending order is coded corresponding to the single electromechanical consumption, and the lowest single electromechanical consumption is selected as the optimal historical operation record of the feed volume partition, and the corresponding optimal recommendation table is obtained.
     A
  6. 根据权利要求5所述的一种水泥原料立磨***的节能优化方法，其特征在于，所述编码与所述历史操作模型一一对应，具体编码计算公式为：编码＝取整函数((变量–变量最低值)/变量步长)，An energy-saving optimization method for a cement raw material vertical mill system according to claim 5, characterized in that the codes correspond one-to-one with the historical operation model, and the specific code calculation formula is: code = rounding function ((variable – Variable minimum value)/variable step size),
        所述变量最低值取决于正太分布情况：The lowest value of the variable depends on the distribution of the positive and negative:
        若是标准正态分布数据，则变量最低值＝均值-3*标准差；In the case of standard normal distribution data, the lowest value of the variable = mean-3 * standard deviation;
        若是左偏态(偏态系数SK<-0 .1)分布数据(分布的左侧有长尾)，当数据集中(峰态系数KT>0)，则变量最低值＝均值-3*标准差，当数据分散(峰态系数KT<0)，则变量最低值＝均值-标准差；If it is left skewed (skew coefficient SK<-0.1) distribution data (the left side of the distribution has a long tail), when the data set (kurtosis coefficient KT>0), the lowest value of the variable = mean-3 * standard deviation , When the data is dispersed (kurtosis coefficient KT<0), the lowest value of the variable = mean-standard deviation;
        若是右偏态(偏态系数SK>0 .1)分布数据(分布的右侧有长尾)，当数据集中(峰态系数KT>0)，则变量最低值＝均值-3*标准差，当数据分散(峰态系数KT<0)，则变量最低值＝均值-标准差；If it is right-skewed (skewness coefficient SK>0.1) distribution data (the distribution has a long tail on the right side), when the data set (kurtosis coefficient KT>0), then the lowest value of the variable = mean-3 * standard deviation, When the data are scattered (kurtosis coefficient KT<0), the lowest value of the variable=mean-standard deviation;
        所述变量步长取决于数据精度变化量：The variable step size depends on the amount of change in data accuracy:
      DIFF＝(均值+3*标准差)-(均值-3*标准差)DIFF = (mean + 3 * standard deviation)-(mean-3 * standard deviation)
        IF DIFF<＝5AND DIFF>＝0 .5：IF DIFF<=5AND DIFF>=0.5:
        变量步长＝0 .1；Variable step size = 0.1;
        ELIF DIFF>5AND DIFF<＝50:ELIF DIFF>5AND DIFF<=50:
        变量步长＝1；Variable step size = 1;
        ELIF DIFF<0 .5：ELIF DIFF<0.5:
        变量步长＝0 .01；Variable step size = 0.01;
        ELIF DIFF>50：ELIF DIFF>50:
        变量步长＝10。Variable step size=10.
     A
  7. 根据权利要求5所述的一种水泥原料立磨***的节能优化方法，其特征在于，根据最优推荐表进行最优操作建议时，对推荐的最优历史操作记录的编码进行解码，解码计算公式如下：变量＝变量编码*变量步长+变量最低值。An energy-saving optimization method for a cement raw material vertical mill system according to claim 5, characterized in that, when performing the optimal operation recommendation according to the optimal recommendation table, the recommended optimal historical operation record is decoded and decoded and calculated The formula is as follows: variable=variable code*variable step size+variable minimum value.
  8. 根据权利要求5所述的一种水泥原料立磨***的节能优化方法，其特征在于，根据实时工况，对最优推荐表对应的工况的最优操作建议进行迭代优化，具体迭代优化过程为，操作产生的操作模型形成实时编码，将实时编码与所述独自编码匹配，当匹配距离为零且实时编码的单机电耗低于独自编码的单机电耗时，判断为较优的操作记录，将该操作记录存储，且重新计算单机电耗均值，即获得该独自编码的新的单机电耗，再与所在喂料量分区的其它独自编码比较出现的频率和单机电耗，从而更新所在喂料量分区的最优历史操作记录；当匹配不到距离为零的独自编码时，将所述实时编码列为新的独自编码，再与所在喂料量分区的其它独自编码比较出现的频率和单机电耗，从而更新所在喂料量分区的最优历史操作记录。An energy-saving optimization method for a cement raw material vertical mill system according to claim 5, characterized in that iterative optimization is performed on the optimal operation recommendations for the operating conditions corresponding to the optimal recommendation table according to the real-time operating conditions, and the specific iterative optimization process For the operation model generated by the operation to form a real-time code, the real-time code is matched with the unique code. When the matching distance is zero and the single-electromechanical consumption of the real-time code is lower than the single-coded single electromechanical consumption, it is judged as a better operation record , Store the operation record, and recalculate the average value of the single electromechanical consumption, that is, obtain the new single electromechanical consumption of the unique code, and then compare the frequency and single electromechanical consumption with the other unique codes of the feed volume partition to update the location The optimal historical operation record of the feed volume partition; when no unique code with a distance of zero is matched, the real-time code is listed as a new unique code, and then compared with the frequency of other unique codes in the feed volume partition And single machine consumption, so as to update the optimal historical operation record of the feed volume partition.
     A
PCT/CN2019/120980 2018-12-26 2019-11-26 Energy-saving optimization method of cement raw material vertical mill system WO2020134801A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811604430.9 2018-12-26
CN201811604430.9A CN109847916B (en) 2018-12-26 2018-12-26 Energy-saving optimization method of cement raw material vertical mill system

Publications (1)

Publication Number Publication Date
WO2020134801A1 true WO2020134801A1 (en) 2020-07-02

Family

ID=66892607

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/120980 WO2020134801A1 (en) 2018-12-26 2019-11-26 Energy-saving optimization method of cement raw material vertical mill system

Country Status (2)

Country Link
CN (1) CN109847916B (en)
WO (1) WO2020134801A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112949183A (en) * 2021-03-04 2021-06-11 上海交通大学宁波人工智能研究院 System and method for detecting abnormal working conditions of cement raw material vertical mill system

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109847916B (en) * 2018-12-26 2021-01-12 厦门邑通软件科技有限公司 Energy-saving optimization method of cement raw material vertical mill system
CN110471375B (en) * 2019-07-08 2020-07-28 杭州电子科技大学 Anti-interference optimization tracking method for cement denitration process
CN112506134A (en) * 2019-09-16 2021-03-16 阿里巴巴集团控股有限公司 Method, device and equipment for determining control variable value
TR202006947A2 (en) * 2020-05-04 2021-11-22 Sabanci Dijital Teknoloji Hizmetleri Anonim Sirketi System and method for ai controlling mill operations to lower electricity consumption
CN116474928B (en) * 2023-06-25 2023-09-26 中才邦业(杭州)智能技术有限公司 Cement mill energy consumption optimization method and system

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101269350A (en) * 2008-04-25 2008-09-24 东南大学 Operation optimizing and energy-saving control method for intermediate storage iron shot mill flour milling system
CN102566551A (en) * 2012-02-03 2012-07-11 北京华电天仁电力控制技术有限公司 Data mining-based method for analyzing thermal power plant operation index optimal target value
WO2014185810A8 (en) * 2013-05-13 2015-06-11 Siemens Aktiengesellschaft Method for adjusting final steel properties at steel mill facility
CN106990763A (en) * 2017-04-20 2017-07-28 浙江大学 A kind of Vertical Mill operation regulator control system and method based on data mining
CN107016176A (en) * 2017-03-24 2017-08-04 杭州电子科技大学 A kind of hybrid intelligent overall boiler burning optimization method
CN107239066A (en) * 2017-06-09 2017-10-10 浙江大学 A kind of Vertical Mill operation closed-loop control device and method based on data mining
CN107818409A (en) * 2017-10-23 2018-03-20 燕山大学 The optimization method of manufacture of cement parameter and regulation section based on fuzzy association rules
CN109847916A (en) * 2018-12-26 2019-06-07 厦门邑通软件科技有限公司 A kind of energy conservation optimizing method of cement raw material vertical mill system

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4833617A (en) * 1987-08-14 1989-05-23 General Electric Company Solid modeling based adaptive feedrate control for NC machining
US7415446B2 (en) * 2003-10-20 2008-08-19 General Cybernation Group, Inc. Model-free adaptive (MFA) optimization
US7591440B2 (en) * 2004-08-13 2009-09-22 Invensys Systems, Inc. Methods and systems for cement finishing mill control
CN100369677C (en) * 2006-04-24 2008-02-20 西安交通大学 Powder-making system automatic control method for heat engine plant steel ball coal grinding mill
CN101318158A (en) * 2008-03-17 2008-12-10 西安艾贝尔科技发展有限公司 Control method for grinding process optimization of cement factory
CN102402186A (en) * 2011-12-20 2012-04-04 西安艾贝尔科技发展有限公司 Optimized control method for milling process in cement plant
CN103028480B (en) * 2012-12-10 2015-04-22 上海凯盛节能工程技术有限公司 Intelligent control system for vertical mill based on fuzzy PID (proportion integration differentiation) algorithm
CN107344138B (en) * 2016-05-05 2019-03-22 中冶长天国际工程有限责任公司 A kind of ore mill control method and device
CN106179633B (en) * 2016-07-28 2018-06-26 洛阳矿山机械工程设计研究院有限责任公司 A kind of method of adjustment of grinding machine intelligence automatic speed regulation system
CN106676934B (en) * 2016-12-28 2018-08-03 东北大学 A kind of high consistency refining systematic economy optimization progress control method and system
CN108268597B (en) * 2017-12-18 2020-10-09 中国电子科技集团公司第二十八研究所 Moving target activity probability map construction and behavior intention identification method
CN108224393B (en) * 2018-01-08 2019-11-08 中国恩菲工程技术有限公司 Intelligent waste heat boiler optimal control method
CN108600756B (en) * 2018-05-04 2019-10-29 上海大学 Coding parameter self-adapting regulation method based on Monte Carlo method Markov chain
CN110302888A (en) * 2019-07-15 2019-10-08 湖北白兆山水泥有限公司 Cement mill Optimization of Energy Saving system

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101269350A (en) * 2008-04-25 2008-09-24 东南大学 Operation optimizing and energy-saving control method for intermediate storage iron shot mill flour milling system
CN102566551A (en) * 2012-02-03 2012-07-11 北京华电天仁电力控制技术有限公司 Data mining-based method for analyzing thermal power plant operation index optimal target value
WO2014185810A8 (en) * 2013-05-13 2015-06-11 Siemens Aktiengesellschaft Method for adjusting final steel properties at steel mill facility
CN107016176A (en) * 2017-03-24 2017-08-04 杭州电子科技大学 A kind of hybrid intelligent overall boiler burning optimization method
CN106990763A (en) * 2017-04-20 2017-07-28 浙江大学 A kind of Vertical Mill operation regulator control system and method based on data mining
CN107239066A (en) * 2017-06-09 2017-10-10 浙江大学 A kind of Vertical Mill operation closed-loop control device and method based on data mining
CN107818409A (en) * 2017-10-23 2018-03-20 燕山大学 The optimization method of manufacture of cement parameter and regulation section based on fuzzy association rules
CN109847916A (en) * 2018-12-26 2019-06-07 厦门邑通软件科技有限公司 A kind of energy conservation optimizing method of cement raw material vertical mill system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112949183A (en) * 2021-03-04 2021-06-11 上海交通大学宁波人工智能研究院 System and method for detecting abnormal working conditions of cement raw material vertical mill system
CN112949183B (en) * 2021-03-04 2024-01-16 上海交通大学宁波人工智能研究院 Abnormal working condition detection system and method for cement raw material vertical mill system

Also Published As

Publication number Publication date
CN109847916A (en) 2019-06-07
CN109847916B (en) 2021-01-12

Similar Documents

Publication Publication Date Title
WO2020134801A1 (en) Energy-saving optimization method of cement raw material vertical mill system
CN109674080B (en) Tobacco leaf conditioning water adding amount prediction method, storage medium and terminal equipment
CN109603494B (en) Desulfurization device absorption circulation system optimized operation method based on big data and absorption circulation system
TWI543102B (en) Method and system of cause analysis and correction for manufacturing data
CN116823826B (en) Numerical control machine tool tipping abnormity detection method
CN107909179A (en) The prediction model construction method and vehicle energy management method of a kind of plug-in hybrid vehicle driving cycle
CN110674120A (en) Wind power plant data cleaning method and device
CN110302888A (en) Cement mill Optimization of Energy Saving system
CN110245390B (en) Automobile engine oil consumption prediction method based on RS-BP neural network
CN108142976B (en) Cut tobacco drying process parameter optimization method
CN105225252B (en) Particle clouds motion Forecasting Methodology
CN110555628B (en) Comprehensive evaluation method for matching degree of new energy power generation output and load characteristics
CN116388150A (en) Matching short-term load prediction method based on digital twin and neural network technology
CN114881374A (en) Multi-element heterogeneous energy consumption data fusion method and system for building energy consumption prediction
CN109447336B (en) Optimized control method for water level between upstream reservoir and reverse regulation reservoir dam thereof
CN111931967B (en) Short-term power prediction method for wind power plant
CN116561692A (en) Dynamic update real-time measurement data detection method
CN117113233A (en) Hierarchical energy structure scene construction method and energy consumption abnormal link tracing method
CN117150323A (en) Cluster division method based on wind field wake model
Pan et al. An adaptive genetic algorithm for the flexible job-shop scheduling problem
CN110286663B (en) Regional cigarette physical index standardized production improving method
CN116738266A (en) Transformer area line loss anomaly identification method based on improved k-medoids clustering algorithm
CN114969017B (en) Wind power data cleaning method, cleaning device and prediction method
CN114721270B (en) Rice hulling and milling cooperative control method and device and storage medium
CN109407614A (en) A kind of chain digital control gear hobbing machine gear hobbing process process parameter optimizing method

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19902250

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19902250

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19902250

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 25.02.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 19902250

Country of ref document: EP

Kind code of ref document: A1