Intelligent Coal Preparation Energy Efficiency Optimization Control Method and System

Abstract

This invention discloses an intelligent coal preparation energy efficiency optimization control method and system, relating to the field of power grid energy efficiency optimization technology. It collects historical energy consumption data for each functional area, trains an energy consumption prediction model for each functional area, and statistically analyzes historical start-up and shutdown data of each flexible device. During the actual coal preparation process, it outputs the predicted energy consumption value for each functional area. Based on the historical start-up and shutdown data of each flexible device, it obtains the predicted energy consumption value of the flexible devices within each functional area, collects the predicted power grid supply energy, and generates a power planning strategy for each functional area based on the predicted energy consumption values, device energy consumption prediction values, and predicted power grid supply energy for each functional area. This invention achieves more accurate energy consumption prediction and more rational power allocation during the coal preparation process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid energy efficiency optimization, in particular to an energy efficiency optimization control method and system for intelligent coal selection. BACKGROUND

[0002] In the field of coal preparation industry, with the continuous improvement of intelligence and automation level, efficient and fine management of electric energy has become a core element to ensure production safety and reduce operating costs. To achieve this goal, the existing technology usually divides the coal preparation plant into multiple functional areas such as crushing, screening, and washing, and collects historical energy consumption data by deploying power metering terminals in each area. Based on these data, a time series prediction model (such as RNN, LSTM, etc.) is used to predict the future total energy consumption of each area, which is the current mainstream technical approach. This method can provide certain reference value when dealing with periodic and regular power consumption scenarios.

[0003] However, the actual electric load of the coal preparation plant is not a single smooth sequence, but a mixture of two loads with completely different characteristics: one is the basic steady load from auxiliary systems such as lighting and ventilation, and the other is the event-driven random load from core production equipment such as crushers, vibrating screens, and flotation machines. The start and stop of the latter is not determined by a fixed schedule, but is randomly triggered by discrete events in the production process, such as the arrival of a new batch of raw coal, changes in coal quality, and downstream process instructions. Its energy consumption in the time dimension is represented as unpredictable sharp pulses. The existing technology generally uses a single model to handle the mixed total load of the entire area. This one-size-fits-all approach creates a fundamental prediction blind spot when dealing with event-driven random loads. The model cannot learn and understand the internal triggering logic of this random load, and will mistakenly identify it as unpredictable noise and smooth or ignore it, thus systematically losing this key energy consumption information.

[0004] The existence of the prediction blind spot makes all power planning based on the prediction results passive and inaccurate. On the one hand, when a random load device (such as a crusher) suddenly starts, the actual power consumption will instantly exceed the predicted value, which can easily cause local power grid voltage drop, impact other sensitive devices, and even result in high punitive electricity charges due to exceeding the planned power consumption. On the other hand, to avoid this risk, excessive power margin must be reserved, which will inevitably lead to excess power supply during most of the time, resulting in serious energy waste. For example, although the Chinese patent with application publication number CN114977169A relates to power grid coordination control, it focuses on protection strategies after a fault occurs and does not solve the problem of how to fill this prediction blind spot, make forward-looking and differentiated predictions on mixed electric loads, and reasonably allocate them during normal production and operation.

[0005] Therefore, there is an urgent need in the field for a brand-new energy efficiency optimization control method to solve the above problems. SUMMARY

[0006] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application proposes an energy efficiency optimization control method and system for intelligent coal separation, which realizes more accurate energy consumption prediction and more reasonable power distribution in the coal separation process.

[0007] To achieve the above purpose, the present application proposes an energy efficiency optimization control method for intelligent coal separation, comprising the following steps:

[0008] Step one: collect the functional area layout in advance, and collect the regional historical energy consumption data in each functional area; for each functional area, collect a list of flexible devices;

[0009] Step two: for each functional area, based on the regional historical energy consumption data, construct an energy consumption training sample; and based on the energy consumption training sample, train an energy consumption prediction model;

[0010] Step three: count the historical start-stop data of each flexible device in each functional area;

[0011] Step four: in the actual process of coal separation, collect the real-time energy consumption feature data of each functional area, and based on the real-time energy consumption feature data and the energy consumption prediction model, output the regional energy consumption prediction value of each functional area;

[0012] Step five: based on the historical start-stop data of each flexible device, obtain the device energy consumption prediction value of the flexible devices in each functional area;

[0013] Step six: collect the predicted power grid supply energy, based on the regional energy consumption prediction value and the device energy consumption prediction value of each functional area and the predicted power grid supply energy, generate the power planning strategy of each functional area;

[0014] The way of collecting regional historical energy consumption data in each functional area is:

[0015] A power metering terminal is installed at the total incoming line of each functional area, which collects power feature data composed of various power parameters in each time period, and the power feature data in each time period constitutes the regional historical energy consumption data;

[0016] The way of constructing an energy consumption training sample based on the regional historical energy consumption data for each functional area is:

[0017] For each functional area:

[0018] Set the input sequence length L and the sliding window step S;

[0019] Mark any time period as t;

[0020] Mark the number of each power parameter as x, and the parameter value of the xth power parameter in the tth time period as x_t;

[0021] Construct a power parameter sequence for the xth power parameter: [x_(t-L+1), x_(t-L+2), …, x_t];

[0022] Mark the power consumption in the t+1th time period as a power consumption label;

[0023] The power parameter sequences of all power parameters form a set of input sample sets;

[0024] The set of input sample sets and the power consumption label form a set of energy consumption training samples;

[0025] Slide t backward by S days to re-construct a new set of energy consumption training samples;

[0026] The way to train the energy consumption prediction model based on the energy consumption training samples is:

[0027] For each functional area:

[0028] Take the input sample set of each time period as the input of the energy consumption prediction model, take the zone energy consumption prediction value of the time period after the time period as the output, take the power consumption label corresponding to the input sample set as the prediction target, take the square of the difference between the zone energy consumption prediction value and the power consumption label as the first prediction error, and take the mean of the first prediction error of all time periods as the training target. Train the energy consumption prediction model until the mean of the first prediction error converges, and stop training; the energy consumption prediction model is a time series prediction model;

[0029] The way to count the historical start-stop data of each flexible device in each functional area is:

[0030] For each flexible device in each functional area:

[0031] Collect the start-stop state parameters of the flexible device in each time period to form historical start-stop data, and use a database table to save the historical start-stop data, with the number of the flexible device as the primary key and each start-stop state parameter as the attribute;

[0032] The way to obtain the device energy consumption prediction value of the flexible device in each functional area based on the historical start-stop data of each flexible device includes the following steps:

[0033] Step 11: Extract additional time features from the timestamps of the historical start-stop data of each flexible device;

[0034] Step 12: Grouping the time periods with the same additional time characteristics to obtain the same type of time groups, and each same type of time group of each flexible device corresponds to a plurality of historical start-stop data;

[0035] Step 13: For each same type of time group of each flexible device:

[0036] Statistically counting the start-up time length of the flexible device in each time period in the same type of time group, and calculating the average value of the start-up time length as the expected time length of the flexible device in the same type of time group;

[0037] Statistically counting the energy consumption value of the flexible device in each time period in the same type of time group, and calculating the average value of the energy consumption value as the expected energy consumption of the flexible device in the same type of time group;

[0038] Statistically counting the average value of the total time length of each start-stop of the flexible device as the expected start-up time length of the flexible device;

[0039] Step 14: For the current time period in the actual coal preparation process, statistically counting the start-up time length of each flexible device as the start-up time length;

[0040] According to the additional time characteristics of the current time period, matching the corresponding same type of time group as the target time group;

[0041] Marking the number of each flexible device as i, and marking the start-up time length of the i-th flexible device as Qi;

[0042] Marking the expected start-up time length of the i-th flexible device as Ei;

[0043] Marking the expected time length corresponding to the target time group of the i-th flexible device as Yi, and marking the expected energy consumption as Wi;

[0044] If Qi is equal to 0, setting the device energy consumption prediction value as Wi;

[0045] If Qi is greater than Ei, setting the device energy consumption prediction value as 0;

[0046] If Qi is less than or equal to Ei, setting the device energy consumption prediction value as ;

[0047] The way of generating the power planning strategy of each functional area based on the regional energy consumption prediction value and the device energy consumption prediction value of each functional area and the predicted power grid supply energy is:

[0048] Marking the number of each functional area as j;

[0049] Mark the total power demand of the jth functional area as Zj; the total power demand Zj is the sum of the energy consumption prediction value of the jth functional area and the energy consumption prediction value of all flexible devices therein;

[0050] Mark the predicted power grid supply energy as H; set the supply energy weight fj for each functional area;

[0051] According to the predicted power grid supply energy, the supply energy weight of each functional area and the total power demand, calculate the electrical energy allocated to each functional area in the next time period, and the allocation strategy of the electrical energy allocated to each functional area is the power planning strategy.

[0052] The application proposes an energy efficiency optimization control system for intelligent coal selection, which comprises a historical data collection module, a model training module, an energy consumption prediction module and a strategy planning module; wherein each module is connected through an electrical way;

[0053] The historical data collection module collects regional historical energy consumption data in each functional area, and counts the historical start-stop data of each flexible device in each functional area, and sends the regional historical energy consumption data to the model training module and the historical start-stop data to the energy consumption prediction module;

[0054] The model training module constructs an energy consumption training sample based on the regional historical energy consumption data for each functional area, and trains an energy consumption prediction model based on the energy consumption training sample, and sends the energy consumption prediction model to the energy consumption prediction module;

[0055] The energy consumption prediction module collects real-time energy consumption characteristic data of each functional area in the actual coal selection process, outputs the regional energy consumption prediction value of each functional area based on the real-time energy consumption characteristic data and the energy consumption prediction model, and obtains the device energy consumption prediction value of the flexible devices in each functional area based on the historical start-stop data of each flexible device, and sends the regional energy consumption prediction value and the device energy consumption prediction value of each functional area to the strategy planning module;

[0056] The strategy planning module collects the predicted power grid supply energy, generates the power planning strategy of each functional area based on the regional energy consumption prediction value and the device energy consumption prediction value of each functional area and the predicted power grid supply energy.

[0057] Compared with the prior art, the application has the following beneficial effects:

[0058] The optimization control method provided by the present application solves the technical defect of "prediction blind area" caused by mixing the "basic steady load" and the "event-driven random load" in the prior art. Specifically, the present application first predicts the overall energy consumption of the functional area through a time series prediction model to accurately capture the basic steady load. In addition, the present application can make a forward-looking quantitative prediction of the peak pulse energy consumption, which was originally considered as unpredictable noise, by analyzing the expected duration and expected energy consumption of the equipment under different time modes and combining the current started duration of the equipment. By adding the regional energy consumption prediction value and the energy consumption prediction value of each flexible device to obtain the total power demand, the present application realizes accurate and comprehensive prediction of the total load of the region, filling the prediction blind area of the prior art.

[0059] Based on this accurate prediction result, the present application further brings more reasonable power planning and distribution effect. On the one hand, accurate total power demand prediction avoids energy waste caused by reserving too much power margin, and avoids the risk of impacting the power grid and causing penalties when the equipment starts due to insufficient prediction, improving the energy utilization efficiency and the stability of system operation. On the other hand, when the predicted power supply is insufficient, the present application adopts a dynamic allocation strategy based on the power supply weight, which can differentially allocate power according to the importance of each functional area, preferentially guaranteeing the power demand of the core production area, ensuring that the overall production efficiency and safety are affected to the minimum in the case of power shortage, and finally realizing fine and intelligent energy efficiency management of the whole coal selection process. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 The flowchart of the energy efficiency optimization control method of intelligent coal selection in embodiment 1 of the present application;

[0061] Figure 2 The module connection relationship diagram of the energy efficiency optimization control system of intelligent coal selection in embodiment 2 of the present application. DETAILED DESCRIPTION

[0062] The technical solutions of the present application will be described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0063] Embodiment 1

[0064] As shown in Figure 1 The energy efficiency optimization control method of intelligent coal selection includes the following steps:

[0065] Step 1: Pre-collect the functional area layout and collect historical energy consumption data for each functional area; for each functional area, collect a list of flexible devices;

[0066] Step 2: For each functional area, construct energy consumption training samples based on the area's historical energy consumption data; and train the energy consumption prediction model based on the energy consumption training samples.

[0067] Step 3: Compile historical start-stop data for each flexible device within each functional area;

[0068] Step 4: In the actual coal preparation process, collect real-time energy consumption characteristic data of each functional area, and output the regional energy consumption prediction value of each functional area based on the real-time energy consumption characteristic data and energy consumption prediction model.

[0069] Step 5: Based on the historical start-stop data of each flexible device, obtain the predicted energy consumption values ​​of the flexible devices in each functional area;

[0070] Step Six: Collect predicted grid energy supply, and generate power planning strategies for each functional area based on the predicted regional energy consumption and equipment energy consumption of each functional area and the predicted grid energy supply.

[0071] Specifically, the division and location of each functional area are obtained from the coal preparation plant design documents and site layout drawings, which serve as the layout of the functional areas. The functional areas include, but are not limited to, crushing area, screening area, washing and beneficiation area, dewatering area, power supply area, etc. Different functional areas are responsible for different functions in the coal preparation process, and the time period for each function and the energy consumed in each time period are also different. Therefore, the timeliness of each functional area needs to be considered when allocating power to different functional areas.

[0072] The method for collecting historical energy consumption data in each functional area is as follows:

[0073] A power metering terminal is installed at the main incoming line of each functional area. The power metering terminal collects power characteristic data composed of various power parameters for each time period. The power characteristic data for each time period constitutes the historical energy consumption data of the area.

[0074] The duration of the time period is set based on practical experience. Generally, it can be set to 1 hour. By setting the time period, the frequency of energy consumption prediction is reduced, thereby saving computing resources of the computing equipment.

[0075] The power characteristic data includes, but is not limited to: total active power (kW), total reactive power (kVar), total apparent power (kVA), total voltage (V), total current (A), and power factor of the region.

[0076] Furthermore, the flexible equipment in the flexible equipment list refers to equipment determined by the functions of different functional areas during the coal preparation process. At different times, the equipment can be turned on or off according to specific functional requirements, or its output power can be manually adjusted.

[0077] Specifically, flexible equipment in the crushing area includes crushers, conveyors, and dust removal fans; flexible equipment in the screening area includes vibrating screens and conveyors; flexible equipment in the washing and screening area includes flotation machines; flexible equipment in the dewatering area includes filter presses; and flexible equipment in the power supply area includes transformers and reactive power compensation devices.

[0078] Furthermore, the method for constructing energy consumption training samples for each functional area based on the area's historical energy consumption data is as follows:

[0079] For each functional area:

[0080] Set the input sequence duration L and the sliding window step size S;

[0081] A preferred example of the duration of L and S is as follows:

[0082] L=9, which contains power characteristic data of various power parameters over the past 9 time periods;

[0083] S=1, meaning that a sample is constructed every time period;

[0084] Let t be any time period;

[0085] The number of each power parameter is labeled as x, and the parameter value of the x-th power parameter in the t-th time period is labeled as x_t;

[0086] Construct a power parameter sequence for the x-th power parameter: [x_(t-L+1),x_(t-L+2),...,x_t];

[0087] The power consumption in the (t+1)th time period is used as the power consumption label.

[0088] Preferably, in order to improve the prediction accuracy, the power consumption time series of the past L days can also be used as a type of power parameter series;

[0089] The sequence of all power parameters forms a set of input samples;

[0090] A set of input samples and power consumption labels form a set of energy consumption training samples;

[0091] S days after t, a new set of energy consumption training samples is reconstructed.

[0092] Furthermore, the method for training the energy consumption prediction model based on energy consumption training samples is as follows:

[0093] For each functional region: the input sample set of each time period is used as the input of the energy consumption prediction model. The energy consumption prediction model outputs the predicted regional energy consumption value for the time period following that time period, uses the power consumption label corresponding to the input sample set as the prediction target, uses the squared difference between the predicted regional energy consumption value and the power consumption label as the first prediction error, and uses minimizing the mean of the first prediction error for all time periods as the training objective. The energy consumption prediction model is trained until the mean of the first prediction error converges, at which point training stops. The energy consumption prediction model is a time series prediction model. The time series prediction model can be either an RNN model or an LSTM model.

[0094] Furthermore, the method for statistically analyzing the historical start-stop data of each flexible device within each functional area is as follows:

[0095] For each flexible device in each functional area:

[0096] Collect the start-stop status parameters of the flexible device in each time period to form historical start-stop data, and use a database table with the flexible device number as the primary key and each start-stop status parameter as an attribute to save the historical start-stop data.

[0097] Preferably, the start / stop status parameters include, but are not limited to:

[0098] Functional area: refers to the functional area where the equipment is located;

[0099] Timestamp: Used to record the precise time of each start and stop;

[0100] Event type: Used to identify whether it is a start or stop event;

[0101] Runtime: refers to the duration from startup to shutdown each time;

[0102] Downtime duration: refers to the duration from each shutdown to the next startup;

[0103] Energy consumption data: used to record the energy consumption of each operating cycle.

[0104] It is understandable that by collecting start-stop status parameter data in various time periods, the probability distribution of future start-stop times can be derived.

[0105] Furthermore, the real-time energy consumption characteristic data includes: the time series of parameter values ​​of various power characteristic data in each functional area during the actual coal preparation process, within the first L time periods; that is, the real-time energy consumption characteristic data has the same form as the input sample set, the difference being the different collection time periods.

[0106] The method for outputting the predicted regional energy consumption values ​​for each functional area based on real-time energy consumption characteristic data and energy consumption prediction model is as follows:

[0107] For each functional area:

[0108] Real-time energy consumption characteristic data is input into the energy consumption prediction model of this functional area to obtain the regional energy consumption prediction value output by the energy consumption prediction model.

[0109] Furthermore, obtaining the predicted energy consumption values ​​of flexible devices in each functional area based on the historical start-stop data of each flexible device includes the following steps:

[0110] Step 11: Extract additional time features from the timestamps of the historical start-stop data of each flexible device; the additional time features are date, weekday or non-weekday, working hours or non-working hours, season, etc.

[0111] Step 12: Group time periods with the same additional time characteristics to obtain time groups of the same type. Each flexible device has multiple historical start and stop data corresponding to the same type of time group.

[0112] Step 13: For each time group of the same type for each flexible device:

[0113] The startup time of the flexible device in each time period within the same time group is statistically analyzed, and the average startup time is calculated as the expected startup time of the flexible device in the same time group.

[0114] The energy consumption of the flexible device is statistically analyzed for each time period in the same time group, and the average energy consumption is calculated as the expected energy consumption of the flexible device in the same time group.

[0115] The average total time from start to stop of the flexible device is calculated and used as the expected startup time of the flexible device.

[0116] Step 14: For the current time period in the actual coal preparation process, calculate the running time of each flexible device as the running time;

[0117] Based on the additional time characteristics of the current time period, match the corresponding time group of the same type as the target time group;

[0118] Each flexible device is labeled with the number i, and the startup duration of the i-th flexible device is labeled with Qi.

[0119] Let Ei be the expected startup time of the i-th flexible device;

[0120] The expected duration corresponding to the target time group of the i-th flexible device is denoted as Yi, and the expected energy consumption is denoted as Wi.

[0121] If Qi equals 0, it means that the flexible device has not been turned on, so the device energy consumption prediction value is set to Wi;

[0122] If Qi is greater than Ei, it means that the flexible device may take longer to start up than expected and may be shut down at any time. In this case, the device energy consumption prediction value should be set to 0.

[0123] If Qi is less than or equal to Ei, it means that the flexible device is still in the startup state during the current time period, and the device energy consumption prediction value is set to... .

[0124] It is understandable that (Ei-Qi) in the formula represents the expected startup time of the flexible device within the current time period. This represents the energy consumption per unit of time expected during startup.

[0125] Furthermore, the predicted grid energy supply can be predicted using the grid's historical power generation and power supply. The relevant prediction schemes are conventional technical means in this field, and will not be elaborated upon here.

[0126] Furthermore, the method for generating power planning strategies for each functional area based on the predicted regional energy consumption, predicted equipment energy consumption, and predicted grid power supply is as follows:

[0127] Label each functional area with the number j;

[0128] Let Zj be the total power demand of the j-th functional area; then the total power demand Zj is the sum of the predicted energy consumption of the j-th functional area and the predicted energy consumption of all flexible devices in it;

[0129] Right now: Rj is the predicted energy consumption of the j-th functional area, U(j) is the set of flexible devices in the j-th functional area, and Pi is the predicted energy consumption of the i-th flexible device.

[0130] The time-series forecasting model in step four accurately captures the steady-state and periodic power consumption patterns of equipment such as lighting and ventilation within the region, while the statistical forecasting method in step five quantifies the previously unpredictable "peak pulse" energy consumption of equipment such as crushers and vibrating screens. This dual-path method, combining regional steady-state forecasting with equipment stochastic forecasting, solves the "prediction blind spot" caused by the mixed processing of two types of loads in existing technologies, resulting in a more accurate and reliable prediction of total power demand compared to the single-model prediction results of existing technologies.

[0131] The predicted grid supply energy is labeled H; a supply weight fj is assigned to each functional region; the supply weight represents the penalty coefficient for power shortage in each functional region, with more important functional regions having larger supply weights. ;

[0132] Based on the predicted power grid supply, the power supply weight of each functional area, and the total power demand, the power energy allocated to each functional area in the next time period is calculated. The power energy allocation strategy for each functional area is the power planning strategy.

[0133] Preferably, the power planning strategy can be:

[0134] like Then the electrical energy allocated to the j-th functional area is Zj;

[0135] like The electrical energy allocated to the j-th functional area is then... It is understandable that for any functional area The value of fj is constant. Therefore, a functional area with a larger fj can obtain electrical energy closer to its total power demand, thus better meeting its power needs when the grid supply is insufficient. Furthermore, it can be found that... This allocation method ensures that all electrical energy provided by the power grid is distributed without any waste, thus maximizing the efficiency of power allocation.

[0136] Example 2

[0137] like Figure 2 As shown, the intelligent coal preparation energy efficiency optimization control system includes a historical data collection module, a model training module, an energy consumption prediction module, and a strategy planning module; the modules are connected to each other by electrical means.

[0138] The historical data collection module collects historical energy consumption data for each functional area and compiles historical start-stop data for each flexible device within each functional area. It then sends the historical energy consumption data to the model training module and the historical start-stop data to the energy consumption prediction module.

[0139] The model training module constructs energy consumption training samples for each functional area based on the area's historical energy consumption data; and trains an energy consumption prediction model based on the energy consumption training samples, and sends the energy consumption prediction model to the energy consumption prediction module.

[0140] The energy consumption prediction module collects real-time energy consumption characteristic data of each functional area during the actual coal preparation process. Based on the real-time energy consumption characteristic data and the energy consumption prediction model, it outputs the regional energy consumption prediction value of each functional area. Based on the historical start-stop data of each flexible device, it obtains the equipment energy consumption prediction value of the flexible devices in each functional area. The regional energy consumption prediction value and equipment energy consumption prediction value of each functional area are sent to the strategy planning module.

[0141] The strategy planning module collects and forecasts the grid energy supply, and generates power planning strategies for each functional area based on the regional energy consumption forecast, equipment energy consumption forecast, and grid energy supply forecast.

[0142] The method for collecting historical energy consumption data in each functional area is as follows:

[0143] A power metering terminal is installed at the main incoming line of each functional area. The power metering terminal collects power characteristic data composed of various power parameters for each time period. The power characteristic data for each time period constitutes the historical energy consumption data of the area.

[0144] The method for constructing energy consumption training samples for each functional area based on the area's historical energy consumption data is as follows:

[0145] For each functional area:

[0146] Set the input sequence duration L and the sliding window step size S;

[0147] Let t be any time period;

[0148] The number of each power parameter is labeled as x, and the parameter value of the x-th power parameter in the t-th time period is labeled as x_t;

[0149] Construct a power parameter sequence for the x-th power parameter: [x_(t-L+1),x_(t-L+2),...,x_t];

[0150] The power consumption in the (t+1)th time period is used as the power consumption label;

[0151] The sequence of all power parameters forms a set of input samples;

[0152] A set of input samples and power consumption labels form a set of energy consumption training samples;

[0153] S days after t, a new set of energy consumption training samples is reconstructed.

[0154] The method for training the energy consumption prediction model based on energy consumption training samples is as follows:

[0155] For each functional region: the input sample set of each time period is used as the input to the energy consumption prediction model. The energy consumption prediction model outputs the predicted regional energy consumption value for the time period following that time period, uses the power consumption label corresponding to the input sample set as the prediction target, uses the squared difference between the predicted regional energy consumption value and the power consumption label as the first prediction error, and uses minimizing the mean of the first prediction error for all time periods as the training objective. The energy consumption prediction model is trained until the mean of the first prediction error converges, at which point training stops. The energy consumption prediction model is a time series prediction model, which can be either an RNN model or an LSTM model.

[0156] The method for collecting historical start-stop data for each flexible device within each functional area is as follows:

[0157] For each flexible device in each functional area:

[0158] Collect the start-stop status parameters of the flexible device in each time period to form historical start-stop data, and use a database table with the flexible device number as the primary key and each start-stop status parameter as an attribute to save the historical start-stop data.

[0159] The method for outputting the predicted regional energy consumption values ​​for each functional area based on real-time energy consumption characteristic data and energy consumption prediction model is as follows:

[0160] For each functional area:

[0161] Real-time energy consumption characteristic data is input into the energy consumption prediction model of this functional area to obtain the regional energy consumption prediction value output by the energy consumption prediction model.

[0162] Furthermore, obtaining the predicted energy consumption values ​​of flexible devices in each functional area based on the historical start-stop data of each flexible device includes the following steps:

[0163] Step 11: Extract additional time features from the timestamps of the historical start-stop data of each flexible device; the additional time features are date, weekday or non-weekday, working hours or non-working hours, season, etc.

[0164] Step 12: Group time periods with the same additional time characteristics to obtain time groups of the same type. Each flexible device has multiple historical start and stop data corresponding to the same type of time group.

[0165] Step 13: For each time group of the same type for each flexible device:

[0166] The startup time of the flexible device in each time period within the same time group is statistically analyzed, and the average startup time is calculated as the expected startup time of the flexible device in the same time group.

[0167] The energy consumption of the flexible device is statistically analyzed for each time period in the same time group, and the average energy consumption is calculated as the expected energy consumption of the flexible device in the same time group.

[0168] The average total time from start to stop of the flexible device is calculated and used as the expected startup time of the flexible device.

[0169] Step 14: For the current time period in the actual coal preparation process, calculate the running time of each flexible device as the running time;

[0170] Based on the additional time characteristics of the current time period, match the corresponding time group of the same type as the target time group;

[0171] Each flexible device is labeled with the number i, and the startup duration of the i-th flexible device is labeled with Qi.

[0172] Let Ei be the expected startup time of the i-th flexible device;

[0173] The expected duration corresponding to the target time group of the i-th flexible device is denoted as Yi, and the expected energy consumption is denoted as Wi.

[0174] If Qi=0, it means that the flexible device has not been turned on, so set the device energy consumption prediction value to Wi;

[0175] If Qi > Ei, it means that the flexible device may take longer to start up than expected and may be shut down at any time. In this case, the device energy consumption prediction value should be set to 0.

[0176] If Qi ≤ Ei, it means that the flexible device is still in the startup state within the current time period, and the device energy consumption prediction value is set to... .

[0177] The method for generating power planning strategies for each functional area based on the predicted energy consumption of each functional area, the predicted energy consumption of equipment, and the predicted power grid supply is as follows:

[0178] Label each functional area with the number j;

[0179] Let Zj be the total power demand of the j-th functional area; then the total power demand Zj is the sum of the predicted energy consumption of the j-th functional area and the predicted energy consumption of all flexible devices in it;

[0180] Right now: Rj is the predicted energy consumption of the j-th functional area, U(j) is the set of flexible devices in the j-th functional area, and Pi is the predicted energy consumption of the i-th flexible device.

[0181] Label the predicted grid power supply as H; set the power supply weight fj for each functional area;

[0182] Based on the predicted power grid supply, the power supply weight of each functional area, and the total power demand, the power energy allocated to each functional area in the next time period is calculated. The power energy allocation strategy for each functional area is the power planning strategy.

[0183] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0184] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0185] The preset parameters or preset thresholds mentioned above are all set by those skilled in the art based on actual conditions or obtained through large-scale data simulation.

[0186] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An energy efficiency optimization control method for intelligent coal separation, characterized in that, The method comprises the following steps: Step 1: collecting functional area layout in advance, collecting area historical energy consumption data in each functional area, and collecting a list of flexible devices in each functional area; Step 2: constructing an energy consumption training sample based on the area historical energy consumption data for each functional area; training an energy consumption prediction model based on the energy consumption training sample; Step 3: counting historical start-stop data of each flexible device in each functional area; Step 4: collecting real-time energy consumption feature data of each functional area in the actual coal selection process, and outputting area energy consumption prediction values of each functional area based on the real-time energy consumption feature data and the energy consumption prediction model; Step 5: obtaining device energy consumption prediction values of the flexible devices in each functional area based on the historical start-stop data of each flexible device; Step 11: extracting additional time features from the time stamps of the historical start-stop data of each flexible device; Step 12: grouping time periods with the same additional time features to obtain the same type of time groups, and each same type of time group of each flexible device corresponds to a plurality of historical start-stop data; Step 13: for each same type of time group of each flexible device: counting the start-up time length of the flexible device in each time period in the same type of time group, and calculating the average value of the start-up time length as the expected time length of the flexible device in the same type of time group; counting the energy consumption value of the flexible device in each time period in the same type of time group, and calculating the average value of the energy consumption value as the expected energy consumption of the flexible device in the same type of time group; counting the average value of the total time length of each start-stop of the flexible device as the expected start-up time length of the flexible device; Step 14: for the current time period in the actual coal selection process, counting the start-up time length of each flexible device as the start-up time length; matching the corresponding same type of time group as the target time group according to the additional time features of the current time period; marking the number of each flexible device as i, and marking the start-up time length of the i-th flexible device as Qi; marking the expected start-up time length of the i-th flexible device as Ei; marking the expected time length corresponding to the target time group of the i-th flexible device as Yi, and marking the expected energy consumption as Wi; if Qi is equal to 0, setting the device energy consumption prediction value as Wi; if Qi is greater than Ei, setting the device energy consumption prediction value as 0; If Qi is less than or equal to Ei, the device energy consumption prediction value is set to ; Step 6: collecting predicted power supply energy, generating power planning strategies of each functional area based on the area energy consumption prediction values and the device energy consumption prediction values of each functional area and the predicted power supply energy.

2. The energy efficiency optimization control method for intelligent coal separation according to claim 1, characterized in that, The area historical energy consumption data is collected in each functional area in the following manner: a power metering terminal is installed at the total incoming line of each functional area, the power metering terminal collects power feature data composed of various power parameters in each time period, and the power feature data in each time period constitutes the area historical energy consumption data.

3. The energy efficiency optimization control method for intelligent coal separation according to claim 2, characterized in that, The energy consumption training sample is constructed based on the area historical energy consumption data for each functional area in the following manner: For each functional area: set the input sequence length L and the sliding window step S; mark any time period as t; mark the number of each power parameter as x, and the parameter value of the xth power parameter in the tth time period as x t; construct a power parameter sequence for the xth power parameter: [x (t-L+1), x (t-L+2), …, x t]; take the power consumption in the t+1th time period as the power consumption label; the power parameter sequences of all power parameters form a set of input sample sets; the set of input sample sets and the power consumption label form a set of energy consumption training samples; slide t by S days to re-construct a new set of energy consumption training samples.

4. The energy efficiency optimization control method for intelligent coal separation according to claim 3, characterized in that, The manner of training the energy consumption prediction model based on the energy consumption training samples is: For each functional area: take the input sample set of each time period as the input of the energy consumption prediction model, the energy consumption prediction value of the time period after the time period as the output, the power consumption label corresponding to the input sample set as the prediction target, the square of the difference between the energy consumption prediction value and the power consumption label as the first prediction error, and the mean of the first prediction errors of all time periods as the training target; train the energy consumption prediction model until the mean of the first prediction errors converges, and then stop training; the energy consumption prediction model is a time series prediction model.

5. The energy efficiency optimization control method for intelligent coal separation according to claim 4, characterized in that, The manner of statistically collecting the historical start-stop data of each flexible device in each functional area is: For each flexible device in each functional area: collect the start-stop state parameters of the flexible device in each time period to form historical start-stop data, and use a database table to save the historical start-stop data, taking the number of the flexible device as the primary key and each start-stop state parameter as the attribute.

6. The energy efficiency optimization control method for intelligent coal selection according to claim 5, characterized in that, The manner of generating the power planning strategy of each functional area based on the regional energy consumption prediction value, the device energy consumption prediction value, and the predicted grid supply energy of each functional area is: Mark the number of each functional area as j; Mark the total power demand of the jth functional area as Zj; the total power demand Zj is the sum of the regional energy consumption prediction value of the jth functional area and the device energy consumption prediction value of all flexible devices therein; Mark the predicted grid supply energy as H; set a supply energy weight fj for each functional area; According to the predicted grid supply energy, the supply energy weight of each functional area, and the total power demand, calculate the allocated energy for each functional area in the next time period, and the allocation strategy of the allocated energy of each functional area is the power planning strategy.

7. The energy efficiency optimization control system for intelligent coal separation, which is used to realize the energy efficiency optimization control method for intelligent coal separation according to any one of claims 1-6, characterized in that, It comprises a historical data collection module, a model training module, an energy consumption prediction module, and a strategy planning module; each module is connected through an electrical method; The historical data collection module collects regional historical energy consumption data in each functional area, and statistically collects the historical start-stop data of each flexible device in each functional area, sends the regional historical energy consumption data to the model training module, and sends the historical start-stop data to the energy consumption prediction module; The model training module constructs an energy consumption training sample based on historical energy consumption data of each functional area, trains an energy consumption prediction model based on the energy consumption training sample, and sends the energy consumption prediction model to the energy consumption prediction module; The energy consumption prediction module collects real-time energy consumption feature data of each functional area in the actual coal preparation process, outputs an area energy consumption prediction value of each functional area based on the real-time energy consumption feature data and the energy consumption prediction model, obtains a device energy consumption prediction value of a flexible device in each functional area based on historical start-stop data of each flexible device, and sends the area energy consumption prediction value and the device energy consumption prediction value of each functional area to the strategy planning module; The strategy planning module collects a predicted power grid supply energy, generates a power planning strategy for each functional area based on the area energy consumption prediction value and the device energy consumption prediction value of each functional area and the predicted power grid supply energy.

Images