A mine mining and selecting whole cycle energy consumption intelligent optimization and regulation method and system

Abstract

The application provides a mine mining and dressing whole cycle energy consumption intelligent optimization and regulation method and system, belongs to the mine mining and dressing energy consumption optimization and regulation technical field, and the method comprises the following steps: obtaining mine mining and dressing each link parameter, calculating each link energy consumption and total energy consumption cost by using a corresponding model, taking the minimization of the total energy consumption cost of mining and dressing and the maximization of the total processing capacity of mining and dressing as the target, and taking the grinding product fineness in the quality constraint interval and the equipment operation parameter in the safety interval as the constraint condition, performing collaborative optimization and solving by using a multi-objective optimization algorithm to obtain an optimal parameter combination; executing the optimal parameter combination, collecting actual operation data and actual energy consumption data; comparing the actual energy consumption data with the predicted corresponding power consumption and consumption to obtain a comprehensive deviation value; when the comprehensive deviation value is greater than a preset deviation threshold value, adjusting the parameters of the crushing size-energy consumption model and / or the grinding energy consumption prediction model by using an optimization algorithm to obtain optimized model parameters and update the corresponding model.

Description

Technical Field

[0001] This application relates to the field of energy consumption optimization and control technology in mining and beneficiation, and in particular to a method and system for intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle. Background Technology

[0002] In existing mining and beneficiation processes, for the blasting stage, technicians determine blasting design parameters based on experience, lacking precise predictive models and failing to accurately estimate post-blast block size distribution and energy consumption. In the crushing stage, operating parameters are set based on the crusher's rated power and experience, without considering the impact of ore block size distribution on energy consumption. For the grinding stage, operation follows standard grinding mill operating procedures without optimization based on the relationship between ore characteristics and energy consumption. The interrelationships between stages throughout the entire mining and beneficiation cycle are ignored, and there is a lack of coordinated optimization of parameters across stages. Due to the lack of predictive models and coordinated optimization mechanisms, the energy consumption of each stage cannot be accurately estimated, resulting in persistently high total energy costs in mining and beneficiation. The inability to adjust parameters based on actual production conditions prevents optimal production efficiency and hinders the ability to maximize total throughput while reducing energy consumption.

[0003] Therefore, there is an urgent need for a method and system for intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a method and system for intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle.

[0005] A first aspect of this application provides a method for intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle, comprising: Obtain blasting design parameters and geological parameters in the mining and beneficiation blasting process, input them into the constructed blasting block size prediction model, and obtain the post-blast block size distribution and blasting energy consumption; The post-explosion block size distribution is input into a preset crushed block size-energy consumption model, and the equipment operating parameters and working condition parameters of the crusher are obtained to calculate the power consumption of the crushing process. The operating parameters of the grinding mill and the characteristics of the ore are obtained. Based on the preset grinding energy consumption prediction model, the power consumption and steel ball consumption in the grinding process are obtained. Based on the power consumption of the blasting, crushing, and grinding stages, as well as the steel ball consumption, the total energy consumption cost of mining and beneficiation is obtained. A comprehensive objective function is constructed with the goal of minimizing the total energy consumption cost of mining and beneficiation and maximizing the total processing capacity. The constraints are that the fineness of the grinding product is within the quality constraint range and the equipment operating parameters are within the safe range. A multi-objective optimization algorithm is used to collaboratively optimize the blasting design parameters, the crusher's equipment operating parameters, and the grinding mill's operating parameters to obtain the optimal parameter combination. Execute the optimal parameter combination and collect actual operating data and actual energy consumption data; The actual energy consumption data is compared with the predicted corresponding power consumption and consumption to obtain a comprehensive deviation value; When the overall deviation value is greater than the preset deviation threshold, the parameters of the crushed block size-energy consumption model and / or the grinding energy consumption prediction model are adjusted using an optimization algorithm to obtain the optimized model parameters and update the corresponding model.

[0006] A second aspect of this application provides an intelligent optimization and control system for energy consumption throughout the entire mining and beneficiation cycle, comprising: The blasting module is used to obtain blasting design parameters and geological parameters in the mining and beneficiation blasting process. By inputting the constructed blasting block size prediction model, the module can obtain the post-blast block size distribution and blasting energy consumption. The crushing module is used to input the post-explosion block size distribution into a preset crushed block size-energy consumption model, and to obtain the equipment operating parameters and working condition parameters of the crusher, and calculate the power consumption of the crushing process. The grinding module is used to obtain the operating parameters of the grinding mill and the characteristics of the ore. Based on the preset grinding energy consumption prediction model, it obtains the power consumption and steel ball consumption of the grinding process. The collaborative optimization module is used to obtain the total energy consumption cost of mining and beneficiation based on the power consumption of the blasting, crushing, and grinding stages, as well as the steel ball consumption. It constructs a comprehensive objective function with the goal of minimizing the total energy consumption cost of mining and beneficiation and maximizing the total processing capacity. The constraints are that the fineness of the grinding product is within the quality constraint range and the equipment operating parameters are within the safe range. The module uses a multi-objective optimization algorithm to collaboratively optimize the blasting design parameters, the equipment operating parameters of the crusher, and the operating parameters of the grinding mill to obtain the optimal parameter combination. The scheme execution module is used to execute the optimal parameter combination and collect actual operating data and actual energy consumption data; The energy efficiency deviation module is used to compare the actual energy consumption data with the predicted corresponding power consumption and consumption to obtain a comprehensive deviation value. The adjustment and update module is used to adjust the parameters of the crushed block size-energy consumption model and / or grinding energy consumption prediction model using an optimization algorithm when the comprehensive deviation value is greater than a preset deviation threshold, so as to obtain the optimized model parameters and update the corresponding model.

[0007] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle.

[0008] In a fourth aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle.

[0009] The beneficial effects of the intelligent optimization and control method and system for energy consumption throughout the entire mining and beneficiation cycle provided in this application are as follows: This application obtains the energy consumption of each stage of mining and beneficiation, constructs a comprehensive objective function, and performs collaborative optimization to obtain the optimal parameter combination to reduce the total energy consumption cost of mining and beneficiation and increase the total processing capacity of mining and beneficiation. At the same time, by comparing the actual and predicted energy consumption to obtain the comprehensive deviation value, the model parameters are adjusted and updated to ensure the accuracy of energy consumption prediction and the effectiveness of optimization and control. Attached Figure Description

[0010] Figure 1 A flowchart illustrating an intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle, provided in an embodiment of this application; Figure 2 This is a structural block diagram of an intelligent optimization and control system for energy consumption throughout the entire mining and beneficiation cycle provided in an embodiment of this application; Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0012] To make the purpose, technical solution, and advantages of this application clearer, the following will be described in conjunction with the appendix. Figure 1-3 The following is an explanation using specific examples.

[0013] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle, provided in an embodiment of this application. The method includes: S101: Obtain blasting design parameters and geological parameters in the mining and beneficiation blasting process, input the constructed blasting block size prediction model, and obtain the post-blast block size distribution and blasting energy consumption.

[0014] In this embodiment, blasting design parameters are the process parameters planned before blasting operations in the mine, which directly determine the blasting effect and energy consumption. These include borehole parameters, such as borehole depth, hole spacing, and row spacing; charge parameters, such as explosive consumption per unit volume, charge structure, and detonation sequence; and detonation parameters, such as delay time and detonation method. These are the input variables for the blasting block size prediction model. Geological parameters are the ore and rock mass characteristics of the blasting area, affecting the transfer of blasting energy and the ore crushing effect. These include ore hardness, rock mass joint development, ore density, and rock mass integrity coefficient, and are influencing factors in predicting the post-blast block size distribution. The blasting block size prediction model is a mathematical model trained based on historical blasting data. In this embodiment, a BP neural network is used. The inputs are the blasting design parameters and geological parameters, and the outputs are the post-blast ore particle size distribution ratio and the energy consumption cost of the blasting process. This is the basic model for the energy consumption correlation of the entire blasting-crushing-grinding process.

[0015] In this embodiment, the blasting block size prediction model adopts a BP neural network with a 3-layer fully connected structure. The specific layer division and connection relationship are as follows: Input layer: A total of 12 neurons are set, corresponding to the input parameters of the mining and beneficiation blasting process, including 6 blasting design parameters: explosive consumption per unit volume, hole spacing, row spacing, charge length, detonation sequence, and packing length; and 6 geological parameters: ore uniaxial compressive strength, rock stratum dip angle, joint density, water content, burial depth, and rock mass integrity coefficient. The input layer has no activation function and is only responsible for receiving and transmitting raw parameter data; Hidden layer: One hidden layer is set, with the number of neurons determined to be 20 after grid search optimization. The ReLU function is selected as the activation function to solve the problem. To address the vanishing gradient problem, the hidden layer and input layer are fully connected via a weight matrix (12×20) and a bias vector (20×1) to achieve nonlinear feature extraction of the input parameters. The output layer consists of 4 neurons, corresponding to the 4 particle size distributions of the post-blast block size (less than 10mm, 10-50mm, 50-100mm, and greater than 100mm), plus 1 neuron outputting the energy consumption of the blasting process (unit: yuan / t), for a total of 5 output neurons. The sigmoid function is used as the activation function to constrain the output value to the 0-1 range, adapting to proportion-type data. The output layer and hidden layer are fully connected via a weight matrix (20×5) and a bias vector (5×1).

[0016] In this embodiment, the post-blast size distribution refers to the proportion of different particle sizes into which the ore is crushed after blasting. It is divided according to particle size ranges, such as d < 10mm, 10mm ≤ d < 50mm, 50mm ≤ d < 100mm, and d > 10mm, directly determining the processing difficulty and energy consumption of the crushing and grinding stages. The energy consumption of the blasting stage includes all energy consumption and material costs generated during the blasting operation, including the cost of explosives (explosive consumption per unit × ore processing volume × explosive unit price) and the energy consumption of blasting equipment (electricity / diesel consumption of drilling rigs, detonators, etc.), which is a component of the total energy cost throughout the mining and beneficiation cycle.

[0017] S102: Input the post-explosion block size distribution into the preset crushed block size-energy consumption model, and obtain the equipment operation parameters and working condition parameters of the crusher to calculate the power consumption of the crushing process.

[0018] In this embodiment, the crushing block size-energy consumption model is a mathematical model that describes the quantitative correlation between ore block size distribution and power consumption in the crushing process. This embodiment adopts a multiple linear regression model, which is constructed based on historical mine operation data (block size distribution, equipment parameters, power consumption data). The inputs are the post-blasting block size distribution, crusher operation and working condition parameters, and the output is the unit ore power consumption in the crushing process. It is a tool for accurate prediction of energy consumption in the crushing process.

[0019] In this embodiment, the input data of the crushing size-energy consumption model closely matches the actual production scenario of the mining crushing process. The proportion of the four particle size distributions is taken from the output results of the blasting size prediction model (or measured data from post-blasting ore image analysis), with values ​​ranging from 0 to 1. The sum of the proportions of each particle size is 1. The main shaft speed of the crusher is a real-time operating parameter collected by the on-site PLC system, with a value range of 500-800 r / min, which is compatible with the rated speed range of the existing jaw crusher in the mine. The output data of the model is the unit ore power consumption in the crushing process, with a value range of 0.8-1.5 kW・h / t, which is consistent with the physical dimensions of the actual power consumption collected by the on-site electricity meter. The scenario correlation logic between the input and output is clear: the higher the proportion of large-diameter ore, the greater the crushing load of the crusher, and the higher the power consumption; when the main shaft speed increases within the rated range, the crushing efficiency increases and the unit power consumption decreases. The sign of the linear coefficient of the model is completely matched with engineering cognition, making the model output results practically instructive and directly applicable to the calculation of total energy consumption cost and parameter optimization in mining and beneficiation.

[0020] In this embodiment, the crusher's operating parameters are process parameters that can be manually adjusted during operation, directly affecting crushing efficiency and energy consumption. These include: feeding speed (the amount of ore fed into the crusher per unit time); crushing chamber clearance (the width of the discharge port, which determines the particle size of the crushed ore); motor rated power; and rotor speed. The crusher's operating condition parameters are parameters representing its operating status, collected by equipment sensors. These cannot be directly adjusted but affect energy consumption. These include: equipment load rate (the ratio of actual load to rated load); bearing temperature; material blockage status; and equipment maintenance coefficient, characterizing the degree of equipment aging and wear. The power consumption in the crushing process is the electrical energy consumed by the crusher motor during the crushing operation, measured per unit of ore. It is a component of the energy cost of the crushing process, and its value is directly related to the size of the blasted ore, equipment operation, and operating condition parameters.

[0021] S103: Obtain the operating parameters of the grinding mill and the characteristics of the ore, and based on the preset grinding energy consumption prediction model, obtain the power consumption and steel ball consumption of the grinding process.

[0022] In this embodiment, the operating parameters of the grinding mill are process parameters that can be manually adjusted during grinding operations, directly affecting grinding efficiency, product fineness, and energy consumption. These include mill rotation speed, such as the cylinder rotation speed, which determines the intensity of steel ball drop; grinding concentration, such as the mass ratio of solid materials in the slurry; filling rate, such as the proportion of steel ball volume to the effective mill volume; and feed rate, such as the amount of ore fed into the mill per unit time. Ore characteristics are the inherent physicochemical properties of the ore entering the grinding stage, and are factors that determine the grinding difficulty. These include ore hardness, feed particle size (the distribution of ore size after crushing before entering the mill), ore density, and ore grindability coefficient (characterizing the ease with which the ore is ground). The grinding energy consumption prediction model is a mathematical model constructed based on historical grinding data from the mine. In this embodiment, the model used is a support vector machine. The inputs are the grinding mill operating parameters and ore characteristics, and the outputs are the unit ore power consumption and steel ball consumption in the grinding stage. It is a tool for calculating grinding costs.

[0023] In this embodiment, the grinding energy consumption prediction model adopts a standard support vector machine regression (SVR) hierarchical structure, consisting of three layers: an input layer, a kernel function mapping layer, and an output layer, with no hidden layers. The input layer includes eight neurons, corresponding to grinding mill operating parameters (mill speed, media filling rate, grinding concentration, feed rate) and ore characteristic parameters (ore grindability, ore density, average particle size of ore entering the mill, ore moisture content), responsible for receiving and transmitting standardized feature data. The kernel function mapping layer is the core of the model, using radial basis function (RBF) as the kernel function to map the 8-dimensional linearly inseparable features of the input layer to a high-dimensional feature space, achieving linear separability. The core parameter of this layer is the kernel function width (γ). The output layer contains two neurons, corresponding to the unit ore power consumption (kW·h / t) and unit ore steel ball consumption (kg / t) in the grinding process, respectively. The optimal hyperplane is solved in the high-dimensional space through linear regression, and the prediction result is output. The model's correlation parameters are the penalty factor (C), kernel width (γ), and insensitive loss coefficient (ε), which together determine the model's fitting accuracy and generalization ability.

[0024] In this embodiment, the power consumption in the grinding process refers to the electrical energy consumed by the mill motor during operation, measured using power consumption per unit of ore. Its value is directly and positively correlated with mill speed, filling rate, and ore hardness. Steel ball consumption refers to the wear and tear of steel balls during the grinding process due to impact and grinding of the ore. It is measured using the amount of steel balls consumed per unit of ore and is related to ore hardness, feed particle size, and steel ball material; it represents the consumable cost of the grinding process.

[0025] S104: Based on the power consumption of the blasting, crushing, and grinding stages, as well as the consumption of steel balls, the total energy consumption cost of mining and beneficiation is obtained. A comprehensive objective function is constructed with the goal of minimizing the total energy consumption cost of mining and beneficiation and maximizing the total processing capacity of mining and beneficiation. The constraints are that the fineness of the grinding product is within the quality constraint range and the equipment operating parameters are within the safe range. A multi-objective optimization algorithm is used to collaboratively optimize the blasting design parameters, the equipment operating parameters of the crusher, and the operating parameters of the grinding mill to obtain the optimal parameter combination.

[0026] In this embodiment, the total energy consumption cost of mining and beneficiation is the sum of energy consumption and material costs in the three major stages of blasting, crushing, and grinding in the entire mining and beneficiation process. It includes: electricity consumption and explosive costs in the blasting stage, electricity consumption costs in the crushing stage, and electricity consumption costs and steel ball consumption costs in the grinding stage. It is an indicator of the economic efficiency of mining and beneficiation, expressed in yuan / t, i.e., the energy cost per ton of ore. The comprehensive objective function is a mathematical function constructed to minimize the total energy consumption cost of mining and beneficiation and maximize the total processing capacity. Through weighting and normalization, the two conflicting objectives are transformed into a solvable unified objective, simultaneously considering both mining and beneficiation economics and production efficiency, serving as a guide for multi-objective optimization. The quality constraint interval is the performance acceptable range set for the grinding product, representing the fineness of the grinding product. If the fineness exceeds this interval, for example, insufficient fineness, it will reduce the efficiency of subsequent processes such as flotation and leaching, and is a hard constraint that must be met. The safety constraint range is the operational safety boundary set for each stage of equipment, that is, the allowable range of values ​​for equipment operating parameters. For example, the load rate of the crusher motor is less than or equal to 85%, the bearing temperature of the grinding mill is less than or equal to 70℃, and the single consumption of blasting explosives is less than or equal to 0.35kg / t. The purpose is to avoid safety risks such as equipment overload, overheating, and explosion accidents, and to ensure stable production.

[0027] In this embodiment, the multi-objective optimization algorithm is a type of heuristic algorithm used to solve multi-objective conflict problems. This embodiment adopts the multi-objective particle swarm optimization algorithm (MOPSO), which can find the optimal solution set that takes into account multiple objectives while satisfying all constraints. Finally, the parameter combination with the best practicality is selected based on production priority.

[0028] In this embodiment, the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm adopts a three-level iterative optimization structure, consisting of a population initialization layer, a particle iterative optimization layer, and a Pareto optimal solution set selection layer. The population initialization layer generates a particle population containing parameters to be optimized. The particle dimensions correspond to 12 optimization variables (blasting design parameters: explosive consumption, hole spacing, row spacing, charge length, detonation sequence, and packing length; crusher operating parameters: spindle speed, feed rate, and discharge port gap; grinding mill operating parameters: mill speed, media filling rate, and grinding concentration). The particle iterative optimization layer is the core of the algorithm, achieving optimization through two sub-modules: particle velocity update and position update. The velocity update module introduces three core parameters: inertia weight, individual learning factor, and global learning factor, controlling the particle's exploration and development capabilities. The position update module adjusts the particle's position in the solution space based on the velocity calculation results. The Pareto optimal solution set selection layer uses crowding calculation and non-dominated sorting to select the optimal particle set that simultaneously minimizes total energy consumption cost and maximizes total processing capacity, avoiding local optima. The algorithm achieves collaborative parameter optimization under multiple objectives and constraints through a closed-loop process of initialization-iteration-selection at each level.

[0029] In this embodiment, collaborative optimization overcomes the limitations of independent optimization of each stage of blasting, crushing, and grinding. Aiming for overall process optimization, it simultaneously optimizes the adjustable parameters of the three stages: blasting design parameters, crusher operating parameters, and grinding mill operating parameters. It considers the coupling relationships between stages; for example, blasting particle size affects crushing energy consumption, and crushing particle size affects grinding energy consumption, thus avoiding local optima leading to global inefficiency. The optimal parameter combination, obtained after solving the multi-objective optimization algorithm, is a set of parameters that satisfies all constraints and optimizes the comprehensive objective function. This set includes adjustable parameters for each stage of blasting, crushing, and grinding, and can be directly used to guide on-site production.

[0030] S105: Execute the optimal parameter combination and collect actual operating data and actual energy consumption data.

[0031] In this embodiment, executing the optimal parameter combination involves distributing the optimal parameters for blasting, crushing, and grinding processes, obtained through multi-objective collaborative optimization, to the corresponding production equipment via the industrial control system to guide actual on-site production operations. This ensures synchronized and precise adjustment of parameters at each stage, preventing deviations in parameters at a single stage from causing the overall optimization effect to fail. Actual operational data refers to the production operation status data collected by equipment sensors and the PLC control system after parameter execution. This includes equipment operating parameters (e.g., actual feed rate, actual mill speed), process parameters (e.g., actual grinding concentration, crushed ore particle size), and equipment operating condition parameters (e.g., actual motor load rate, bearing temperature), used to verify the accuracy of parameter execution. Actual energy consumption data refers to the actual energy and material consumption data at each stage during parameter execution, including: actual explosive consumption and actual power consumption of the equipment in the blasting stage; actual motor power consumption in the crushing stage; and actual motor power consumption and actual steel ball consumption in the grinding stage. This data needs to be collected from metering instruments (electricity meters, weighbridges, steel ball consumption statistics ledgers) and serves as the basis for evaluating the optimization effect.

[0032] S106: Compare the actual energy consumption data with the predicted corresponding power consumption and consumption to obtain the comprehensive deviation value.

[0033] In this embodiment, the actual energy consumption data refers to the real energy and material consumption data collected by metering instruments and sensors at each stage after executing the optimal parameter combination. This includes the actual explosive cost and equipment power consumption in the blasting stage; the actual unit power consumption in the crushing stage; and the actual unit power consumption and steel ball consumption in the grinding stage. This data serves as the benchmark for evaluating the accuracy of the model's predictions. The predicted power consumption and consumption figures are calculated using the blasting block size prediction model, the crushing block size-energy consumption model, and the grinding energy consumption prediction model. These predicted values ​​correspond one-to-one with the actual energy consumption data and serve as a comparative reference for deviation calculations.

[0034] In this embodiment, the comprehensive deviation value is a comprehensive index obtained by weighting and summing the relative deviation rates of each stage based on the weight ratio of energy consumption in the total energy consumption cost of mining and beneficiation. The calculation formula is: Comprehensive deviation value = (|actual value - predicted value| / predicted value) × 100%. It is used to comprehensively evaluate the overall prediction accuracy of the blasting-crushing-grinding whole process model and is the basis for determining whether to trigger model parameter adjustment.

[0035] S107: When the overall deviation value is greater than the preset deviation threshold, the parameters of the crushing block size-energy consumption model and / or the grinding energy consumption prediction model are adjusted using the optimization algorithm to obtain the optimized model parameters and update the corresponding model.

[0036] In this embodiment, the preset deviation threshold is a pre-set critical value used to determine whether the model's prediction accuracy meets the standard. It is determined by historical mine model operation data and production accuracy requirements, and ranges from 5% to 10%. When the overall deviation value exceeds the preset deviation threshold, it indicates that the model's prediction results deviate too much from actual production, failing to accurately guide energy consumption optimization, and the model parameter adjustment process must be initiated. The optimization algorithm is a heuristic optimization algorithm used to correct model parameters. This embodiment employs the differential evolution algorithm, which iteratively searches within the feasible region of the model parameters to find the optimal parameter combination that minimizes the deviation between the model's predicted and actual values, adapting to nonlinear, multi-parameter model optimization scenarios.

[0037] In this embodiment, the differential evolution algorithm used for model parameter adjustment adopts a four-layer iterative optimization structure, consisting of a population initialization layer, a mutation operation layer, a crossover operation layer, and a selection operation layer. The population initialization layer generates a particle population containing the parameters to be optimized. The particle dimensions correspond to the five linear coefficients and one bias term of the fragmentation-energy consumption model (multiple linear regression), and the penalty factor, kernel width, and insensitive loss coefficient of the grinding energy consumption prediction model (support vector machine), totaling 10 dimensions of parameters to be optimized. The mutation operation layer generates mutated particles by weighting the parent particles through differential expression and a scaling factor, with the scaling factor controlling the mutation amplitude. The crossover operation layer controls the gene exchange ratio between mutated particles and parent particles through crossover probability, generating experimental particles. The selection operation layer selects and retains the optimal particles based on the objective function value, allowing the algorithm to iterate towards the direction with the minimum model prediction bias. Each layer of the algorithm achieves collaborative optimization of the parameters of the two types of energy consumption models through a closed-loop process of initialization-mutation-crossover-selection.

[0038] In this embodiment, the optimized model parameters are a set of parameters obtained through iterative search using an optimization algorithm, which maximizes the model's prediction accuracy. This parameter set reduces the model's prediction error regarding energy consumption in the crushing and grinding stages. Updating the corresponding model involves replacing the old parameters of the original model with the optimized parameter set, generating a new version of the crushing size-energy consumption model or grinding energy consumption prediction model, and deploying it to the mine energy consumption optimization system for energy consumption prediction in the next production cycle, thus achieving closed-loop iterative optimization of the model.

[0039] As can be seen from the above, this application obtains the energy consumption of each stage of mining and beneficiation, constructs a comprehensive objective function, and performs collaborative optimization to obtain the optimal parameter combination to reduce the total energy consumption cost of mining and beneficiation and increase the total processing capacity. At the same time, by comparing the actual and predicted energy consumption to obtain the comprehensive deviation value, the model parameters are adjusted and updated to ensure the accuracy of energy consumption prediction and the effectiveness of optimization and control.

[0040] In one embodiment of this application, the total energy consumption cost of mining and beneficiation is obtained based on the power consumption of the blasting stage, the power consumption of the crushing stage, the power consumption of the grinding stage, and the consumption of steel balls, including: Obtain cost calculation benchmark data, which includes time-of-use electricity price, diesel unit price, steel ball unit price, and the comprehensive efficiency coefficient of equipment in each stage; The electricity consumption of the blasting, crushing and grinding processes is converted into electricity costs based on the electricity price at the specific time of occurrence. The steel ball consumption in the grinding process is converted into steel ball loss cost based on the unit price of the steel ball; Identify and acquire the non-electric direct energy costs related to blasting, crushing, and grinding processes; including the cost of explosives used in blasting and the fuel consumption costs of large diesel equipment. Based on the comprehensive efficiency coefficient of each stage of equipment, the cost of electricity and the cost of non-electric direct energy consumption are corrected to obtain the corrected cost of electricity; the comprehensive efficiency coefficient is used to characterize the energy conversion efficiency loss caused by equipment aging and maintenance. The total energy cost of mining and beneficiation is obtained by summing the corrected electricity cost, steel ball loss cost, and non-electric direct energy cost.

[0041] In this embodiment, the cost calculation benchmark data consists of essential pricing and efficiency parameters necessary for calculating the energy consumption costs of each stage of mining and beneficiation. These parameters include time-of-use electricity prices (electricity unit prices at different times, such as peak, flat, and off-peak prices), diesel unit prices (fuel purchase prices for large mining equipment), steel ball unit prices (purchase prices for steel balls used in grinding), and the equipment comprehensive efficiency coefficient, which represents the energy conversion efficiency loss caused by equipment aging and maintenance. The coefficient is less than or equal to 1, with values ​​closer to 1 indicating better equipment condition. Time-of-use electricity prices are differentiated pricing policies formulated by the power sector based on changes in electricity load. They are divided into peak electricity prices (highest prices during peak electricity consumption periods), flat electricity prices (medium prices during stable load periods), and off-peak electricity prices (lowest prices during off-peak periods), used to calculate the actual cost of equipment electricity consumption at different times.

[0042] In this embodiment, the overall equipment efficiency coefficient is an indicator representing the actual operating efficiency of the equipment. It is calculated by weighting the equipment's load rate, maintenance coefficient, and aging coefficient, and is used to correct the deviation between theoretical and actual energy consumption costs. The more severe the equipment aging and the worse the maintenance, the smaller the coefficient, and the higher the corrected cost. Non-electric direct energy costs are the energy and material costs directly generated in each stage besides electricity consumption, including the cost of blasting explosives (the core material cost of the blasting stage) and the fuel consumption cost of diesel equipment (e.g., the diesel consumption cost of blasting drills and crushing and transfer equipment). These are an important component of the total energy consumption cost of mining and beneficiation. The corrected electricity cost is the actual electricity cost obtained by multiplying the theoretical electricity cost of each stage by the overall efficiency coefficient of the corresponding equipment. The formula is: Corrected electricity cost = Theoretical electricity cost / Overall efficiency coefficient. This eliminates the impact of equipment efficiency on cost accounting, making the cost data more accurate and reliable.

[0043] As can be seen from the above, this embodiment comprehensively considers benchmark data such as time-of-use electricity prices, diesel unit prices, steel ball unit prices, and equipment comprehensive efficiency coefficients, converts the electricity consumption of each link into electricity costs, converts the steel ball consumption in the grinding link into steel ball loss costs, identifies and obtains non-electricity direct energy costs, corrects the electricity costs, and sums them to obtain the total energy consumption cost of mining and beneficiation. This makes the calculation of the total energy consumption cost of mining and beneficiation more accurate and comprehensive, and can reflect the impact of equipment aging and maintenance status on energy consumption costs.

[0044] In one embodiment of this application, the optimization algorithm is a differential evolution algorithm; the optimization algorithm is used to adjust the parameters of the blasting block size prediction model, the crushed block size-energy consumption coupling model, and / or the grinding energy consumption prediction model to obtain optimized model parameters, including: The solution space of the differential evolution algorithm is determined based on the dimension and physical feasible range of the adjustable parameter subset of the model to be adjusted. Based on the combined deviation between actual energy consumption data and the predicted corresponding power consumption and energy expenditure, the hyperparameter combination of the differential evolution algorithm is determined.

[0045] In this embodiment, the differential evolution algorithm is a population-based heuristic optimization algorithm that iteratively searches for the optimal solution through three operations: mutation, crossover, and selection. It is suitable for multi-dimensional, nonlinear model parameter optimization scenarios, featuring fast convergence and strong robustness, and serves as the tool for adjusting prediction model parameters in this application. The solution space is the feasible range of values ​​for the parameters of the model to be optimized, determined by the physical meaning of the parameters, historical fitting intervals, and equipment operating boundaries. For example, the solution space for the fitting coefficients of multiple linear regression is [-1, 1], and the solution space for the weights of a neural network is [-0.5, 0.5]. The solution space is the search boundary of the differential evolution algorithm, ensuring that the optimized parameters are engineering-feasible.

[0046] In this embodiment, the hyperparameter combination refers to the control parameters of the differential evolution algorithm itself. These parameters do not participate in the direct calculation of the model, but they determine the algorithm's search efficiency and optimization accuracy. They include the scaling factor (which controls the magnitude of mutation operations) and the crossover probability (which controls the probability of crossover operations). The hyperparameter combination needs to be adjusted based on the comprehensive deviation value to adapt to different optimization scenarios. The comprehensive deviation value is the weighted deviation rate between the actual energy consumption of each stage and the model's predicted energy consumption. It is an indicator for determining whether the model's accuracy meets the standard and a direct basis for determining the hyperparameter combination. The larger the deviation value, the more the algorithm needs to strengthen its global search capability; the smaller the deviation value, the more the algorithm needs to strengthen its local refinement capability.

[0047] As can be seen from the above, this embodiment uses the differential evolution algorithm to adjust the model parameters, determines the solution space based on the dimension and physical feasible range of the adjustable parameter subset of the model to be adjusted, and determines the hyperparameter combination based on the comprehensive deviation between the actual energy consumption data and the predicted value. This can improve the accuracy and effectiveness of parameter adjustment for the blasting block size prediction model, the crushed block size-energy consumption coupling model, and / or the grinding energy consumption prediction model, thereby optimizing the model to more accurately represent the actual energy consumption.

[0048] In one embodiment of this application, the hyperparameters include a scaling factor and a crossover probability; Based on the comprehensive deviation between actual energy consumption data and the predicted corresponding power consumption and energy expenditure, the hyperparameter combination of the differential evolution algorithm is determined, including: If the overall deviation value is greater than or equal to the preset first-level deviation threshold, the reference value of the scaling factor is increased based on the first preset step size, and the reference value of the crossover probability is decreased based on the second preset step size. If the overall deviation value is less than the preset second-level deviation threshold, the reference value of the scaling factor is reduced based on the third preset step size, and the reference value of the crossover probability is increased based on the fourth preset step size.

[0049] In this embodiment, the scaling factor is one of the hyperparameters of the differential evolution algorithm, used to control the magnitude of the mutation operation, with a value range of [0,2]. A larger scaling factor results in a larger mutation magnitude, stronger global search capability of the differential evolution algorithm, and less susceptibility to local optima; a smaller scaling factor results in a smaller mutation magnitude, making the differential evolution algorithm more inclined towards local fine-grained search and faster convergence. The crossover probability is another hyperparameter of the differential evolution algorithm, used to control the probability of the crossover operation, with a value range of [0,1]. A higher crossover probability results in a lower proportion of offspring inheriting genes from their parents, leading to stronger population diversity; a lower crossover probability results in offspring being closer to their parents, making the algorithm converge more stably, but more prone to getting trapped in local optima.

[0050] In this embodiment, the first-level deviation threshold is a preset high-deviation critical value. When the overall deviation value is greater than or equal to the first-level deviation threshold, it indicates that the model's predicted value deviates significantly from the actual value, requiring an enhancement of the algorithm's global search capability. This can be achieved by increasing the scaling factor and decreasing the crossover probability to expand the search range and find better model parameters. The second-level deviation threshold is a preset low-deviation critical value, lower than the first-level deviation threshold. When the overall deviation value is lower than the second-level deviation threshold, it indicates that the model's prediction accuracy is high, requiring only an enhancement of the algorithm's local search capability. This can be achieved by decreasing the scaling factor and increasing the crossover probability to refine the parameters and further improve model accuracy.

[0051] In this embodiment, the first, second, third, and fourth preset step sizes are fixed amplitude values ​​for hyperparameter adjustment. This avoids algorithm oscillations caused by excessively large step sizes and ineffective adjustments by excessively small step sizes. The step sizes are based on statistical calibration using historical mine optimization data, rather than calculation using a fixed formula. This is to adapt the search efficiency of the differential evolution algorithm under different deviation scenarios, and needs to be determined through reverse verification based on the historical performance of model optimization.

[0052] Specifically, historical data from the past 10-20 model optimizations in the mining process were collected, including: the overall deviation value for each optimization, such as 3.94%, 1.2%, and 5.1%; candidate step sizes for hyperparameter adjustment, such as 5%, 10%, and 15% for scaling factors; and 10%, 15%, and 20% for crossover probabilities. The adjusted algorithm performance metrics included: the rate of reduction in model prediction bias and the number of algorithm convergence iterations.

[0053] Scenario 1: When the overall deviation value is greater than or equal to the first-level deviation threshold, it is necessary to increase the scaling factor and reduce the crossover probability. By adjusting the step size, the global search capability of the differential evolution algorithm can be enhanced. First preset step size: First, optimize cases with deviations greater than or equal to the first-level deviation threshold are selected from historical data. A baseline value of 0.8 is chosen, and the optimization effects of step sizes of 5%, 10%, and 15% are tested respectively. Second, the deviation reduction rate (deviation after optimization / deviation before optimization) and the number of convergence iterations corresponding to different step sizes are calculated. Finally, the step size with the highest deviation reduction rate and a moderate number of convergence iterations is selected as the first preset step size. For example, as shown in Table 1: Table 1 First Preset Step Size Table Based on the above, we can conclude that 10% should be selected as the first preset step size.

[0054] Similarly, for the second preset step size, with a fixed baseline value of 0.6, the optimization effects of step sizes of 10%, 15%, and 20% were tested. The prediction deviation stability fluctuation of the index model was less than or equal to 5%, so 15% was selected as the second preset step size.

[0055] Scenario 2: If the overall deviation value is less than the second-level deviation threshold, it is necessary to reduce the scaling factor and increase the crossover probability. By adjusting the step size, the local search capability can be strengthened. Third preset step size: Select cases where the overall deviation value is less than the second-level deviation threshold, and test the step size reduction of 5%, 8%, and 10%. The deviation of the indicator model needs to be further reduced by more than or equal to 10%, so 8% is selected as the third preset step size.

[0056] Fourth preset step size: Test the step size by increasing it by 10%, 12%, and 15%; meet the criteria that the algorithm converges and the number of iterations is reduced by more than or equal to 20%; and select 12% as the fourth preset step size.

[0057] As can be seen from the above, this embodiment, by increasing the reference value of the scaling factor based on the first preset step size and decreasing the reference value of the crossover probability based on the second preset step size when the comprehensive deviation value is greater than or equal to the preset first-level deviation threshold, can more effectively adjust the differential evolution algorithm, thereby adjusting the parameters of the crushing block size-energy consumption model and / or the grinding energy consumption prediction model; when the comprehensive deviation value is less than the preset second-level deviation threshold, by decreasing the reference value of the scaling factor based on the third preset step size and increasing the reference value of the crossover probability based on the fourth preset step size, the differential evolution algorithm can more accurately adjust the model parameters to adapt to the deviation between the actual energy consumption data and the predicted value, thereby optimizing the intelligent optimization and control method of energy consumption throughout the entire mining and beneficiation cycle.

[0058] In one embodiment of this application, the actual energy consumption data is compared with the predicted corresponding power consumption and usage to obtain a comprehensive deviation value, including: Calculate the relative deviation rate of energy consumption prediction for the blasting, crushing and grinding stages respectively; Based on the historical average weights of energy consumption in the blasting, crushing, and grinding stages in the total energy cost of mining and beneficiation, the relative deviation rates of each stage are weighted and summed to obtain the comprehensive deviation value.

[0059] In this embodiment, the relative deviation rate is an indicator representing the degree of deviation between the predicted and actual energy consumption values ​​of a single stage, indicating the accuracy of the prediction model for that stage. The calculation formula is: (|actual energy consumption value - predicted energy consumption value / predicted energy consumption value|) × 100%. The larger the value, the lower the prediction accuracy of the model for that stage; the smaller the value, the more accurate the model prediction. The historical average weight is based on historical data from long-term mine production, statistically calculating the average proportion of energy consumption costs in each stage (blasting, crushing, and grinding) to the total energy consumption cost of mining and beneficiation. The weight indicates the degree of influence of energy consumption in that stage on the total energy consumption cost; the higher the proportion, the larger the weight, and the greater the contribution of its deviation rate to the overall deviation value.

[0060] As can be seen from the above, this embodiment calculates the relative deviation rate of energy consumption prediction for each stage separately, and then weights and sums the relative deviation rates according to the historical average weight of energy consumption in the total energy consumption cost of mining and beneficiation. This allows for an accurate comprehensive deviation between the actual energy consumption data and the predicted corresponding electricity consumption and consumption, providing a reliable basis for subsequent model adjustments.

[0061] In one embodiment of this application, a weighted sum of the relative deviation rates of each stage is obtained based on the historical average weights of energy consumption in the blasting, crushing, and grinding stages in the total energy consumption cost of mining and beneficiation, resulting in a comprehensive deviation value, including: Based on historical energy consumption data of blasting, crushing and grinding processes and their corresponding historical comprehensive deviation assessment data, a historical deviation dataset is constructed. Using historical comprehensive deviation assessment data as the dependent variable, and the relative deviation rates of the blasting, crushing, and grinding processes as independent variables, the importance score of the relative deviation rate of each process to the comprehensive deviation assessment is calculated using the random forest algorithm. Based on the importance score, the relative deviation rates of each step are weighted and fused to obtain a comprehensive deviation value.

[0062] In this embodiment, the historical deviation dataset is a dataset constructed based on historical data accumulated from long-term mine production. It includes two types of data: first, the predicted and actual energy consumption values ​​and corresponding relative deviation rates for each stage of blasting, crushing, and grinding; second, the historical comprehensive deviation assessment data corresponding to each batch of production, which can be determined through expert scoring, measured values ​​of energy consumption deviation throughout the entire process, etc. This dataset forms the basis for the subsequent calculation of weights using the random forest algorithm. The dependent variable, in statistical modeling, refers to the variable that is explained or predicted. In this embodiment, the historical comprehensive deviation assessment data is the dependent variable, and its value is determined by the deviation rates of each stage, used to characterize the overall degree of deviation in the prediction of energy consumption throughout the entire process. The independent variables, in statistical modeling, refer to the variables used to explain and predict the dependent variable. In this embodiment, the relative deviation rates of the blasting, crushing, and grinding stages are three independent variables, which are factors affecting the comprehensive deviation. The random forest algorithm is an ensemble learning algorithm based on decision trees. It constructs multiple independent decision trees and outputs results using voting or averaging methods. In this embodiment, its function is to represent the contribution of the relative deviation rate of each link to the overall deviation, and to output the importance score of each variable, replacing the traditional historical average weight.

[0063] In this embodiment, the random forest algorithm used to calculate the importance score of the deviation rate adopts an ensemble learning hierarchical structure, consisting of three layers: a decision tree base learner layer, a feature sampling layer, and a voting fusion layer. The decision tree base learner layer contains multiple independent CART regression decision trees, each responsible for learning the local mapping relationship between the independent and dependent variables. The feature sampling layer employs a random subspace strategy, randomly sampling the three input independent variables (relative deviation rates of blasting, crushing, and grinding stages), ensuring differences in the input features of each decision tree and improving the model's generalization ability. The voting fusion layer performs mean fusion of the outputs of all decision trees, ultimately outputting the importance score of the relative deviation rate for each stage. Algorithm parameters include the number of decision trees, the maximum tree depth, and the feature sampling ratio, which together determine the model's fitting accuracy and the reliability of the feature importance assessment.

[0064] In this embodiment, the importance score is an indicator output by the Random Forest algorithm, representing the degree of influence of each independent variable on the dependent variable. A higher score indicates a greater contribution of the relative deviation rate of that stage to the overall deviation, and thus a higher weight; a lower score indicates a smaller contribution and a lower weight. Weighted fusion involves multiplying the relative deviation rates of each stage by the importance score weights calculated by the Random Forest algorithm, then summing these values ​​to obtain the final overall deviation value. Compared to the traditional historical average weight, this method more accurately represents the actual impact of deviations at each stage on the entire process.

[0065] As can be seen from the above, this embodiment uses the random forest algorithm and historical deviation dataset to determine the importance score of the relative deviation rate of each link and then weights and fuses them to obtain a comprehensive deviation value, which can more accurately represent the overall deviation of energy consumption prediction in each link of the selection process.

[0066] In one embodiment of this application, after executing the optimal parameter combination, the method further includes: Correction and verification of post-explosion fragmentation distribution; Post-blast ore images are acquired, and based on image segmentation and particle size analysis algorithms, the post-blast ore images are analyzed to identify and statistically obtain the actual block size distribution data of the ore. The actual block size distribution data is compared with the post-blast block size distribution to obtain the block size distribution difference. Compare the block size distribution difference with a preset difference threshold; If the difference in block size distribution is greater than or equal to the preset difference threshold, the blasting block size prediction model is optimized based on the actual block size distribution data of historical multiple cycles and the preset decision rules, and the optimized blasting block size prediction model is applied to the next production cycle.

[0067] In this embodiment, the post-blast block size distribution correction verification is a process of detecting, comparing, and verifying the actual block size distribution of the ore after blasting, following the execution of optimal blasting parameters. Its purpose is to verify whether the output of the blasting block size prediction model matches reality, and it is a key step in the closed-loop optimization of the blasting process. The post-blast ore image is a visual image of the ore pile after blasting, captured by equipment such as industrial cameras and drones. It is the original data source for obtaining the actual block size distribution. To ensure the image is clear and has sufficient coverage, it needs to include ore blocks from different regions. Image segmentation and grain size analysis algorithms: Image segmentation is an algorithm that separates individual ore blocks from the background in an ore image (e.g., based on threshold segmentation, edge detection, or deep learning segmentation models), identifying the contour of each independent ore block. The grain size analysis algorithm, based on the segmented ore block contours, calculates the equivalent grain size, area, and other parameters of each ore block, and then statistically analyzes the proportion of ore in different grain size ranges. It is a technique for obtaining the actual block size distribution.

[0068] In this embodiment, the algorithm for post-blast ore image segmentation adopts the U-Net deep learning semantic segmentation hierarchical structure, which is divided into three layers: an encoding layer, a decoding layer, and a skip connection layer. The encoding layer consists of four sets of convolutional + pooling modules. Each convolutional module includes two 3×3 convolution operations and one 2×2 max pooling operation, responsible for downsampling the input image and extracting multi-scale ore features from shallow textures to deep contours. The decoding layer consists of four sets of deconvolution + convolutional modules. Each deconvolutional module includes one 2×2 deconvolution operation and two 3×3 convolution operations, responsible for upsampling the feature map of the encoding layer and gradually restoring the image resolution. The skip connection layer concatenates and fuses the feature map of the encoding layer at the corresponding scale with the feature map of the decoding layer to compensate for the ore edge details lost during downsampling. The core related parameters of the algorithm include the convolution kernel size, pooling / deconvolution stride, and activation function type, which together determine the segmentation accuracy of the ore region and the background region.

[0069] In this embodiment, the granularity analysis algorithm for ore block size statistics adopts a three-level feature calculation and statistical structure, consisting of a connected component extraction layer, an ore block geometric parameter calculation layer, and a granularity classification statistical layer. The connected component extraction layer, based on the binary label image output from image segmentation, uses an 8-neighborhood connectivity rule to divide the ore region with a pixel value of 1 in the label image into multiple independent connected components, each corresponding to one independent ore block. The geometric parameter calculation layer is the core of the algorithm, calculating three key parameters for each connected component: equivalent diameter, projected area, and aspect ratio. The core parameter is the pixel-to-physical size conversion coefficient (determined by on-site calibration). The granularity classification statistical layer, based on a preset block size classification standard, categorizes each connected component into its corresponding granularity interval according to its equivalent diameter, and then calculates the proportion of projected area for each granularity, achieving quantitative output of block size distribution data. Each level of the algorithm completes the transformation from segmented image to block size distribution data through a process of domain extraction, parameter calculation, and granularity classification statistics.

[0070] In this embodiment, the actual block size distribution data is the true distribution of ore block size after blasting, obtained through image analysis. It is represented by the mass percentage of ore in different particle size ranges, for example, less than 10mm accounts for 15%, 10-50mm for 60%, 50-100mm for 20%, and greater than 100mm for 5%. The block size distribution difference is an indicator of the degree of deviation between the actual block size distribution and the block size distribution output by the blasting block size prediction model. In this embodiment, the calculation method is Barcol distance; a larger value indicates a greater deviation. The preset difference threshold is a pre-set critical value for determining whether the blasting block size prediction model needs optimization. For example, the Barcol distance threshold is 0.15. When the block size distribution difference is greater than or equal to the preset difference threshold, it indicates that the prediction accuracy of the blasting block size prediction model is insufficient, and optimization needs to be initiated. The historical multi-period actual block size distribution data is the actual block size distribution data of the ore after blasting collected from multiple production cycles in the past (e.g., the last 3 months, 50 batches), including block size characteristics under different blasting parameters and geological conditions. This data serves as training data for optimizing the blasting block size prediction model. The preset decision rules are optimization rules for the blasting block size prediction model based on the block size distribution difference. For example, when the block size distribution difference is greater than or equal to the preset difference threshold, the training set is supplemented with high deviation data from the last 10 periods. These are specific criteria for guiding the optimization of the blasting block size prediction model.

[0071] As can be seen from the above, this embodiment improves the accuracy of prediction by correcting and verifying the post-blast block size distribution after executing the optimal parameter combination; it obtains the actual block size distribution data by processing the post-blast ore image using image segmentation and particle size analysis algorithms, thus acquiring a more accurate picture of the actual situation; it obtains the difference degree by comparing the actual block size distribution data with the post-blast block size distribution, which can intuitively represent the gap between the prediction and the actual situation; it determines whether the model needs to be optimized by comparing the difference degree with a threshold; if the block size distribution difference degree is greater than or equal to the preset difference degree threshold, it optimizes the blasting block size prediction model based on historical multi-cycle actual block size distribution data and decision rules and applies it to the next cycle, which can improve the accuracy and applicability of the model, thereby enhancing the effect of intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle.

[0072] In one embodiment of this application, based on image segmentation and particle size analysis algorithms, the post-blast ore image is analyzed to identify and statistically obtain the actual particle size distribution data of the ore, including: The post-explosion ore image is preprocessed to obtain the preprocessed ore image; The preprocessed image is input into a pre-trained deep semantic segmentation network model to obtain a label map that marks each ore pixel; The pixel regions belonging to the ore in the label image are divided into independent connected components to obtain the ore connected components. Calculate the equivalent diameter, aspect ratio, and projected area of ​​each ore connectivity region; Based on the preset block size classification standard, each ore connectivity domain is classified into the corresponding particle size according to the equivalent diameter; The actual block size distribution data is obtained by calculating the area ratio of the ore connected domain in the corresponding grain size based on the projected area.

[0073] In this embodiment, image preprocessing involves a series of basic processing operations performed on the original post-blast ore image. The purpose is to eliminate interference such as noise, uneven lighting, and shooting distortion, thereby improving the accuracy of subsequent segmentation and analysis. The deep semantic segmentation network model is a deep learning-based image segmentation model. This embodiment uses U-Net, which can accurately identify the category of each pixel in the image, such as ore pixels and background pixels, and outputs a label map annotating the pixel categories. It is a tool for achieving accurate ore block segmentation.

[0074] In this embodiment, the U-Net model used for post-blast ore image segmentation adopts a symmetrical encoder-decoder hierarchical structure, consisting of three layers: an encoder downsampling layer, a skip connection layer, and a decoder upsampling layer. The encoder downsampling layer includes four convolutional modules, each consisting of two 3×3 convolution operations plus one 2×2 max pooling operation. The activation function of the convolutional layer is ReLU, which is responsible for feature extraction and dimensionality compression of the input image, gradually transforming the input 3-channel RGB image into multi-scale ore texture and contour feature maps. The skip connection layer concatenates and fuses the feature maps of each scale of the encoder layer with the feature maps of the corresponding scale of the decoder layer, compensating for the ore edge details lost during downsampling. The decoder upsampling layer includes four deconvolutional modules, each consisting of one 2×2 deconvolution operation plus two 3×3 convolution operations. Finally, a 1×1 convolution maps the feature map into a single-channel binary label map, achieving accurate differentiation between ore pixels and background pixels. The core parameters of the model include the kernel size, pooling / deconvolution stride, and number of feature map channels, which together determine the segmentation accuracy and edge integrity of the ore region.

[0075] In this embodiment, the label map is the output of the deep semantic segmentation model. It is a pixel-level annotation map with the same size as the original image. Each pixel is labeled as a category such as ore or background, intuitively presenting the location and outline of the ore region in the image. A connected component is a set of pixels of the same category that are interconnected in the label map; in this embodiment, it is the pixel region corresponding to a single, independent ore block. By segmenting the connected components, multiple ore blocks in the image can be separated one by one, achieving single-block recognition. The equivalent diameter is an equivalent parameter used to represent the size of the ore block. It is calculated as the diameter of a circle with the same projected area as the ore connected component and is an indicator of the ore block size. The aspect ratio is the length of the long side and the short side of the circumscribed rectangle of the ore connected component. It is used to help describe the shape of the ore block; for example, an aspect ratio close to 1 indicates a spherical shape, while an aspect ratio much greater than 1 indicates a long strip shape. This can serve as a supplementary feature for block size analysis.

[0076] In this embodiment, the projected area is the pixel area of ​​the ore connected regions on the image plane, which needs to be converted into actual physical area and is the basis for calculating the mass and area proportions of ore blocks. The block size classification standard is a pre-defined rule for dividing ore particle size ranges, such as less than 10mm, 10-50mm, 50-100mm, and greater than 100mm, which is the basis for classifying ore connected regions into the corresponding particle size class. The area proportion is the proportion of the total projected area of ​​all ore connected regions at a certain particle size class to the total projected area of ​​all ore connected regions in the image. It represents the block size distribution and can be approximately equivalent to the mass proportion, conforming to the statistical conventions of mine block size.

[0077] As can be seen from the above, this embodiment can improve image quality and remove interference factors by preprocessing the post-blast ore image; accurately segment the ore by labeling ore pixels using a pre-trained deep semantic segmentation network model; divide the ore pixel region into connected components for further analysis of shape features; calculate the equivalent diameter, aspect ratio, and projected area of ​​the connected components to provide a basis for subsequent classification; clearly classify ore size grades according to preset grading standards and equivalent diameter classification granularity; and obtain accurate actual block size distribution data by calculating the proportion of ore connected component area in the granularity based on the projected area, which is beneficial for correcting and verifying the post-blast block size distribution, optimizing the blasting block size prediction model, making the model more consistent with the actual situation, improving the accuracy of energy consumption prediction in subsequent production cycles, and helping to achieve intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle.

[0078] In one embodiment of this application, preprocessing is performed on a post-blast ore image to obtain a preprocessed ore image, including: Illumination correction is performed on the post-blast ore image data to obtain corrected post-blast ore image data; Noise suppression processing is performed on the corrected post-blast ore image data to obtain denoised post-blast ore image data; The denoised post-blast ore image data is input into a preset attention enhancement model to obtain enhanced post-blast ore image data, which serves as the preprocessed ore image.

[0079] In this embodiment, illumination correction addresses the uneven illumination problem in post-explosion ore images caused by the shooting environment (e.g., strong light, shadow, backlight). It adjusts the brightness and contrast of image pixels through algorithms to make the texture and contour of the ore area clearly visible. This embodiment uses the Retinex algorithm to eliminate the interference of illumination on subsequent segmentation.

[0080] In this embodiment, the Retinex algorithm for post-blast ore image illumination correction employs a multi-scale decomposition-luminance reconstruction hierarchical structure, consisting of three layers: an image color space conversion layer, a multi-scale Gaussian filtering layer, and a reflection component reconstruction layer. The color space conversion layer converts the input RGB format ore image to the HSV color space, separating the luminance channel (V channel) and chrominance channel (H / S channel), correcting only the luminance channel to avoid chrominance distortion. The multi-scale Gaussian filtering layer is the core of the algorithm, filtering the luminance channel using three sets of Gaussian kernels at different scales to calculate the illumination components at different scales. The core parameters are the Gaussian kernel scale, gain coefficient, and offset coefficient. The reflection component reconstruction layer, based on the Retinex theory of image = illumination component × reflection component, eliminates illumination component interference, reconstructs a uniformly illuminated luminance channel, and then merges it with the original chrominance channel to convert it back to RGB space, outputting the corrected image. Each layer of the algorithm, through a channel-multi-scale filtering-reconstruction process, solves the problems of uneven illumination, overexposure, and underexposure in ore images from open-pit mine scenes.

[0081] In this embodiment, noise suppression processing is the operation of removing random noise from the image, such as camera sensor noise, salt-and-pepper noise caused by environmental interference, and Gaussian noise. This embodiment uses Gaussian filtering to prevent noise from being misidentified as ore contours, thereby improving segmentation accuracy. The attention enhancement model is a deep learning-based image enhancement model that uses an attention mechanism to focus on key features of ore blocks in the image, such as edges and textures, while weakening background interference, such as dirt and weeds, and strengthening the pixel features of the ore region, enabling more accurate identification of ore blocks in subsequent steps.

[0082] In this embodiment, the attention enhancement model for ore image feature enhancement adopts a Convolutional Block Attention Module (CBAM) hierarchical structure, consisting of a three-layer core architecture: a channel attention sublayer, a spatial attention sublayer, and a feature fusion layer. The channel attention sublayer extracts channel-level features from the input feature map through global average pooling and global max pooling. Channel weights are then calculated via a shared two-layer fully connected network, strengthening the feature channels corresponding to the ore texture and weakening background noise channels. The spatial attention sublayer, based on the channel attention-weighted feature map, generates a spatial attention weight map through channel-dimensional average pooling and max pooling, focusing on key spatial features such as the edges and contours of the ore block. The feature fusion layer performs a weighted multiplication of the channel attention weights, spatial attention weights, and the original feature map to output the enhanced feature map. The core parameters of the model, including the number of neurons in the hidden layers of the fully connected network, the pooling kernel size, and the attention weight activation function, collectively determine the accuracy and efficiency of feature enhancement.

[0083] In this embodiment, the corrected post-blast ore image data is the ore image data after illumination correction, which solves the problem of uneven illumination and makes the brightness difference between the ore and the background more reasonable, laying the foundation for subsequent processing. The denoised post-blast ore image data is the image data after noise suppression processing, which eliminates random noise, makes the image smoother, and removes false burrs from the ore contours. The enhanced post-blast ore image data is the final preprocessed image after processing by the attention enhancement model. The edge and texture features of the ore blocks are enhanced, and background interference is weakened. It is the final data input to the semantic segmentation model.

[0084] As can be seen from the above, this embodiment can eliminate the influence of illumination by performing illumination correction on the post-blast ore image data, making the image brightness uniform; noise suppression processing can remove noise interference in the image; inputting the denoised image into the attention enhancement model can highlight the key features of the ore in the image, providing a high-quality pre-processed image for the subsequent accurate identification and statistical acquisition of the actual block size distribution data of the ore, thereby helping to correct and verify the post-blast block size distribution, improve the accuracy of the blasting block size prediction model, and ultimately realize the intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle.

[0085] Corresponding to the intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle in the above embodiment, Figure 2 This is a structural block diagram of a smart energy consumption optimization and control system for the entire mining and beneficiation cycle, provided as an embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 2 The intelligent optimization and control system for energy consumption throughout the entire mining and beneficiation cycle 20 includes: blasting module 21, crushing module 22, grinding module 23, collaborative optimization module 24, scheme execution module 25, energy efficiency deviation module 26, and adjustment and update module 27.

[0086] Among them, the blasting module 21 is used to obtain blasting design parameters and geological parameters in the mining and beneficiation blasting process, input the constructed blasting block size prediction model, and obtain the post-blast block size distribution and blasting energy consumption. The crushing module 22 is used to input the post-explosion block size distribution into the preset crushed block size-energy consumption model, and to obtain the equipment operating parameters and working condition parameters of the crusher, and calculate the power consumption of the crushing process. The grinding module 23 is used to obtain the operating parameters of the grinding mill and the characteristics of the ore, and to obtain the power consumption and steel ball consumption of the grinding process based on the preset grinding energy consumption prediction model. The collaborative optimization module 24 is used to obtain the total energy consumption cost of mining and beneficiation based on the power consumption of blasting, crushing and grinding, and steel ball consumption. It constructs a comprehensive objective function with the goal of minimizing the total energy consumption cost of mining and beneficiation and maximizing the total processing capacity of mining and beneficiation. The constraints are that the fineness of the grinding product is within the quality constraint range and the equipment operating parameters are within the safe range. The multi-objective optimization algorithm is used to collaboratively optimize the blasting design parameters, the equipment operating parameters of the crusher and the operating parameters of the grinding mill to obtain the optimal parameter combination. The scheme execution module 25 is used to execute the optimal parameter combination and collect actual operating data and actual energy consumption data; The energy efficiency deviation module 26 is used to compare the actual energy consumption data with the predicted corresponding power consumption and consumption to obtain a comprehensive deviation value. The adjustment and update module 27 is used to adjust the parameters of the crushed block size-energy consumption model and / or the grinding energy consumption prediction model using an optimization algorithm when the comprehensive deviation value is greater than the preset deviation threshold, so as to obtain the optimized model parameters and update the corresponding model.

[0087] See Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 3 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned device embodiments, for example... Figure 2 The functions of the following modules are shown: blasting module 21, crushing module 22, grinding module 23, collaborative optimization module 24, scheme execution module 25, energy efficiency deviation module 26, and adjustment and update module 27.

[0088] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0089] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0090] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0091] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation methods described in any embodiment of the intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle provided in the embodiments of this application, or they can execute the implementation methods of the electronic devices described in the embodiments of this application, which will not be repeated here.

[0092] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0093] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0094] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0095] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0096] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0097] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0098] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0099] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent optimization and control of energy consumption throughout the entire mining and beneficiation cycle, characterized in that, include: Obtain blasting design parameters and geological parameters in the mining and beneficiation blasting process, input them into the constructed blasting block size prediction model, and obtain the post-blast block size distribution and blasting energy consumption; The post-explosion block size distribution is input into a preset crushed block size-energy consumption model, and the equipment operating parameters and working condition parameters of the crusher are obtained to calculate the power consumption of the crushing process. The operating parameters of the grinding mill and the characteristics of the ore are obtained. Based on the preset grinding energy consumption prediction model, the power consumption and steel ball consumption in the grinding process are obtained. Based on the power consumption of the blasting, crushing, and grinding stages, as well as the steel ball consumption, the total energy consumption cost of mining and beneficiation is obtained. A comprehensive objective function is constructed with the goal of minimizing the total energy consumption cost of mining and beneficiation and maximizing the total processing capacity. The constraints are that the fineness of the grinding product is within the quality constraint range and the equipment operating parameters are within the safe range. A multi-objective optimization algorithm is used to collaboratively optimize the blasting design parameters, the crusher's equipment operating parameters, and the grinding mill's operating parameters to obtain the optimal parameter combination. Execute the optimal parameter combination and collect actual operating data and actual energy consumption data; The actual energy consumption data is compared with the predicted corresponding power consumption and consumption to obtain a comprehensive deviation value; When the overall deviation value is greater than the preset deviation threshold, the parameters of the crushed block size-energy consumption model and / or the grinding energy consumption prediction model are adjusted using an optimization algorithm to obtain the optimized model parameters and update the corresponding model.

2. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle as described in claim 1, characterized in that, The total energy cost of mining and beneficiation, derived from the power consumption of the blasting, crushing, and grinding processes, as well as the consumption of steel balls, includes: Obtain cost calculation benchmark data, which includes time-of-use electricity price, diesel unit price, steel ball unit price, and comprehensive efficiency coefficient of equipment in each stage; The power consumption of the blasting, crushing and grinding processes are converted into electricity costs based on the electricity price corresponding to the specific time point in time. The steel ball consumption in the grinding process is converted into steel ball loss cost based on the unit price of the steel ball; Identify and acquire the non-electric direct energy costs related to the blasting, crushing, and grinding processes, including the cost of explosives used in blasting and the fuel consumption costs of large diesel equipment. Based on the comprehensive efficiency coefficient of each component, the electrical cost and non-electric direct energy cost are corrected to obtain the corrected electrical cost; the comprehensive efficiency coefficient is used to characterize the energy conversion efficiency loss caused by equipment aging and maintenance. The total energy consumption cost of mining and beneficiation is obtained by summing the corrected electrical energy cost, the steel ball loss cost, and the non-electric direct energy cost.

3. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle according to claim 1, characterized in that, The optimization algorithm is a differential evolution algorithm; the optimization algorithm is used to adjust the parameters of the blasting block size prediction model, the crushed block size-energy consumption coupling model, and / or the grinding energy consumption prediction model to obtain optimized model parameters, including: The solution space of the differential evolution algorithm is determined based on the dimension and physical feasible range of the adjustable parameter subset of the model to be adjusted. Based on the combined deviation between the actual energy consumption data and the predicted corresponding power consumption and energy expenditure, the hyperparameter combination of the differential evolution algorithm is determined.

4. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle as described in claim 3, characterized in that, The hyperparameters include scaling factor and crossover probability; The step of determining the hyperparameter combination of the differential evolution algorithm based on the comprehensive deviation between the actual energy consumption data and the predicted corresponding power consumption and energy consumption includes: If the overall deviation value is greater than or equal to the preset first-level deviation threshold, the reference value of the scaling factor is increased based on the first preset step size, and the reference value of the crossover probability is decreased based on the second preset step size. If the overall deviation value is less than the preset second-level deviation threshold, the reference value of the scaling factor is reduced based on the third preset step size, and the reference value of the crossover probability is increased based on the fourth preset step size.

5. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle according to claim 1, characterized in that, The step of comparing the actual energy consumption data with the predicted corresponding electricity consumption and usage to obtain a comprehensive deviation value includes: Calculate the relative deviation rate of energy consumption prediction for the blasting, crushing and grinding processes respectively; Based on the historical average weight of the energy consumption of the blasting, crushing and grinding processes in the total energy consumption cost of mining and beneficiation, the relative deviation rates of each process are weighted and summed to obtain the comprehensive deviation value.

6. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle according to claim 5, characterized in that, The comprehensive deviation value is obtained by weighting and summing the relative deviation rates of the blasting, crushing, and grinding processes according to their historical average weights in the total energy consumption cost of mining and beneficiation, and includes: Based on the historical energy consumption data of the blasting, crushing and grinding processes and their corresponding historical comprehensive deviation assessment data, a historical deviation dataset is constructed. Using the historical comprehensive deviation assessment data as the dependent variable, and the relative deviation rates of the blasting, crushing, and grinding processes as independent variables, the importance score of the relative deviation rate of each process to the comprehensive deviation assessment is calculated using the random forest algorithm. Based on the importance score, the relative deviation rates of each step are weighted and fused to obtain a comprehensive deviation value.

7. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle according to claim 1, characterized in that, After executing the optimal parameter combination, the process further includes: The post-explosion fragmentation distribution was corrected and verified. The post-blast ore image is acquired, and the image is analyzed based on image segmentation and particle size analysis algorithms to identify and statistically obtain the actual block size distribution data of the ore. The actual block size distribution data is compared with the post-blast block size distribution to obtain the block size distribution difference degree; The block size distribution difference is compared with a preset difference threshold; If the block size distribution difference is greater than or equal to a preset difference threshold, the blasting block size prediction model is optimized based on historical multi-cycle actual block size distribution data and preset decision rules, and the optimized blasting block size prediction model is applied to the next production cycle.

8. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle according to claim 7, characterized in that, The image segmentation and particle size analysis algorithm is used to analyze the post-blast ore image, identify and statistically obtain the actual particle size distribution data of the ore, including: The post-explosion ore image is preprocessed to obtain a preprocessed ore image; The preprocessed image is input into a pre-trained deep semantic segmentation network model to obtain a label map that marks each ore pixel; The pixel regions belonging to the ore in the label image are divided into independent connected regions to obtain the ore connected regions; Calculate the equivalent diameter, aspect ratio, and projected area of ​​each of the ore connected regions; According to the preset block size classification standard, each ore connectivity domain is classified into the corresponding particle size based on the equivalent diameter; The area ratio of the ore connected domain in the corresponding grain size is calculated based on the projected area to obtain the actual block size distribution data.

9. The intelligent optimization and control method for energy consumption throughout the entire mining and beneficiation cycle as described in claim 8, characterized in that, The preprocessing of the post-blast ore image to obtain a preprocessed ore image includes: The post-blast ore image data is subjected to illumination correction to obtain corrected post-blast ore image data; The corrected post-blast ore image data is subjected to noise suppression processing to obtain denoised post-blast ore image data; The denoised post-blast ore image data is input into a preset attention enhancement model to obtain enhanced post-blast ore image data, which serves as the preprocessed ore image.

10. A smart optimization and control system for energy consumption throughout the entire mining and beneficiation cycle, characterized in that, include: The blasting module is used to obtain blasting design parameters and geological parameters in the mining and beneficiation blasting process. By inputting the constructed blasting block size prediction model, the module can obtain the post-blast block size distribution and blasting energy consumption. The crushing module is used to input the post-explosion block size distribution into a preset crushed block size-energy consumption model, and to obtain the equipment operating parameters and working condition parameters of the crusher, and calculate the power consumption of the crushing process. The grinding module is used to obtain the operating parameters of the grinding mill and the characteristics of the ore. Based on the preset grinding energy consumption prediction model, it obtains the power consumption and steel ball consumption of the grinding process. The collaborative optimization module is used to obtain the total energy consumption cost of mining and beneficiation based on the power consumption of the blasting, crushing, and grinding stages, as well as the steel ball consumption. It constructs a comprehensive objective function with the goal of minimizing the total energy consumption cost of mining and beneficiation and maximizing the total processing capacity. The constraints are that the fineness of the grinding product is within the quality constraint range and the equipment operating parameters are within the safe range. The module uses a multi-objective optimization algorithm to collaboratively optimize the blasting design parameters, the equipment operating parameters of the crusher, and the operating parameters of the grinding mill to obtain the optimal parameter combination. The scheme execution module is used to execute the optimal parameter combination and collect actual operating data and actual energy consumption data; The energy efficiency deviation module is used to compare the actual energy consumption data with the predicted corresponding power consumption and consumption to obtain a comprehensive deviation value. The adjustment and update module is used to adjust the parameters of the crushed block size-energy consumption model and / or grinding energy consumption prediction model using an optimization algorithm when the comprehensive deviation value is greater than a preset deviation threshold, so as to obtain the optimized model parameters and update the corresponding model.

Images