Energy-saving control system of coal mining equipment for continuous mining and continuous filling of coal mine

Abstract

The invention discloses an energy-saving control system for coal mining equipment for continuous mining and continuous filling of a coal mine, and relates to the technical field of intelligent control of the coal mine. A multi-parameter acquisition module is used for acquiring multi-source state data such as equipment running current, vibration signals, shuttle car tracks and operation shifts in real time; the artificial intelligence analysis module performs feature extraction and state recognition on the data based on a trained random forest model, and outputs an idling prediction label and a confidence score; the intelligent idling judgment module adopts a confidence weighting mechanism and trend residual analysis to judge whether to enter an idling state or not; the energy-saving control execution module executes a grading response strategy according to the idling severity level; the pre-judgment and intervention module judges whether feeding is about to be recovered or not through track trend modeling, current jump recognition and cycle matching analysis, and stops shutdown or starts preheating standby according to the judgment result; the system has the advantages of high recognition precision, high response flexibility, remarkable energy-saving effect and the like, and is suitable for efficient energy-saving control in a complex coal mine operation environment.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine intelligent control, and in particular to an energy-saving control system for continuous mining and continuous filling coal mining equipment in a coal mine. Background Art

[0002] With the continuous development of smart mining and green coal mining, continuous mining and filling processes are playing an increasingly important role in improving resource recovery and controlling goaf areas. In this process, crushers and conveying equipment are key components in achieving continuous conveying and filling of mined coal. Because the feeding rhythm of shuttle cars at coal mines is often intermittent due to factors such as terrain, mining progress, and shift scheduling, crushing and conveying equipment often runs idle for extended periods. This not only increases energy consumption and causes equipment wear, but also affects overall system efficiency.

[0003] Existing technologies often use a "timed delayed shutdown" approach to identify idling. Specifically, if the equipment runs continuously for a fixed period (e.g., 60 seconds) without detecting coal or a car approaching, it is identified as idling and a shutdown command is issued. However, this control strategy fails to fully account for operating fluctuations in the shuttle car feeding interval. Especially during night shifts, low loads, or remote operations, the shuttle car interval may exceed the specified threshold, leading to false shutdowns and frequent starts and stops, severely limiting the system's energy efficiency and stability.

[0004] Therefore, it is urgent to propose an intelligent control system based on multi-parameter adaptive identification, which can combine coal mining condition data to determine whether it is in idling state in real time and dynamically adjust the shutdown strategy, so as to achieve precise energy-saving control and improve the operating efficiency of coal mine mining and transportation systems. Summary of the Invention

[0005] The purpose of the present invention is to provide an energy-saving control system for continuous mining and continuous filling coal mining equipment in a coal mine, so as to solve the shortcomings of the background technology.

[0006] In order to achieve the above-mentioned purpose, the present invention provides the following technical solution: an energy-saving control system for continuous mining and continuous filling coal mining equipment in a coal mine, comprising:

[0007] A multi-parameter acquisition module is used to collect multi-source operating status data related to the continuous mining and charging system in coal mines, including the distance between the shuttle car and the equipment, the current load of the crusher and conveyor, operating vibration values, material flow detection signals, and the current operating shift type;

[0008] An artificial intelligence analysis module, based on a pre-trained random forest model, performs feature extraction and time series modeling on the operating status data, and outputs a predicted probability value and a credibility score for whether the coal mining equipment is in an idling state within the current time window;

[0009] The intelligent idling judgment module uses a confidence-weighted judgment mechanism to determine whether the vehicle has entered the idling state based on the idling prediction results in multiple consecutive time windows. If the set judgment conditions are met, an energy-saving control instruction is generated;

[0010] An energy-saving control execution module executes a hierarchical control strategy after receiving an energy-saving control instruction;

[0011] The predictive intervention module is used to determine whether the system is about to resume feeding based on information such as the shuttle trajectory trend, current fluctuation rebound, and historical feeding cycles. If it is determined that the system is about to resume feeding, the current shutdown logic is interrupted or the system is switched to preheating standby mode.

[0012] Preferably, the multi-parameter acquisition module includes:

[0013] Collect the operating current signals of the crusher and conveyor, and extract the load fluctuation feature vector through short-time Fourier transform to characterize the typical current pattern of the equipment when it is unloaded;

[0014] Real-time measurement of the dynamic distance change between the shuttle car and the crusher, and filtering of environmental interference through a differential sliding window algorithm to achieve continuous identification of the shuttle car's dwell time;

[0015] Obtain the equipment's operating vibration value and identify the equipment's micro-vibration trend when it is not operating by clustering its frequency domain envelope energy.

[0016] A temporal association rule model is constructed by integrating the collection timestamp and mining operation scheduling data to identify the operation shift types and analyze the periodic structure.

[0017] Preferably, the artificial intelligence analysis module includes:

[0018] Based on the collected multi-source operating status data, a multidimensional feature vector set X is constructed, which consists of current fluctuation amplitude, shuttle car parking time, vibration power spectrum density, material flow interruption period and operation shift label.

[0019] Using a random forest classification model that has been trained offline, the feature vector set X is input into the model structure. Each decision tree outputs a preliminary classification result for the idling state. The model as a whole aggregates the results through a majority voting mechanism and generates an idling prediction label.

[0020] In response to the fluctuations in the output of each decision tree, the confidence score β of the predicted label is calculated and dynamically adjusted based on the leaf node information gain entropy value and the weighted distribution of feature importance;

[0021] The predicted label and confidence score are combined to form an output vector, which serves as the decision basis for idling determination.

[0022] Preferably, the intelligent idling determination module includes:

[0023] Construct a state prediction window sequence W = {Y1, Y2, ..., Y n}, each Yᵢ represents the idling prediction result output by the artificial intelligence analysis module in the i-th time slice and its corresponding confidence βᵢ, and n is the total number of time slices;

[0024] A dynamic weight function ωᵢ = f(βᵢ, Δtᵢ) is set, where Δtᵢ is the delay of the time slice relative to the current time. This function is used to give higher weight to recent prediction results, thus implementing a credibility weighting mechanism that enhances temporal priority.

[0025] Aggregate all weighted results in the window sequence to obtain the total confidence score γ = Σ(ωᵢ·Yᵢ) and compare it with the adaptive threshold θ. If γ ≥ θ, it is determined to be in a stable idling state.

[0026] The energy-saving control instruction generation module is triggered and the status is passed to the control execution unit, so that the equipment energy-saving shutdown is executed only when both idling continuity and confidence stability are met.

[0027] Preferably, the intelligent idling determination module further includes:

[0028] Construct a set of lightweight time series residual prediction models that run in parallel, use a one-dimensional convolutional network to model the state change trend in the prediction window sequence W, and output the sequence trend prediction value and its relationship with reality The residual ε between

[0029] If the residual ε in two or more consecutive time slices exceeds the preset fluctuation tolerance threshold ε t , judge that there is nonlinear mutation or short-term interference in the current prediction state, suspend the current idling judgment process and enter the self-stabilization observation period;

[0030] During the observation period, an improved exponentially weighted smoothing algorithm is used to perform time series correction on the γ value to reduce the impact of short-term sharp fluctuations on the judgment results.

[0031] Preferably, the energy-saving control execution module includes:

[0032] Analyze the priority tags and confidence level parameters in the energy-saving control instructions, and divide the control objects into core power units, auxiliary feeding units, and control support units according to their functional links;

[0033] Build a hierarchical execution strategy based on priority tags, and use a deeply configurable state mapping table to map idle severity levels to grouped device states;

[0034] During the execution process, a dynamic unloading curve function is introduced to adjust the unloading time difference of each unit according to the current load level and the energy consumption rate reduction model;

[0035] A state holding lock mechanism is set. If the upstream feeding trend recovery signal is detected during the unloading process, the shutdown process will be immediately terminated and switched to the preheating standby state.

[0036] Preferably, the prejudgment intervention module includes:

[0037] A shuttle trajectory trend model based on time series embedding representation is constructed. By performing multi-scale convolution feature extraction on the shuttle displacement data, the velocity vector and acceleration trend curve of the current trajectory change are generated.

[0038] Combined with the real-time current fluctuation sequence of the conveying equipment, a jump recognition algorithm is used to detect whether there is a typical rebound feature, that is, a slight increase in current exceeding the set threshold in the idling state, prompting the mechanical component to pre-start action;

[0039] A historical feeding cycle database is introduced, and a pattern matching neural network is used to analyze whether the current period coincides with the high-frequency section of the typical feeding cycle, so as to determine whether it has entered a high-incidence feeding period;

[0040] If any two of the three trend information meet the judgment conditions for resuming feeding, the current energy-saving shutdown process will be suspended, and the equipment preheating standby strategy will be started according to the system temperature rise model to realize the switching from idling control to recovery control.

[0041] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0042] The energy-saving control system for continuous mining and continuous filling coal mining equipment provided by this invention utilizes multi-parameter data acquisition, artificial intelligence recognition, and a confidence-weighted idling discrimination mechanism. This system breaks through the traditional control methods that rely on fixed time delay judgment and single signal triggering, achieving precise identification of idling conditions and flexible energy-saving control. The system significantly improves recognition accuracy through multi-dimensional feature vector construction and random forest model inference. Furthermore, it utilizes residual trend analysis and a dynamic smoothing algorithm to effectively reduce the rate of false stops caused by data fluctuations, ensuring the stability and interpretability of the control response.

[0043] 2. This invention innovatively integrates predictive mechanisms such as shuttle trajectory trend analysis, current spike identification, and feed cycle matching to create a proactive intervention module capable of proactive intervention, shifting energy-saving control from passive response to proactive management. Through a hierarchical control execution strategy and preheat standby logic, the system improves energy efficiency and operational flexibility, making it widely adaptable to the multi-shift and highly volatile coal mine operating environment. While ensuring equipment safety and stable operation, it achieves significant energy savings and engineering value. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0045] Figure 1 This is a mind map of the system modules of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0047] Example 1, please refer to Figure 1 As shown, the energy-saving control system for continuous mining and continuous filling coal mining equipment in this embodiment includes:

[0048] A multi-parameter acquisition module is used to collect multi-source operating status data related to the continuous mining and charging system in coal mines, including the distance between the shuttle car and the equipment, the current load of the crusher and conveyor, operating vibration values, material flow detection signals, and the current operating shift type;

[0049] An artificial intelligence analysis module, based on a pre-trained random forest model, performs feature extraction and time series modeling on the operating status data, and outputs a predicted probability value and a credibility score for whether the coal mining equipment is in an idling state within the current time window;

[0050] The intelligent idling judgment module uses a confidence-weighted judgment mechanism to determine whether the vehicle has entered the idling state based on the idling prediction results in multiple consecutive time windows. If the set judgment conditions are met, an energy-saving control instruction is generated;

[0051] An energy-saving control execution module executes a hierarchical control strategy after receiving an energy-saving control instruction;

[0052] The predictive intervention module is used to determine whether the system is about to resume feeding based on information such as the shuttle trajectory trend, current fluctuation rebound, and historical feeding cycles. If it is determined that the system is about to resume feeding, the current shutdown logic is interrupted or switched to preheating standby mode.

[0053] The multi-parameter acquisition module includes at least the following four types of data channels and processing steps:

[0054] This part is used to collect the operating current signals of the crusher and conveyor. The collected signals are connected to the system through an intelligent current sensing terminal (such as a three-phase mutual induction sampler) and are continuously collected at a sampling frequency of not less than 200Hz to ensure the complete capture of small fluctuations and idle waveforms.

[0055] To extract stable features from the raw current signal that can be used to identify the idling state, the system uses a short-time Fourier transform (STFT) algorithm to analyze the current signal in the time and frequency domains. Within each fixed window (e.g., 2 seconds), the signal is segmented using a window-weighted Fourier transform, and the frequency domain characteristic amplitude spectrum is extracted.

[0056] Next, by setting spectral feature thresholds and normalized offset criteria, a load fluctuation feature vector set was constructed. This set includes features such as the current main frequency distribution, frequency offset rate, low-frequency noise amplitude, and amplitude fluctuation mean square error. These features are used to characterize the device's operating modes under no-load, light-load, and loaded conditions. Experiments have shown that the current spectrum distribution during no-load operation is characterized by low amplitude, broadband, and stability. This feature vector significantly improves the state recognition accuracy of subsequent models.

[0057] To continuously monitor the shuttle's docking status, a multi-array laser ranging sensor system is deployed in the crusher's feed area. Each array consists of three independent ranging units, covering a range of 0 to 20 meters from the front of the crusher. The sensor data sampling period is 50ms, and multi-channel time synchronization is supported.

[0058] To avoid false triggering and false data disturbances caused by factors like coal dust and humidity in the mine environment, the system applies a differential sliding window algorithm to the raw distance data for filtering. This algorithm calculates the difference ΔD between the current and previous window's mean distance measurement, using a sliding window period of one second. If ΔD is less than a dynamic jitter threshold (e.g., 0.15 meters) for more than three consecutive window periods, the shuttle is considered stationary.

[0059] Finally, the shuttle car stop time characteristic item T_car is generated by accumulating the continuous stop time, which serves as an auxiliary basis for determining whether the shuttle car is feeding, and is linked with the current signal to analyze the start and end time of feeding.

[0060] Considering that the equipment is in operation but lacks material medium transmission in the idling state, the vibration characteristics often show an atypical pattern of low amplitude, single frequency peak and small fluctuation. The system is set up to install high-sensitivity MEMS inertial sensors (3 axes) to collect operating vibration data.

[0061] After collection, the vibration signal is extracted by frequency domain envelope analysis algorithm. First, the acceleration signal is subjected to fast Fourier transform (FFT), and the energy distribution curve in the frequency range is extracted. Then, a clustering algorithm (such as K-means) is used to classify and train the vibration envelope spectrum, and three modes of "high-load vibration type", "medium-load stable type" and "empty-micro-vibration type" are divided.

[0062] The system takes the similarity score of matching the current vibration spectrum with the "empty-micro-vibration type" as one of the model input features, to improve the accuracy of the idle state determination under low speed or intermittent working conditions.

[0063] Considering the significant influence of the coal mine shift system on the shuttle car feeding rhythm, the system introduces a time dimension feature processing mechanism. First, the timestamp information of each working condition data is collected, and the mine operation shift schedule table is digitized and modeled to distinguish between day shift, night shift and maintenance period, etc.

[0064] Based on the collection timestamp and shift data, a time sequence association rule model is constructed, and the Apriori algorithm is used to extract the typical feeding interval distribution interval and shuttle car average in-out cycle T_batch under each shift. The model outputs the "operation shift type label" and its related cycle factor as auxiliary feature input of the artificial intelligence analysis module.

[0065] By introducing this type of non-process signal data, the system's forward-looking perception ability for working condition mode changes can be significantly enhanced, enabling the AI model to have context prior perception ability, improving its robustness and fault tolerance under special conditions such as night shift.

[0066] Through the integration of the above four types of data collection and processing methods, the multi-parameter collection module realizes multi-dimensional perception and high-precision extraction of key working condition variables in coal mine continuous mining and filling operations. Especially in the identification of idle state, it no longer relies on a single distance or current signal, but constructs a composite data structure that integrates multi-source information such as electric frequency vibration features, dynamic trajectory changes, and non-structured time sequence rules.

[0067] The artificial intelligence analysis module described in the present application is the core intelligent discrimination unit of the coal mine continuous mining and filling coal mining equipment energy-saving control system, mainly used to construct an idle state recognition model based on multi-source operating state data, and output the prediction label and confidence score of the idle state as the basis for subsequent control decisions.

[0068] The module uses a trained random forest classification model as the recognition main algorithm, integrates the high-dimensional data obtained by the multi-parameter collection module, and realizes intelligent discrimination of whether the current operating state of the coal mining equipment is idle through feature engineering, model reasoning and dynamic confidence calculation, etc.

[0069] To achieve high-precision state recognition, the artificial intelligence analysis module first aggregates and preprocesses the original signals from the multi-parameter acquisition module to construct structured feature input data. This process includes the following steps:

[0070] Current fluctuation amplitude extraction: Through short-time Fourier transform (STFT) analysis, the amplitude spectrum of the crusher and conveyor current signal is obtained, from which indicators such as low-frequency band mean, medium-frequency fluctuation amplitude, and high-frequency energy proportion are extracted to form current fluctuation features (denoted as Ivar) reflecting the degree of load fluctuation.

[0071] Shuttle car stop duration statistics: Combined with the differential sliding window algorithm, the continuous stationary time of the shuttle car in front of the crusher is calculated to form a stable stop duration index (denoted as Tcar), which is used to quantify whether the shuttle car is in an effective feeding state.

[0072] Vibration power spectral density estimation: Through frequency domain envelope clustering analysis method, the characteristic frequency band energy distribution of the inertial sensor signal is extracted to obtain the vibration power spectral density vector (denoted as Vpsd), which is used to identify the typical micro-vibration mode under the condition of empty load of the device.

[0073] Material flow interruption period calculation: From the material flow sensor, the continuous no-material signal time sequence of the belt section is analyzed to calculate the periodicity characteristics of the current material flow interruption (denoted as Fgap) as one of the input features.

[0074] Work shift label archiving: Through the time sequence association rule model, the work shift type corresponding to the current period (such as day shift, night shift, maintenance shift) is extracted to form a discrete shift label variable (denoted as Btype), which is used to correct the judgment prior of the model.

[0075] Finally, the artificial intelligence analysis module constructs a multi-dimensional feature vector set containing the above five types of core features: X={Ivar, Tcar, Vpsd, Fgap, Btype}.

[0076] The idle recognition model adopted by the present application is a multi-classification random forest model (Random Forest Classifier), which is composed of a plurality of decision trees with structural independence (set as N trees, denoted as T1, T2,..., TN). Each tree learns different subsets of the sample feature space to construct different discrimination paths.

[0077] In the model inference stage:

[0078] The feature vector set X is input to each decision tree;

[0079] Each decision tree independently completes the classification task of whether the current device is in idle state, and outputs a binary label Yi∈{0,1}, where 1 represents idle;

[0080] The model as a whole integrates the results of each tree through the majority voting mechanism and outputs the final idle prediction label Yfinal.

[0081] To improve the reliability of the system's judgment of boundary conditions, the AI ​​analysis module introduces a confidence scoring mechanism β to measure the credibility of the model's output labels. This scoring is based on the following two dimensions:

[0082] Classification consistency measurement: Count the proportion r of the label "idle" in all decision tree output results. If r is close to 1 or 0, it means that the consistency is high and the β value is large; if r is close to 0.5, it means that the prediction divergence is large and the β value is low.

[0083] Feature contribution weight correction: Based on the feature importance distribution vector ϕ={ϕ1,ϕ2,...,ϕm} statistically calculated during the training process, combined with the deviation value and information gain contribution of each feature in the current input X, the final confidence result is adjusted at the model level to obtain: ;in Represents the leaf node information gain corresponding to the j-th feature.

[0084] Finally, the artificial intelligence analysis module combines the idling prediction label Yfinal with the confidence score β to form an output vector: O=[Yfinal,β]. This output serves as the input basis for the system idling judgment module to determine whether to trigger the energy-saving control strategy.

[0085] Through the construction and inference process of the aforementioned AI analysis module, this invention effectively integrates operational data from multiple dimensions. Leveraging the diversity and robustness of random forests, it addresses the issues of traditional idling recognition technology, which relies on a single threshold, suffers from poor adaptability, and is prone to misjudgment. Furthermore, the introduction of a confidence scoring mechanism enables the system to self-regulate when determining edge states, improving recognition stability and the accuracy of energy-saving control.

[0086] The intelligent idling identification module described in this invention constructs a time-sliding prediction structure based on the state prediction results output by the artificial intelligence analysis module, dynamically determining whether the equipment is in a stable idling state. By combining a multi-time-slice confidence weighting mechanism with a trend residual compensation mechanism, a highly robust intelligent identification process with a low false positive rate is achieved, providing a highly interpretable and reliable judgment basis for energy-saving control strategies.

[0087] In order to introduce time continuity constraints and dynamic weight correction mechanisms, the system sets a sliding time window structure, and the sampling period is set to a fixed duration (for example, 5 seconds). In each period, the idling state prediction result and the corresponding confidence score output by the artificial intelligence analysis module are recorded once, forming a time series structure as follows: ;in, represents the idling prediction label in the i-th time slice (1 for idling, 0 for non-idling), This is the confidence score for the tag. n is the time window length, with a recommended range of 612 time slices (3060 seconds). This structure is used to describe the state evolution path and provide timing support in subsequent weight calculation and trend modeling.

[0088] In order to achieve intelligent weighting of state change trends, the system sets the dynamic weight function as follows: ;in, represents the delay of the i-th time slice relative to the current moment; λ is a time decay factor, which is used to control the influence of old data. This mechanism makes predictions closer to the current state more influential, thereby improving response sensitivity and reducing sensitivity to early interference values.

[0089] Next, the system calculates the weighted total confidence score γ: ; and compare γ with the set adaptive threshold θ (θ is dynamically adjusted according to the shift and environment, and the general value range is 0.6~0.8). If γ≥θ is satisfied, the system determines that the current coal mining equipment is in a stable idling state and triggers the generation process of energy-saving control instructions.

[0090] In order to further enhance the system's ability to identify and suppress nonlinear fluctuations or sudden interference, the intelligent idling discrimination module introduces a set of lightweight time series residual prediction models running in parallel.

[0091] The model is based on a one-dimensional convolutional network (1D-CNN) structure and uses the state label in the current window sequence W As input, for the next time slice Perform trend forecasts and generate forecast values . The prediction residuals are then calculated: If the system detects that the residual ε in two or more consecutive time slices exceeds the preset fluctuation tolerance threshold εt (for example, 0.3), it determines that the current state has "nonlinear mutation" or "short-term signal interference", suspends the original idling judgment process, and enters the self-stabilization observation period.

[0092] During the observation period, the system uses an improved exponentially weighted smoothing algorithm (EWMA+) to perform time-weighted correction on the historical γ value to suppress the impact of abnormal volatility on the final judgment result. The smoothing formula is: ; where α is the adaptive smoothing factor, which is dynamically adjusted according to the current residual size to enhance the elastic response capability to sudden interference. Indicates the total idling confidence score calculated at the current time (the t-th time slice), without smoothing. Indicates the total confidence score of idling after smoothing at the current moment, that is, the value corrected by EWMA. Indicates the last moment ( The smoothed score value of the time slice is the “historical inertia” of the current γ′.

[0093] If the state fluctuation returns to stability during the observation period and the residual ε continuously returns to the normal range, the system will restore the original judgment logic and re-judge whether to enter the energy-saving control state based on the corrected γ value.

[0094] By leveraging the aforementioned intelligent idling identification module's timing weighting strategy and trend residual modeling, this invention effectively integrates static prediction results with dynamic trend analysis capabilities, achieving continuous identification of idling states, credibility integration, and anti-interference correction. This mechanism overcomes the limitations of traditional idling identification, which relies on a single output point or fixed time threshold, significantly improving the system's energy-saving response reliability, state identification accuracy, and adaptability.

[0095] Especially in scenarios where the feeding rhythm of coal mine shuttle cars fluctuates greatly and signals are frequently interfered with, the module demonstrates good robustness and engineering applicability, providing stable and highly confident logical support for downstream energy-saving control strategies.

[0096] In the energy-saving control system of the continuous mining and continuous filling coal mining equipment in the coal mine described in the present invention, the energy-saving control execution module is used to orderly shut down or switch different types of equipment units to a preheating standby state based on the control instructions output by the intelligent idling judgment module according to a preset grading strategy after the system determines that it is currently in an idling state and the confidence level meets the standard.

[0097] The control instructions received by the energy-saving control execution module come from the intelligent idling judgment module and include the following two core fields:

[0098] Priority Label (P_tag): This tag indicates the severity of the current idling state. It is divided into three levels (L1 - low idling; L2 - medium idling; L3 - severe idling), corresponding to different degrees of energy-saving response actions;

[0099] Confidence Score (β): This is the idling prediction credibility score output by the AI ​​analysis module (value range: 0-1), which is used to determine the execution strength and safety margin of the current control action.

[0100] The energy-saving control execution module analyzes the instruction and starts the corresponding control strategy logic.

[0101] To improve the response resolution and flexibility of energy-saving control, the present invention groups the entire set of continuous mining and charging equipment according to the functional path of the control object, mainly including:

[0102] Core power unit: such as crusher main motor, main conveyor belt drive;

[0103] Auxiliary feeding units: such as coal storage tank vibrators, trolley loading and unloading equipment;

[0104] Control support unit: such as PLC logic control cabinet, signal communication equipment, solenoid valve control circuit, etc.

[0105] Each unit is connected to a deeply configurable state map that defines the response actions of each unit to different idling severity levels (P_tag), for example:

[0106] P_tag core power unit Auxiliary feeding unit Control support unit L1 Keep running Reduced speed operation Keep All L2 Reduced frequency operation Intermittent energy supply Partial disconnection L3 Shutdown and power outage Stop unloading Main circuit standby

[0107] The mapping table is configured by system engineering parameter setting personnel according to actual process requirements and can be updated online, with high flexibility and deployment efficiency.

[0108] During the equipment shutdown process, to avoid current shock, system voltage fluctuation, and equipment mechanical stress concentration caused by sudden power outages, this paper introduces a dynamic unloading function control strategy based on the operating load and energy consumption curve. The core concept is:

[0109] Calculate the recommended unloading time difference ΔT based on the current operating current (I) and energy consumption rate (E) of the equipment unit and the remaining transportation task (such as the amount of coal to be carried);

[0110] Introduce the unloading curve function U(t), such as: ;in is the current load percentage, k is the unloading coefficient, and t is the control time. The system dynamically adjusts the shutdown sequence and delay time of each controlled object according to the curve, forming a gradient, non-sudden shutdown process to improve safety and control stability.

[0111] Considering that the idling state may be misjudged due to instantaneous fluctuations, in order to avoid the false triggering of shutdown when the feeding is about to be resumed, the present invention establishes a state holding lock mechanism in the energy-saving control execution module, which specifically includes:

[0112] During the unloading process, real-time monitoring of feedback signals from the prediction and intervention module is performed, including shuttle acceleration trajectory, current rebound trend, and operation high-frequency cycle matching signals.

[0113] If the probability score of the feeding trend recovery is detected to be higher than the set threshold (e.g. 0.75), the current shutdown process is immediately terminated and the "preheating standby" state is activated. In this state:

[0114] The main power supply of the equipment is not cut off;

[0115] Key electronic control links maintain low-frequency power supply;

[0116] The system maintains the "soft start ready" state and can re-enter full load operation within 2 seconds.

[0117] This mechanism significantly improves the system's adaptability to complex and alternating working conditions, reducing mechanical losses and energy waste caused by frequent starts and stops.

[0118] Through the technical implementation of the above-mentioned energy-saving control execution module, the present invention constructs a set of energy-saving execution strategies that are safe, flexible and reversible, effectively solving key problems such as "extensive control, high false stop rate, and delayed system response" existing in traditional coal mine equipment energy-saving control methods.

[0119] In the energy-saving control system for continuous mining and continuous charging equipment in coal mines, the prejudgment and intervention module, serving as the intermediary logic layer connecting intelligent idling detection and energy-saving control execution, performs three core functions: forward-looking judgment, policy interruption, and state transition. Its primary purpose is to intelligently predict the likelihood of future equipment feed resumption after the system initially determines it is idling and before a shutdown is executed, thereby avoiding false shutdowns caused by short-term misjudgment. Furthermore, the module uses temperature control strategies to guide the equipment to a smooth transition to a preheated standby state.

[0120] This module first constructs a shuttle trajectory model based on the Time-Series Embedding method. The specific process is as follows:

[0121] The system continuously collects shuttle displacement data D = {d1, d2, ..., d t}; t is the total length of the time step;

[0122] The original displacement sequence is normalized and denoised to construct an embedding matrix E, where each column contains a time window of fixed length;

[0123] A multi-scale one-dimensional convolutional neural network (Multi-Scale 1D-CNN) was used to extract the features of E and extract the shuttle speed variation characteristics (V) and acceleration trend curves (A) at different time scales.

[0124] When it is identified that the speed continuously increases from zero to a positive value and the acceleration trend is stable in a positive direction for more than a set time window (such as 3 seconds), it is regarded as a "shuttle car approaching" signal, which serves as the first reference for feeding resumption.

[0125] This trend model is significantly superior to the traditional "fixed-distance trigger" mechanism, with stronger foresight and anti-interference capabilities. It is particularly suitable for complex shuttle operation scenarios such as night shifts, long distances, and non-uniform movements.

[0126] When the equipment is idling, the current signals of the crusher and conveying equipment should remain low and steady. However, when the upstream shuttle car or coal storage tank begins to release coal and the equipment enters the pre-start state, the motor current may fluctuate slightly. The current jump recognition algorithm designed in this module specifically includes:

[0127] Construct the current sequence I = {i1, i2, ..., i t};

[0128] The "current micro-jump threshold" δI is introduced by comparing whether the current has continuously increased monotonically at multiple consecutive time points and the average increase is greater than δI (e.g., 2-5%).

[0129] If the current change curve shows the characteristics of the initial stage of the "S-shaped" starting curve (typically load fluctuations before the belt motor starts), it is judged as a sign of pre-start.

[0130] This feature has shown stability in multiple field tests and can be used as a second criterion for "possibility of equipment recovery," effectively compensating for the delay in shuttle trajectory perception.

[0131] Considering the distinct shift scheduling and equipment operation cycles of coal mine continuous mining systems, this module introduces a historical data modeling mechanism to construct a feeding cycle database and deploys a pattern matching neural network (Pattern Matching RNN) to identify high-incidence feeding sections. The implementation process is as follows:

[0132] Feeding history records are grouped by shift, and the frequency of shuttle feeding per unit time is counted to establish a "feeding thermal curve";

[0133] The current system operation timestamp is used as input and mapped to the historical cycle curve;

[0134] The model outputs the similarity score S (range 0~1) between the current period and the high-frequency band of feeding;

[0135] If S≥0.75, it is judged as "may enter the high-incidence period of feeding".

[0136] This mechanism significantly improves the system's prediction accuracy under periodic operating conditions, and is particularly important for night shifts.

[0137] The prediction and intervention module uses the above three types of trend information as independent prediction signal sources and sets the fusion judgment logic as follows:

[0138] If any two of the three trend judgment results meet the positive recovery conditions, it will be regarded as a high-confidence signal that "feed will be restored soon";

[0139] The system immediately suspends the original energy-saving shutdown instruction transmission, and starts the "preheating standby control flow", which includes:

[0140] The core power unit maintains low-frequency power-on;

[0141] The conveying motor maintains low-speed slip operation;

[0142] The control support unit maintains full-function active state;

[0143] The temperature prediction model predicts the preparation time (T_ready) required to enter full speed according to the current motor temperature rise state, and serves as a reference for subsequent switching rhythm;

[0144] This process ensures that the device is in a stable preheating state before resuming operation, ensuring fast response and reducing energy consumption peaks.

[0145] The pre-judgment intervention module of the present application deeply integrates three types of non-traditional means, namely time series trajectory modeling, current micro-jump identification and periodic pattern matching, to build a forward-looking, multi-dimensional and high-credibility feed recovery prediction mechanism.

[0146] Compared with the existing energy-saving control system which relies on fixed threshold or single sensor signal judgment, this module can effectively avoid the risk of misjudgment and misstop, realize intelligent switching of idle running judgment logic to device recovery control logic, and significantly improve the stability, sensitivity and flexibility level of system energy-saving control.

[0147] Example 2: In order to verify the effectiveness of the coal mine continuous mining and continuous charging coal mining equipment energy-saving control system proposed by the present application, the applicant constructed an experimental platform including a complete "data acquisition-AI analysis-idle running discrimination-energy-saving control-prediction intervention" closed-loop structure, and compared it with the traditional idle running control logic under the actual mine simulation conditions. The results are as follows.

[0148] This experiment was carried out in the digital twin control simulation system of a certain coal mining enterprise, and the specific configuration is as follows:

[0149] Sampling period: 5 seconds / frame;

[0150] Shuttle car trajectory simulator: true running trajectory and feed cycle reproduction;

[0151] Current sensor sampler: three-phase motor, sampling accuracy ±0.5A;

[0152] Control object: 1 crusher (power 110kW), 2 conveyors (power 55kW);

[0153] Test period: 3 consecutive work shifts (total 36 hours);

[0154] Comparison system: traditional timed delayed shutdown control (60-second idling shutdown) vs. the system of the present invention.

[0155] Parameter settings:

[0156] Idle severity level threshold θ: 0.7;

[0157] Confidence score trigger threshold β_min: 0.8;

[0158] Current jump threshold ΔI: +3% for more than 3 seconds; Equipment preheating slip frequency: 15Hz;

[0159] Shuttle trajectory model CNN convolution kernel size: [3, 5, 7];

[0160] Feeding cycle matching similarity S threshold: ≥0.75.

[0161] Experimental indicator definition:

[0162] Idle recognition accuracy rate (AAR): the rate at which the model correctly identifies the idling state;

[0163] False Stop Rate (FER): The percentage of times the system stops due to misjudgment but is actually in the feeding state;

[0164] Response delay (RTD): The average time difference between the restoration of feed and the resumption of equipment operation;

[0165] Energy Saving Rate (SER): The ratio of electricity saved in a unit operating cycle to total electricity consumption.

[0166] Comparison of experimental results:

[0167] Indicator name Traditional control systems System of the present invention Relative improvement AAR(%) 81.4 96.7 ↑+18.8% FER(%) 12.5 2.1 ↓−83.2% RTD (seconds) 9.7 3.2 ↓−66.9% SER(%) 6.8 15.4 ↑+126.5%

[0168] Analysis and explanation: This invention significantly improves the accuracy of idling judgment (by nearly 20 percentage points) through multi-parameter fusion and random forest recognition model;

[0169] Traditional control methods often experience false stops, especially during night shifts, when extended feeding intervals can lead to misjudgments. This system uses shuttle trend modeling and current jump identification to pre-start feeding, reducing the false stop rate by over 83%.

[0170] Due to the introduction of the preheating standby strategy, the system feeding recovery response delay is reduced from the traditional average of nearly 10 seconds to less than 3 seconds, avoiding startup energy consumption and mechanical shock;

[0171] The energy-saving control execution module accurately executes the hierarchical control strategy through gradient unloading and priority mapping mechanism, and the overall energy saving rate is increased to more than 15%.

[0172] By deploying the system in the actual mine working condition and systematically comparing with the traditional scheme, the coal mine continuous mining and continuous filling coal mining equipment energy-saving control system has not only remarkable identification accuracy and energy-saving efficiency, but also good response robustness and scene adaptability, which fully proves its application potential and popularization value in the intelligent energy-saving management of coal mines.

[0173] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. An energy-saving control system for continuous mining and continuous filling coal mining equipment in a coal mine, characterized by: include: A multi-parameter acquisition module is used to collect multi-source operating status data related to the continuous mining and charging system in coal mines, including the distance between the shuttle car and the equipment, the current load of the crusher and conveyor, operating vibration values, material flow detection signals, and the current operating shift type; An artificial intelligence analysis module, based on a pre-trained random forest model, performs feature extraction and time series modeling on the operating status data, and outputs a predicted probability value and a credibility score for whether the coal mining equipment is in an idling state within the current time window; The intelligent idling judgment module uses a confidence-weighted judgment mechanism to determine whether the vehicle has entered the idling state based on the idling prediction results in multiple consecutive time windows. If the set judgment conditions are met, an energy-saving control instruction is generated; An energy-saving control execution module executes a hierarchical control strategy after receiving an energy-saving control instruction; The predictive intervention module is used to determine whether the system is about to resume feeding based on the shuttle trajectory trend, current fluctuation rebound, and historical feeding cycle information. If it is determined that it is about to resume, the current shutdown logic is interrupted or switched to preheating standby mode.

2. The energy-saving control system for continuous mining and continuous filling coal mining equipment according to claim 1, characterized in that: The multi-parameter acquisition module includes: Collect the operating current signals of the crusher and conveyor, and extract the load fluctuation feature vector through short-time Fourier transform to characterize the typical current pattern of the equipment when it is unloaded; Real-time measurement of the dynamic distance change between the shuttle car and the crusher, and filtering of environmental interference through a differential sliding window algorithm to achieve continuous identification of the shuttle car's dwell time; Obtain the equipment's operating vibration value and identify the equipment's micro-vibration trend when it is not operating by clustering its frequency domain envelope energy. A temporal association rule model is constructed by integrating the collection timestamp and mining operation scheduling data to identify the operation shift types and analyze the periodic structure.

3. The energy-saving control system for continuous mining and continuous filling coal mining equipment according to claim 1, characterized in that: Artificial intelligence analysis modules include: Based on the collected multi-source operating status data, a multidimensional feature vector set X is constructed, which consists of current fluctuation amplitude, shuttle car parking time, vibration power spectrum density, material flow interruption period and operation shift label. Using a random forest classification model that has been trained offline, the feature vector set X is input into the model structure. Each decision tree outputs a preliminary classification result for the idling state. The model as a whole aggregates the results through a majority voting mechanism and generates an idling prediction label. In response to the fluctuations in the output of each decision tree, the confidence score β of the predicted label is calculated and dynamically adjusted based on the leaf node information gain entropy value and the weighted distribution of feature importance; The predicted label and confidence score are combined to form an output vector, which serves as the decision basis for idling determination.

4. The energy-saving control system for continuous mining and continuous filling coal mining equipment according to claim 1, characterized in that: The intelligent idling detection module includes: Construct a state prediction window sequence W = {Y1, Y2, ..., Y n }, each Yᵢ represents the idling prediction result output by the artificial intelligence analysis module in the i-th time slice and its corresponding confidence βᵢ, and n is the total number of time slices; A dynamic weight function ωᵢ = f(βᵢ, Δtᵢ) is set, where Δtᵢ is the delay of the time slice relative to the current time. This function is used to give higher weight to recent prediction results, thus implementing a credibility weighting mechanism that enhances temporal priority. Aggregate all weighted results in the window sequence to obtain the total confidence score γ = Σ(ωᵢ·Yᵢ) and compare it with the adaptive threshold θ. If γ ≥ θ, it is determined to be in a stable idling state. The energy-saving control instruction generation module is triggered and the status is passed to the control execution unit, so that the equipment energy-saving shutdown is executed only when both idling continuity and confidence stability are met.

5. The energy-saving control system for continuous mining and continuous filling coal mining equipment according to claim 4, characterized in that: The intelligent idling determination module further includes: Construct a set of lightweight time series residual prediction models that run in parallel, use a one-dimensional convolutional network to model the state change trend in the prediction window sequence W, and output the sequence trend prediction value and its relationship with reality The residual ε between If the residual ε in two or more consecutive time slices exceeds the preset fluctuation tolerance threshold ε t , judge that there is nonlinear mutation or short-term interference in the current prediction state, suspend the current idling judgment process and enter the self-stabilization observation period; During the observation period, an improved exponentially weighted smoothing algorithm is used to perform time series correction on the γ value to reduce the impact of short-term sharp fluctuations on the judgment results.

6. The energy-saving control system for continuous mining and continuous filling coal mining equipment according to claim 1, characterized in that: The energy-saving control execution module includes: Analyze the priority tags and confidence level parameters in the energy-saving control instructions, and divide the control objects into core power units, auxiliary feeding units, and control support units according to their functional links; Build a hierarchical execution strategy based on priority tags, and use a deeply configurable state mapping table to map idle severity levels to grouped device states; During the execution process, a dynamic unloading curve function is introduced to adjust the unloading time difference of each unit according to the current load level and the energy consumption rate reduction model; A state holding lock mechanism is set. If the upstream feeding trend recovery signal is detected during the unloading process, the shutdown process will be immediately terminated and switched to the preheating standby state.

7. The energy-saving control system for continuous mining and continuous filling coal mining equipment in a coal mine according to claim 1, characterized in that: The predictive intervention module includes: A shuttle trajectory trend model based on time series embedding representation is constructed. By performing multi-scale convolution feature extraction on the shuttle displacement data, the velocity vector and acceleration trend curve of the current trajectory change are generated. Combined with the real-time current fluctuation sequence of the conveying equipment, a jump recognition algorithm is used to detect whether there is a typical rebound feature, that is, a slight increase in current exceeding the set threshold in the idling state, prompting the mechanical component to pre-start action; A historical feeding cycle database is introduced, and a pattern matching neural network is used to analyze whether the current period coincides with the high-frequency section of the typical feeding cycle, so as to determine whether it has entered a high-incidence feeding period; If any two of the three trend information meet the judgment conditions for resuming feeding, the current energy-saving shutdown process will be suspended, and the equipment preheating standby strategy will be started according to the system temperature rise model to realize the switching from idling control to recovery control.

Images