Enterprise electrical energy monitoring and analyzing method and system based on industrial internet

Abstract

The invention discloses an enterprise electric power energy monitoring analysis method and system based on the industrial internet, relates to the technical field of energy monitoring, and provides reliable data for a system by collecting operation electric power data and production management system data through a sensor, preprocessing the data and storing the data in a historical data set Dhis. And constructing an energy consumption prediction model through the historical data set Dhis to provide data support for a response strategy. And constructing a reward and punishment function based on the energy consumption prediction model, comparing the reward and punishment function with a reward and punishment threshold Rth, generating a power supply mode decision Mpow, and intelligently adjusting the power supply mode. And outputting the optimized production schedule Sopt by taking the total energy consumption cost TEC and the schedule stability SSP as optimization targets, so as to improve the production continuity. Model parameters are dynamically corrected through comparison of an optimization effect evaluation value Oscore and a threshold value Oth, it is ensured that the system continuously approaches the actual working condition in long-term operation, and strategy failure caused by equipment aging or process change is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of energy monitoring, and specifically to an enterprise electric power energy monitoring and analysis method and system based on the industrial Internet. Background Art

[0002] In the wave of the deep integration of the industrial Internet and intelligent manufacturing, industrial automation control technology is gradually evolving towards a data-driven refinement direction. As an important branch of industrial automation, in the field of enterprise energy management, through real-time perception and intelligent decision-making, the monitoring and analysis of electric power energy has become the core breakthrough point for enterprise energy efficiency optimization due to its high energy consumption attribute and complex working condition characteristics. However, traditional power monitoring systems are mostly limited to the acquisition of the operating status of single devices, and it is difficult to support the collaborative analysis and dynamic regulation of multi-source heterogeneous data. A solution covering the closed loop of "monitoring - prediction - control - feedback" is required to minimize energy consumption costs while ensuring production continuity.

[0003] Currently, enterprise electric power energy management generally faces three major bottlenecks: First, data fragmentation. Most systems rely on independently deployed sensors and SCADA monitoring platforms, and electric power parameters such as current and voltage are separated from production plans and external electricity price data, resulting in a lack of cross-dimensional correlation in energy consumption analysis and an inability to identify high-energy-consuming devices during high electricity price periods. Second, decision-making lag. Traditional methods rely on manual experience to formulate scheduling and power supply strategies, and it is difficult to respond in a timely manner to grid fluctuations or abnormal energy storage charge conditions. Finally, prediction and control are disjointed. Although existing prediction models can estimate energy consumption trends, they are not linked to dynamic electricity prices and energy storage states, resulting in optimization strategies deviating from actual working conditions. These problems have left enterprises in a passive position among high electricity bills, unplanned outages, and excessive energy consumption. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides an enterprise electric power energy monitoring and analysis method and system based on the industrial Internet, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: An enterprise electric power energy monitoring and analysis method based on the industrial Internet includes the following steps: S1. Real-time collect the operating electric power data of industrial devices through electric power data collection devices, and real-time synchronize the order scheduling S_ord, grid time-of-use electricity price C_gri, and energy storage charge state SOC in the production management system data. Preprocess the operating electric power data and production management system data, extract the characteristics of the operating electric power data, construct the current electric power feature vector Vec, and store the current electric power feature vector Vec and production management system data in the historical data set D_his; S2. Establish an energy consumption prediction baseline E_base based on the historical data set D_his, build an energy consumption prediction model based on the energy consumption prediction baseline E_base and the current power feature vector Vec, and obtain an energy consumption prediction value E_pred through the energy consumption prediction model; S3, construct a reward and punishment function based on the energy consumption prediction value E_pred and the grid time-of-use electricity price C_gri, calculate the reward and punishment function value R of the current strategy and compare it with the preset reward and punishment threshold R_th, and generate a power supply mode decision M_pow according to the comparison result; S4, construct a multi-objective scheduling optimization algorithm based on the energy consumption prediction value E_pred, the power supply mode decision M_pow and the order schedule S_ord, calculate the multi-objective scheduling optimization value F_s and output the optimized production schedule S_opt; S5. Collect the actual electricity price C_act and actual average energy consumption E_avg of the optimized production schedule S_opt and build an optimization effect evaluation algorithm. Calculate the optimization effect evaluation value O_score and compare it with the optimization effect threshold O_th. According to the comparison results, execute the iterative optimization strategy and store the data involved in the algorithm calculation into the historical data set D_his.

[0006] Preferably, S1 includes S11 and S12; S11, collect the operating power data of industrial equipment in real time through the power data collection equipment, the operating power data includes the real-time current data I_t at time t, the real-time voltage data V_t at time t and the power factor PF_t at time t, and the order schedule S_ord of the production management system, the time-of-use electricity price C_gri of the power grid and the energy storage charge state SOC_t at time t in real time synchronization; Among them, the real-time current data I_t is obtained through the current sensor, the real-time voltage data V_t is obtained through the smart meter, the power factor PF_t at time t is obtained through the power quality analyzer, and the actual power consumption E_act(t) at time t is obtained through the energy consumption calculation formula. The order schedule S_ord includes the preset order number O_ID, order quantity O_qua, planned start time T_sta, planned end time T_end and order priority P_prio. The grid time-of-use electricity price C_gri is obtained from the power company by using the API interface, and the energy storage charge state SOC_t at time t is obtained through the bus protocol of the BMS.

[0007] Preferably, S12, performing data cleaning on the operating power data and the production management system data, removing data outliers and aligning timestamps of different devices, and performing Z-Score standardization on the operating power data and the production management system data after removing data outliers; Extract the frequency-domain features of the real-time current data \(I_t\) at time \(t\) from the operating power data through Fourier transform, obtain the main frequency amplitude \(FFT(I_t)\) of the current spectrum at time \(t\), calculate and extract the time-domain features of the real-time voltage data \(V_t\) and the power factor \(PF_t\) at time \(t\) from the operating power data, obtain the root mean square voltage \(RMS(V_t)\) at time \(t\) and the rate of change of power factor \(\Delta PF_t\) at time \(t\), construct the current power feature vector \(Vec = \{FFT(I_t), RMS(V_t), \Delta PF_t, E_{act}(t)\}\) based on the operating power data after feature extraction, and store the current power feature vector \(Vec\) and the production management system data in the historical data set \(D_{his}\); Among them, the formula for calculating the root mean square voltage \(RMS(V_t)\) at time \(t\) is as follows: ; In the formula, \(N\) represents the number of sampling points within a single time window, \(k\) represents the sampling point traversal index value, and \(V_t(k)\) represents the real-time voltage data of the \(k\)-th sampling point at time \(t\); Among them, the formula for calculating the rate of change of power factor \(\Delta PF_t\) at time \(t\) is as follows: ; In the formula, \(\Delta t\) represents the time sliding window.

[0008] Preferably, S2 includes S21 and S22; S21. Based on the actual power consumption \(E_{act}(t)\) at time \(t\) in the historical data set \(D_{his}\), obtain the average actual power consumption \(\mu E_{his}(t)\) in the same time period in the historical data, and establish the power consumption prediction baseline \(E_{base}(t + \Delta t)\) at time \(t+\Delta t\) in combination with the rate of change of power factor \(\Delta PF_t\) at time \(t\); Among them, the formula for the power consumption prediction baseline \(E_{base}(t + \Delta t)\) at time \(t+\Delta t\) is as follows: ; In the formula, \(w1\) represents the power factor influence factor set based on historical data.

[0009] Preferably, S22. Based on the power consumption prediction baseline \(E_{base}(t + \Delta t)\) at time \(t+\Delta t\), the main frequency amplitude \(FFT(I_t)\) of the current spectrum at time \(t\) in the current power feature vector \(Vec\), the root mean square voltage \(RMS(V_t)\) at time \(t\), the average main frequency amplitude \(\mu FFT(I_t)\) of the current spectrum in the same time period in the historical data set \(D_{his}\), and the average root mean square voltage \(\mu RMS(V_t)\) in the same time period, construct a power consumption prediction model, and obtain the power consumption prediction value \(E_{pred}(t + \Delta t)\) at time \(t+\Delta t\) through the power consumption prediction model; Among them, the energy consumption prediction model formula is as follows: ; ; In the formula, ECC represents the energy consumption correction term, and w2 represents the preset frequency domain correction coefficient.

[0010] Preferably, S3 includes S31 and S32; S31. Construct a reward and punishment function based on the predicted energy consumption value E_pred(t + △t) at time t + △t and the time-of-use grid electricity price C_gri(t + △t) at time t + △t, and calculate the reward and punishment function value R of the current strategy; Among them, based on the predicted energy consumption prediction value Ehis_pred(t1) in the historical database at time t1 and the actual power consumption E_act(t1) in the historical dataset D_his at time t1, calculate the energy consumption prediction confidence E. The calculation formula of the energy consumption prediction confidence E is as follows: ; The reward and punishment function formula is as follows: ; In the formula, C_base represents the base electricity price, α represents the weight coefficient of the preset electricity price cost item, and β represents the weight coefficient of the preset prediction error penalty.

[0011] Preferably, S32. Compare the reward and punishment function value R of the current strategy with the preset reward and punishment threshold R_th, and compare the state of charge SOC_t of the energy storage at time t with the set minimum safe state of charge SOC_min of the energy storage. Generate the power supply mode decision M_pow(t) at time t according to the comparison result; If the reward and punishment function value R of the current strategy ≥ the reward and punishment threshold R_th and the state of charge SOC_t of the energy storage at time t ≥ 1.5 * the minimum safe state of charge SOC_min of the energy storage, then the power supply mode decision M_pow(t) at time t is to use the energy storage for power supply; If the reward and punishment function value R of the current strategy < the reward and punishment threshold R_th or the state of charge SOC_t of the energy storage at time t < 1.5 * the minimum safe state of charge SOC_min of the energy storage, then the power supply mode decision M_pow(t) at time t is to use the mains power supply.

[0012] Preferably, S4 includes S41; S41. Obtain the production scheduling time period set M by counting the production scheduling time periods, assign a unique identifier to each time period, M = {t_1, t_2, ……, t_m}, and construct a multi-objective scheduling optimization algorithm based on the predicted energy consumption value E_pred(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, the power supply mode decision M_pow(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, the time-of-use electricity price C_gri(t_i) of the power grid in the time period t_i with the unique identifier i in the production scheduling time period set M, and the order scheduling S_ord. Calculate and traverse to find the maximum value of the multi-objective scheduling optimization value F_s, and output the parameters when the maximum value of the multi-objective scheduling optimization value F_s is obtained as the optimized production scheduling S_opt and output it; Among them, the formula of the multi-objective scheduling optimization algorithm is as follows: ; ; ; ; In the formula, TEC represents the predicted total energy consumption cost, SSP represents the scheduling stability, L represents the total number of orders, j represents the order number, t(j, new) represents the execution time of order j in the optimized scheduling, t(j, old) represents the execution time of order j in the original scheduling, T_total represents the total production cycle, δ represents the influence factor of the power supply mode decision M_pow(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M on the total energy consumption cost TEC, λ represents the weight coefficient of the scheduling stability SSP, and exp represents the natural exponential function.

[0013] Preferably, S5 includes S51; S51. Construct an optimization effect evaluation algorithm based on the actual electricity price C_act, the actual average energy consumption E_avg, and the predicted total energy consumption cost TEC of the optimized production scheduling S_opt implemented, and calculate the optimization effect evaluation value O_score; Among them, the formula of the optimization effect evaluation algorithm is as follows: ; Compare the optimization effect evaluation value O_score with the optimization effect threshold O_th, execute the iterative optimization strategy according to the comparison result, and store the data participating in the algorithm calculation in the historical data set D_his; If the optimization effect evaluation value O_score ≤ the optimization effect threshold O_th, it is determined that the optimization is qualified and the current strategy is retained; If the optimization effect evaluation value O_score > the optimization effect threshold O_th, it is determined that the optimization is unqualified, and the model is retrained, and the weight coefficients w1 and w2 of the prediction model and the weight coefficients α and β in the reward and punishment function are corrected.

[0014] An enterprise power energy monitoring and analysis system based on the industrial Internet, including a data acquisition and processing module, a prediction model construction module, a power supply mode decision-making module, an optimized production scheduling module, and an iterative optimization module; The data acquisition and processing module collects the operating power data of industrial equipment in real time through power data acquisition devices, synchronizes the order scheduling S_ord, the grid time-of-use electricity price C_gri, and the energy storage charge state SOC in the production management system data in real time, preprocesses the operating power data and the production management system data, extracts the operating power data features, constructs the current power feature vector Vec, and stores the current power feature vector Vec and the production management system data in the historical data set D_his. The prediction model construction module establishes an energy consumption prediction baseline E_base based on the historical data set D_his, constructs an energy consumption prediction model based on the energy consumption prediction baseline E_base and the current power feature vector Vec, and obtains the energy consumption prediction value E_pred through the energy consumption prediction model. The power supply mode decision-making module constructs a reward and punishment function based on the energy consumption prediction value E_pred and the grid time-of-use electricity price C_gri, calculates the reward and punishment function value R of the current strategy, compares it with the preset reward and punishment threshold R_th, and generates a power supply mode decision M_pow according to the comparison result. The optimized production scheduling module constructs a multi-objective scheduling optimization algorithm based on the energy consumption prediction value E_pred, the power supply mode decision M_pow, and the order scheduling S_ord, calculates the multi-objective scheduling optimization value F_s, and outputs the optimized production scheduling S_opt. The iterative optimization module collects the actual electricity price C_act and the actual average energy consumption E_avg of the implemented optimized production scheduling S_opt, constructs an optimization effect evaluation algorithm, calculates the optimization effect evaluation value O_score, compares it with the optimization effect threshold O_th, executes the iterative optimization strategy according to the comparison result, and stores the data participating in the algorithm calculation in the historical data set D_his.

[0015] The present invention provides an enterprise power energy monitoring and analysis method and system based on the industrial Internet, having the following beneficial effects: (1) By collecting operation power data and production management system data in real time, and performing data cleaning, timestamp alignment, and Z-Score standardization processing, a unified historical dataset D_his is constructed, solving the problem of fragmentation of multi-source heterogeneous data, and realizing cross-dimensional correlation analysis of features such as the main frequency amplitude of the current spectrum FFT(I_t) at time t, the root mean square value of voltage RMS(V_t) at time t, and the rate of change of power factor ΔPF_t at time t. By dynamically generating the energy consumption prediction value E_pred based on the energy consumption prediction baseline E_base and the energy consumption prediction model, combining the time-of-use electricity price C_gri of the power grid to construct a reward and punishment function, comparing the reward and punishment function value R with the threshold R_th in real time, and linking the state of charge SOC of the energy storage to generate the power supply mode decision M_pow, a closed-loop automatic response from prediction to control is realized, avoiding decision-making lag caused by relying on manual experience. By taking the total energy consumption cost TEC and the scheduling stability SSP as the optimization objectives to output the optimized production schedule S_opt, and calculating the optimization effect evaluation value O_score based on the actual electricity price C_act and the actual average energy consumption E_avg, the model parameters are dynamically corrected to ensure that the energy consumption prediction and the power supply strategy are long-term consistent with the actual working conditions, overcoming the problem of the prediction model failure caused by equipment aging or process change. Finally, while ensuring production continuity, the energy consumption cost is significantly reduced.

[0016] (2) By using the power data acquisition device to collect the operation power data of industrial equipment in real time, synchronously integrating the order scheduling S_ord, the time-of-use electricity price C_gri of the power grid, and the state of charge SOC of the energy storage, and using data cleaning, timestamp alignment, and Z-Score standardization processing, a unified historical dataset D_his is constructed, solving the problem of fragmentation of multi-source heterogeneous data in traditional systems, and realizing the cross-dimensional deep integration of key features such as the main frequency amplitude of the current spectrum FFT(I_t) at time t, the root mean square value of voltage RMS(V_t) at time t, and the rate of change of power factor ΔPF_t at time t with production scheduling and electricity price information. Based on the average actual energy consumption μE_his(t) and the rate of change of power factor ΔPF_t in the same time period in the historical dataset D_his, the energy consumption prediction baseline E_base is dynamically established, and combined with the main frequency amplitude of the current spectrum FFT(I_t) at time t and the mean value of the main frequency amplitude of the current spectrum μFFT(I_t) in the historical dataset D_his in the same time period, an energy consumption prediction model is constructed to calculate the energy consumption prediction value E_pred, breaking through the limitation that the traditional static prediction model cannot correlate with the dynamic characteristics of equipment conditions, providing high-confidence data support for the subsequent closed-loop linkage of the real-time reward and punishment function and the power supply strategy, and improving the prediction accuracy and response timeliness.

[0017] (3) By constructing a reward and punishment function based on the predicted energy consumption value E_pred and the time-of-use electricity price C_gri of the power grid, dynamically calculating the reward and punishment function value R through the energy consumption prediction confidence E, and comparing it with the preset reward and punishment threshold R_th in real time, and generating a power supply mode decision M_pow in combination with the state of charge SOC of the energy storage, the transformation of the power supply strategy from "static preset" to "dynamic response" is realized, solving the problems of electricity bill waste and over-discharge risk of the energy storage caused by the lag of traditional manual decision-making. Through the multi-objective scheduling optimization algorithm, the predicted energy consumption value E_pred, the power supply mode decision M_pow and the order scheduling S_ord are linked. With the total energy consumption cost TEC and the scheduling stability SSP as the dual objectives, the optimized production scheduling S_opt is output, ensuring the stability of the order execution timing while reducing the electricity cost, and avoiding production interruption caused by electricity price fluctuations or power supply strategy adjustments. By collecting the actual electricity price C_act and the actual average energy consumption E_avg of the optimized scheduling, calculating the optimization effect evaluation value O_score, and comparing it with the optimization effect evaluation threshold O_th, the model weight coefficient is dynamically corrected to form a "prediction-optimization-feedback" closed loop, ensuring that the model can adapt to dynamic working conditions such as equipment aging or process changes in the long term, and overcoming the problem of strategy failure caused by the lack of self-adaptability of traditional prediction models. It realizes a full-link automation closed loop from real-time decision-making to continuous optimization, significantly reducing the total energy consumption cost on the basis of ensuring production continuity. Description of the Drawings

[0018] Figure 1 It is a schematic diagram of the steps of an enterprise electric power energy monitoring and analysis method based on the industrial Internet according to the present invention; Figure 2 It is a schematic block diagram of an enterprise electric power energy monitoring and analysis system based on the industrial Internet according to the present invention; Figure 3 It is a schematic diagram of the power supply decision data processing flow. Detailed Embodiments

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. Embodiment

[0020] The present invention provides an enterprise electric power energy monitoring and analysis method based on the industrial Internet. Please refer to Figure 1 , including the following steps: S1. Real-time collect the operating power data of industrial equipment through power data acquisition devices, and synchronize the order scheduling S_ord, grid time-of-use electricity price C_gri, and energy storage charge state SOC in the production management system data in real time. Preprocess the operating power data and production management system data and extract the operating power data features. Construct the current power feature vector Vec, and store the current power feature vector Vec and production management system data in the historical data set D_his; S2. Establish an energy consumption prediction baseline E_base based on the historical data set D_his, construct an energy consumption prediction model based on the energy consumption prediction baseline E_base and the current power feature vector Vec, and obtain the energy consumption prediction value E_pred through the energy consumption prediction model; S3. Construct a reward and punishment function based on the energy consumption prediction value E_pred and the grid time-of-use electricity price C_gri, calculate the reward and punishment function value R of the current strategy and compare it with the preset reward and punishment threshold R_th, and generate a power supply mode decision M_pow according to the comparison result; S4. Construct a multi-objective scheduling optimization algorithm based on the energy consumption prediction value E_pred, power supply mode decision M_pow, and order scheduling S_ord, calculate the multi-objective scheduling optimization value F_s and output the optimized production schedule S_opt; S5. Collect the actual electricity price C_act and actual average energy consumption E_avg of the optimized production schedule S_opt, construct an optimization effect evaluation algorithm, calculate the optimization effect evaluation value O_score and compare it with the optimization effect threshold O_th, and execute the iterative optimization strategy according to the comparison result and store the data participating in the algorithm calculation in the historical data set D_his.

[0021] In this embodiment, first, the operating power data of industrial equipment is collected in real time by sensors, and the order scheduling S_ord of the production management system, the time-of-use grid electricity price C_gri, and the state of charge SOC of the energy storage are synchronously integrated. After data cleaning, timestamp alignment, and Z-Score standardization processing, a unified historical dataset D_his is constructed, which solves the problem of the fragmentation of electrical parameters such as current and voltage from production plans and electricity price data in traditional systems, and realizes the cross-dimensional correlation of the main frequency amplitude FFT(I_t) of the current spectrum at time t, the root mean square value RMS(V_t) of the voltage at time t, the rate of change of power factor ΔPF_t with the order scheduling S_ord, the time-of-use electricity price C_gri, and the state of charge SOC of the energy storage, providing a complete data basis for subsequent analysis. Second, based on the average actual energy consumption μE_his(t) and the rate of change of power factor ΔPF_t in the same time period in the historical dataset D_his, an energy consumption prediction baseline E_base is dynamically established. Combining the main frequency amplitude FFT(I_t) of the current spectrum at time t with the mean value μFFT(I_t) of the main frequency amplitude of the current spectrum in the same time period in the historical dataset D_his, an energy consumption prediction model is constructed to generate an energy consumption prediction value E_pred, breaking through the limitation of traditional static models that cannot capture the dynamic operating conditions of equipment and providing high-confidence support for subsequent strategies. The reward and punishment function R is used to compare the reward and punishment threshold R_th in real time, and the state of charge SOC_t of the energy storage at time t is linked to generate the power supply mode decision M_pow(t) at time t. When the value of the reward and punishment function R ≥ the reward and punishment threshold R_th and the state of charge SOC_t of the energy storage at time t ≥ 1.5 times the minimum safe state of charge SOC_min of the energy storage, the energy storage power supply is automatically switched, solving the problems of electricity waste caused by lagging manual decision-making and the risk of over-discharging of the energy storage. The production schedule S_opt is optimized with the total energy consumption cost TEC and the schedule stability SSP as the dual objectives, ensuring the temporal stability of order execution through SSP while reducing electricity costs. Finally, in step S5, the model weight coefficients are dynamically corrected through the optimization effect evaluation value O_score to form a "prediction-optimization-feedback" closed loop, ensuring that the system can adapt to equipment aging or process changes in the long term and overcoming the strategy failure caused by the lack of self-adaptability of traditional models. Through the above mechanism, a full-link closed loop from data integration to dynamic optimization is realized, which has significant advantages in eliminating data islands, improving decision-making timeliness, and enhancing strategy robustness. Embodiment

[0022] This embodiment is an explanatory description carried out in Embodiment 1. Please refer to Figure 1 , specifically: S1 includes S11 and S12; S11, collect the operating power data of industrial equipment in real time through the power data collection equipment, the operating power data includes the real-time current data I_t at time t, the real-time voltage data V_t at time t and the power factor PF_t at time t, and the order schedule S_ord of the production management system, the time-of-use electricity price C_gri of the power grid and the energy storage charge state SOC_t at time t in real time synchronization; Among them, the real-time current data I_t is obtained through the current sensor, the real-time voltage data V_t is obtained through the smart meter, the power factor PF_t at time t is obtained through the power quality analyzer, and the actual power consumption E_act(t) at time t is obtained through the energy consumption calculation formula. The order schedule S_ord includes the preset order number O_ID, order quantity O_qua, planned start time T_sta, planned end time T_end and order priority P_prio. The grid time-of-use electricity price C_gri is obtained from the power company by using the API interface, and the energy storage charge state SOC_t at time t is obtained through the bus protocol of the BMS.

[0023] S12. Perform data cleaning on the operating power data and the production management system data, remove data outliers and align timestamps of different devices, and perform Z-Score standardization on the operating power data and the production management system data after removing data outliers; The frequency domain features of the real-time current data I_t at time t in the operating power data are extracted by Fourier transform, and the main frequency amplitude FFT (I_t) of the current spectrum at time t is obtained. The time domain features of the real-time voltage data V_t at time t and the power factor PF_t at time t in the operating power data are extracted by calculation, and the voltage effective value RMS (V_t) at time t and the power factor change rate △PF_t at time t are obtained. Based on the operating power data after feature extraction, the current power feature vector Vec = {FFT (I_t), RMS (V_t), △PF_t, E_act (t)} is constructed, and the current power feature vector Vec and the production management system data are stored in the historical data set D_his; Among them, the calculation formula of the voltage effective value RMS (V_t) at time t is as follows: ; In the formula, N represents the number of sampling points in a single time window, k represents the sampling point traversal index value, V_t(k) represents the real-time voltage data of the kth sampling point at time t, and the voltage RMS (V_t) at time t can accurately reflect the impact of voltage fluctuations on energy consumption and avoid the interference of instantaneous voltage abnormal values on the model; Among them, the power factor change rate △PF_t at time t is calculated as follows: ; Where, △t represents the time sliding window, and the power factor change rate △PF_t at time t quantifies the dynamic characteristics of the power factor, providing a trend input for the prediction model.

[0024] In this embodiment, the current sensor, smart meter, and power quality analyzer are used to accurately collect the current data I_t at time t, the voltage data V_t at time t, and the power factor PF_t at time t, respectively. The order scheduling S_ord, the grid time-of-use electricity price C_gri, and the state of charge SOC_t of the energy storage at time t are synchronously integrated. After data cleaning, timestamp alignment, and Z-Score standardization processing, the interference of abnormal data and dimensional differences on model training is eliminated, and a unified historical data set D_his is constructed, solving the problems of heterogeneous data formats and misaligned time series caused by independent deployment of sensors in traditional systems. By performing Fourier transform on the current data I_t at time t, the main frequency amplitude FFT(I_t) of the current spectrum at time t is obtained. Combining time-domain calculations, the root mean square value RMS(V_t) of the voltage at time t and the power factor change rate △PF_t at time t are obtained, and a multi-dimensional power feature vector Vec={FFT(I_t), RMS(V_t), ΔPF_t, E_act(t)} is constructed. The main frequency amplitude FFT(I_t) of the current spectrum at time t can capture the periodic fluctuations of the device load, and the power factor change rate △PF_t at time t reflects the dynamic change trend of the power factor. The combination of the two provides a fine-grained characterization of the device operating conditions for the energy consumption prediction model, breaking through the limitations of traditional single-feature analysis. Embodiment

[0025] This embodiment is an explanatory description carried out in Embodiment 2. Please refer to Figure 1 and Figure 3 , specifically: S2 includes S21 and S22; S21. Based on the actual power consumption E_act(t) at time t in the historical data set D_his, the average actual power consumption μE_his(t) in the same time period in the historical data is obtained, and the energy consumption prediction baseline E_base(t + △t) at time t + △t is established by combining the power factor change rate △PF_t at time t; Among them, the formula for the energy consumption prediction baseline E_base(t + △t) at time t + △t is as follows: ; In the formula, w1 represents the power factor influence factor set based on historical data.

[0026] S22. Based on the energy consumption prediction baseline E_base(t + △t) at time t + △t, the main frequency amplitude FFT(I_t) of the current spectrum at time t in the current power feature vector Vec, the root mean square value RMS(V_t) of the voltage at time t, the mean value μFFT(I_t) of the main frequency amplitude of the current spectrum in the same time period in the historical dataset D_his, and the mean value μRMS(V_t) of the root mean square value of the voltage in the same time period, construct an energy consumption prediction model, and obtain the energy consumption prediction value E_pred(t + △t) at time t + △t through the energy consumption prediction model; Among them, the formula of the energy consumption prediction model is as follows: ; ; In the formula, ECC represents the energy consumption correction term, and w2 represents the preset frequency domain correction coefficient.

[0027] S3 includes S31 and S32; S31. Based on the energy consumption prediction value E_pred(t + △t) at time t + △t and the time-of-use grid electricity price C_gri(t + △t) at time t + △t, construct a reward and punishment function, and calculate the reward and punishment function value R of the current strategy; Among them, based on the predicted energy consumption prediction value Ehis_pred(t1) at time t1 in the historical database and the actual power energy consumption E_act(t1) at time t1 in the historical dataset D_his, calculate the energy consumption prediction confidence E. The calculation formula of the energy consumption prediction confidence E is as follows: ; The formula of the reward and punishment function is as follows: ; In the formula, C_base represents the base electricity price, α represents the preset weight coefficient of the electricity price cost item, and β represents the preset weight coefficient of the prediction error penalty.

[0028] S32. Compare the reward and punishment function value R of the current strategy with the preset reward and punishment threshold R_th, and compare the state of charge SOC_t of the energy storage at time t with the set minimum safe state of charge SOC_min of the energy storage. Generate the power supply mode decision M_pow(t) at time t according to the comparison result; If the reward and punishment function value R of the current strategy ≥ the reward and punishment threshold R_th and the state of charge SOC_t of the energy storage at time t ≥ 1.5 * the minimum safe state of charge SOC_min of the energy storage, then the power supply mode decision M_pow(t) at time t is to use the energy storage for power supply; If the reward and punishment function value R of the current policy < R_th (the reward and punishment threshold) or the state of charge of the energy storage SOC_t at time t < 1.5 * SOC_min (the minimum safe state of charge of the energy storage), then the power supply mode decision M_pow(t) at time t is to use the mains power supply.

[0029] Specific examples of the power supply mode decision M_pow(t) at time t: The parameters are set as follows: The average actual energy consumption μE_his(t) in the same time period in the historical data set D_his: 0.5; The rate of change of power factor ΔPF_t: 0.02; The power factor influence factor w1 set based on historical data: 0.1; Substitute into the calculation formula of the energy consumption prediction baseline E_base(t + Δt) at time t + Δt: ; The main frequency amplitude FFT(I_t) of the current spectrum at time t: 0.12; The effective voltage value RMS(v_t) at time t: 0.22; The average value of the main frequency amplitude of the current spectrum μFFT(I_t) in the historical data set D_his in the same time period: 0.1; The effective voltage value μRMS(v_t) in the same time period: 0.23; The frequency domain correction coefficient w2: 0.5; Substitute into the energy consumption prediction model: ; ; The time-of-use electricity price of the power grid C_gri(t + Δt) at time t + Δt: 0.35; The base electricity price C_base: 0.15; The actual power energy consumption E_act(t1) at time t1 in the historical data set D_his: 0.6; The weight coefficients: α: 0.6, β: 0.4; ; The preset reward and punishment threshold R_th: 0.02, the current state of charge of the energy storage SOC_t: 75%, the minimum safe state of charge of the energy storage SOC_min: 15%; If the reward and punishment function value R > the reward and punishment threshold R_th and the current state of charge of the energy storage SOC_t > 1.5 times the minimum safe state of charge of the energy storage SOC_min, then the power supply mode decision M_pow(t) at time t is to use the energy storage for power supply.

[0030] In this embodiment, based on the average actual energy consumption μE_his(t) in the same time period in the historical data set D_his and the power factor change rate ΔPF_t at time t, an energy consumption prediction baseline E_base(t + Δt) at time t + Δt is dynamically established. The influence of the dynamic change of the power factor on energy consumption is quantified by the weight coefficient w1, which solves the defect that the traditional baseline only depends on the historical mean and ignores the working condition fluctuations. Secondly, an energy consumption prediction model is constructed by combining the main frequency amplitude FFT(I_t) of the current spectrum at time t and the mean value μFFT(I_t) of the main frequency amplitudes of the current spectra in the same time period in the historical data set D_his. The energy consumption prediction value E_pred(t + Δt) at time t + Δt is dynamically adjusted by the frequency domain correction coefficient w2 to accurately capture the periodic fluctuations of the equipment load and break through the limitation of the traditional time domain model's insufficient response to transient working conditions. The electricity price cost item and the prediction error penalty item are dynamically weighted by the reward and punishment function R, and combined with the energy consumption prediction confidence E to quantify the model credibility, realizing the balance between the economy and reliability of the power supply strategy. Further, based on the real-time comparison between the state of charge SOC_t of the energy storage at time t and 1.5 times the minimum safe state of charge SOC_min of the energy storage, a power supply mode decision M_pow(t) is generated. When the value of the reward and punishment function R ≥ the reward and punishment threshold R_th and the state of charge SOC_t of the energy storage is sufficient at time t, the energy storage power supply is preferentially enabled to avoid the sharp increase in costs caused by relying on the mains power during high electricity price periods. At the same time, the safety redundancy is determined by 1.5 times the minimum safe state of charge SOC_min of the energy storage to prevent over-discharge of the energy storage, solving the problem that it is difficult to balance economy and safety in traditional strategies, realizing the closed-loop optimization from dynamic prediction to safety control, solving the problem of strategy deviation caused by static parameters in traditional models, and significantly improving the system adaptability under high-fluctuation working conditions through frequency-time domain feature fusion and multi-objective weight regulation. Embodiment

[0031] This embodiment is an explanatory description carried out in Embodiment 3. Please refer to Figure 1 , specifically: S4 includes S41; S41. Obtain the production scheduling time period set M by statistically analyzing the production scheduling time periods and assign a unique identifier to each time period M = {t_1, t_2,..., t_m}. Based on the energy consumption prediction value E_pred(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, the power supply mode decision M_pow(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, the grid time-of-use electricity price C_gri(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, and the order scheduling S_ord, construct a multi-objective scheduling optimization algorithm, calculate and traverse to find the maximum value of the multi-objective scheduling optimization value F_s, and output the parameters when the maximum value of the multi-objective scheduling optimization value F_s is obtained as the optimized production scheduling S_opt and output; Among them, the formula of the multi-objective scheduling optimization algorithm is as follows: ; ; ; ; In the formula, TEC represents the total energy consumption cost, SSP represents the scheduling stability, L represents the total number of orders, j represents the order number, t(j, new) represents the execution time of order j in the optimized scheduling, t(j, old) represents the execution time of order j in the original scheduling, T_total represents the total production cycle, δ represents the influence factor of the power supply mode decision M_pow(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M on the total energy consumption cost TEC, λ represents the weight coefficient of the scheduling stability SSP, and exp represents the natural exponential function; The purpose achieved by this formula: The total energy consumption cost TEC directly reflects the total energy consumption cost under the time-of-use electricity price through the product of the energy consumption prediction value E_pred(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M and the time-of-use electricity price C_gri(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, embeds the electricity price sensitivity into the scheduling optimization, the scheduling stability SSP converts the time discreteness into a continuous stability index, and the multi-objective scheduling optimization value F_s balances the cost and stability through the scheduling stability weight coefficient λ.

[0032] S5 includes S51; S51. Based on the actual electricity price C_act, the actual average energy consumption E_avg, and the predicted total energy consumption cost TEC of the optimized production scheduling S_opt implemented after collection, construct an optimization effect evaluation algorithm to calculate the optimization effect evaluation value O_score; Among them, the formula of the optimization effect evaluation algorithm is as follows: ; Compare the optimization effect evaluation value O_score with the optimization effect threshold O_th, execute the iterative optimization strategy according to the comparison result, and store the data participating in the algorithm calculation in the historical data set D_his; If the optimization effect evaluation value O_score ≤ the optimization effect threshold O_th, it is determined that the optimization is qualified and the current strategy is retained; If the optimization effect evaluation value O_score > the optimization effect threshold O_th, it is determined that the optimization is unqualified, and the model is retrained and the weight coefficients w1 and w2 of the prediction model and the weight coefficients α and β in the reward and punishment function are corrected.

[0033] In this embodiment: By constructing a multi-objective scheduling optimization algorithm, the total energy consumption cost TEC and the scheduling stability SSP are taken as dual objectives, and the comprehensive optimization value F_S is used as the decision basis to output the optimized production schedule S_opt. When the time-of-use electricity price C_gri(t_i) of the power grid at the time period t_i with the unique identifier i in the production schedule time period set M is relatively high, the system automatically reduces the energy consumption prediction value E_pred(t_i) at the time period t_i with the unique identifier i in the production schedule time period set M. At the same time, the scheduling stability is quantified by the scheduling stability SSP index to ensure that while reducing the total energy consumption cost TEC, unplanned downtime caused by order delays is avoided, solving the pain point that it is difficult to balance cost and efficiency in traditional single-objective optimization. Secondly, through the calculation of the optimization effect evaluation algorithm, the optimization effect threshold O_th is dynamically compared. If the optimization effect evaluation value O_score > the optimization effect threshold O_th, the model is retrained and the weight coefficients are corrected to form a "optimization - feedback - correction" closed loop. When high abnormal values appear in the actual electricity price C_act or the actual energy consumption E_avg and the predicted total energy consumption cost TEC, the optimization effect evaluation value O_score rises to trigger model parameter iteration, avoiding the long-term strategy failure caused by power grid price fluctuations or equipment aging, and breaking through the limitation of the lack of self-adaptability of traditional static optimization models. Embodiment

[0034] An enterprise electric power energy monitoring and analysis system based on the industrial Internet, please refer to Figure 2 , specifically: including a data acquisition and processing module, a prediction model construction module, a power supply mode decision-making module, an optimized production scheduling module, and an iterative optimization module; The data acquisition and processing module collects the operating power data of industrial equipment in real time through power data acquisition devices, synchronizes the order schedule S_ord, the time-of-use electricity price C_gri of the power grid, and the energy storage charge state SOC in the production management system data in real time, preprocesses the operating power data and the production management system data, extracts the operating power data features, constructs the current power feature vector Vec, and stores the current power feature vector Vec and the production management system data in the historical data set D_his; The prediction model construction module establishes an energy consumption prediction baseline E_base based on the historical data set D_his, constructs an energy consumption prediction model based on the energy consumption prediction baseline E_base and the current power feature vector Vec, and obtains the energy consumption prediction value E_pred through the energy consumption prediction model; The power supply mode decision-making module constructs a reward and punishment function based on the energy consumption prediction value E_pred and the time-of-use electricity price C_gri, calculates the reward and punishment function value R of the current strategy, compares it with the preset reward and punishment threshold R_th, and generates a power supply mode decision M_pow according to the comparison result; The optimized production scheduling module constructs a multi-objective scheduling optimization algorithm based on the energy consumption prediction value E_pred, the power supply mode decision M_pow, and the order scheduling S_ord, calculates the multi-objective scheduling optimization value F_s, and outputs the optimized production scheduling S_opt. The iterative optimization module collects the actual electricity price C_act and the actual average energy consumption E_avg of the optimized production scheduling S_opt, constructs an optimization effect evaluation algorithm, calculates the optimization effect evaluation value O_score, compares it with the optimization effect threshold O_th, executes the iterative optimization strategy according to the comparison result, and stores the data participating in the algorithm calculation in the historical data set D_his.

[0035] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An enterprise power energy monitoring and analysis method based on the industrial Internet, characterized in that: It includes the following steps: S1. Real-time collect the operating power data of industrial equipment through power data acquisition equipment, and real-time synchronize the order scheduling S_ord, grid time-of-use electricity price C_gri, and energy storage charge state SOC in the production management system data. Preprocess the operating power data and production management system data and extract the operating power data features. Construct the current power feature vector Vec, and store the current power feature vector Vec and production management system data into the historical data set D_his; S2. Establish an energy consumption prediction baseline E_base based on the historical data set D_his, construct an energy consumption prediction model based on the energy consumption prediction baseline E_base and the current power feature vector Vec, and obtain the energy consumption prediction value E_pred through the energy consumption prediction model; S3. Construct a reward and punishment function based on the energy consumption prediction value E_pred and the grid time-of-use electricity price C_gri, calculate the reward and punishment function value R of the current strategy and compare it with the preset reward and punishment threshold R_th, and generate a power supply mode decision M_pow according to the comparison result; S4. Construct a multi-objective scheduling optimization algorithm based on the energy consumption prediction value E_pred, power supply mode decision M_pow, and order scheduling S_ord, calculate the multi-objective scheduling optimization value F_s and output the optimized production schedule S_opt; S5. Collect the actual electricity price C_act and actual average energy consumption E_avg of the implemented optimized production schedule S_opt and construct an optimization effect evaluation algorithm, calculate the optimization effect evaluation value O_score and compare it with the optimization effect threshold O_th, and execute the iterative optimization strategy according to the comparison result and store the data participating in the algorithm calculation into the historical data set D_his.

2. The enterprise electric power energy monitoring and analysis method based on industrial Internet according to claim 1, characterized in that: S1 includes S11 and S12; S11. Real-time collect the operating power data of industrial equipment through power data acquisition equipment. The operating power data includes the real-time current data I_t at time t, the real-time voltage data V_t at time t, and the power factor PF_t at time t. Real-time synchronize the order scheduling S_ord of the production management system, the grid time-of-use electricity price C_gri, and the energy storage charge state SOC_t at time t; Among them, the real-time current data I_t is obtained through a current sensor, the real-time voltage data V_t is obtained through a smart meter, the power factor PF_t at time t is obtained through a power quality analyzer, and the actual power consumption E_act(t) at time t is obtained through the energy consumption calculation formula. The order scheduling S_ord includes the preset order number O_ID, order quantity O_qua, planned start time T_sta, planned end time T_end, and order priority P_prio. The grid time-of-use electricity price C_gri is obtained from the power company by using the API interface, and the energy storage charge state SOC_t at time t is obtained through the bus protocol of the BMS.

3. The method for monitoring and analyzing enterprise electric power energy based on industrial Internet according to claim 2, wherein: S12. Clean the operation power data and production management system data, remove data outliers, align the timestamps of different devices, and perform Z-Score standardization on the operation power data and production management system data after removing data outliers; Extract the frequency-domain features of the real-time current data \(I_t\) at time \(t\) from the operation power data through Fourier transform, obtain the main frequency amplitude \(FFT(I_t)\) of the current spectrum at time \(t\). Calculate and extract the time-domain features of the real-time voltage data \(V_t\) and the power factor \(PF_t\) at time \(t\) from the operation power data, obtain the root mean square voltage \(RMS(V_t)\) at time \(t\) and the power factor change rate \(\Delta PF_t\) at time \(t\). Based on the operation power data after feature extraction, construct the current power feature vector \(Vec = \{FFT(I_t), RMS(V_t), \Delta PF_t, E_{act}(t)\}\), and store the current power feature vector \(Vec\) and the production management system data in the historical dataset \(D_{his}\); Among them, the formula for calculating the root mean square voltage \(RMS(V_t)\) at time \(t\) is as follows: ； In the formula, \(N\) represents the number of sampling points within a single time window, \(k\) represents the sampling point traversal index value, and \(V_t(k)\) represents the real-time voltage data of the \(k\)-th sampling point at time \(t\); Among them, the formula for calculating the power factor change rate \(\Delta PF_t\) at time \(t\) is as follows: ； In the formula, \(\Delta t\) represents the time sliding window.

4. A method for monitoring and analyzing enterprise electric power energy based on industrial Internet according to claim 1, characterized in that: S2 includes S21 and S22; S21. Based on the actual power consumption \(E_{act}(t)\) at time \(t\) in the historical dataset \(D_{his}\), obtain the average actual power consumption \(\mu E_{his}(t)\) in the same time period in the historical data, and establish an energy consumption prediction baseline \(E_{base}(t + \Delta t)\) at time \(t+\Delta t\) in combination with the power factor change rate \(\Delta PF_t\) at time \(t\); Among them, the formula for the energy consumption prediction baseline \(E_{base}(t + \Delta t)\) at time \(t+\Delta t\) is as follows: ； In the formula, \(w1\) represents the power factor influence factor set based on historical data.

5. A method for monitoring and analyzing enterprise electric power energy based on industrial Internet according to claim 4, characterized in that: S22. Based on the energy consumption prediction baseline \(E_{base}(t + \Delta t)\) at time \(t+\Delta t\), the main frequency amplitude \(FFT(I_t)\) of the current spectrum at time \(t\) in the current power feature vector \(Vec\), the root mean square voltage \(RMS(V_t)\) at time \(t\), the average main frequency amplitude \(\mu FFT(I_t)\) of the current spectrum in the same time period in the historical dataset \(D_{his}\), and the average root mean square voltage \(\mu RMS(V_t)\) in the same time period, construct an energy consumption prediction model, and obtain the energy consumption prediction value \(E_{pred}(t + \Delta t)\) at time \(t+\Delta t\) through the energy consumption prediction model; Among them, the formula for the energy consumption prediction model is as follows: ； ； In the formula, \(ECC\) represents the energy consumption correction term, and \(w2\) represents the preset frequency-domain correction coefficient.

6. The enterprise electric power energy monitoring and analysis method based on industrial Internet according to claim 5, characterized in that: S3 includes S31 and S32; S31. Construct a reward and punishment function based on the predicted energy consumption value E_pred(t + △t) at time t + △t and the time-of-use electricity price C_gri(t + △t) of the power grid at time t + △t, and calculate the reward and punishment function value R of the current strategy; Among them, calculate the energy consumption prediction confidence E based on the predicted energy consumption prediction value Ehis_pred(t1) in the historical database at time t1 and the actual power energy consumption E_act(t1) at time t1 in the historical dataset D_his. The calculation formula for the energy consumption prediction confidence E is as follows: ； The formula for the reward and punishment function is as follows: ； In the formula, C_base represents the base electricity price, α represents the preset weight coefficient of the electricity price cost item, and β represents the preset weight coefficient of the prediction error penalty.

7. According to an enterprise electric power energy monitoring and analysis method based on industrial Internet according to claim 6, it is characterized in that: S32. Compare the reward and punishment function value R of the current strategy with the preset reward and punishment threshold R_th, and compare the state of charge SOC_t of the energy storage at time t with the set minimum safe state of charge SOC_min of the energy storage. Generate a power supply mode decision M_pow(t) at time t according to the comparison result; If the reward and punishment function value R of the current strategy ≥ the reward and punishment threshold R_th and the state of charge SOC_t of the energy storage at time t ≥ 1.5 * the minimum safe state of charge SOC_min of the energy storage, then the power supply mode decision M_pow(t) at time t is to use energy storage for power supply; If the reward and punishment function value R of the current strategy < the reward and punishment threshold R_th or the state of charge SOC_t of the energy storage at time t < 1.5 * the minimum safe state of charge SOC_min of the energy storage, then the power supply mode decision M_pow(t) at time t is to use mains power for power supply.

8. A method for monitoring and analyzing enterprise electric power energy based on industrial Internet according to claim 7, characterized in that: S4 includes S41; S41. Obtain the production scheduling time period set M by counting the production scheduling time periods and assign a unique identifier to each time period M = {t_1, t_2,..., t_m}. Based on the predicted energy consumption value E_pred(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, the power supply mode decision M_pow(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, the time-of-use electricity price C_gri(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M, and the order scheduling S_ord, construct a multi-objective scheduling optimization algorithm, calculate and traverse to find the maximum value of the multi-objective scheduling optimization value F_s, and output the parameters when the maximum value of the multi-objective scheduling optimization value F_s is obtained as the optimized production scheduling S_opt and output; Among them, the formula for the multi-objective scheduling optimization algorithm is as follows: ； ； ； ； In the formula, TEC represents the predicted total energy consumption cost, SSP represents the scheduling stability, L represents the total number of orders, j represents the order number, t(j, new) represents the execution time of order j in the optimized schedule, t(j, old) represents the execution time of order j in the original schedule, T_total represents the total production cycle, δ represents the influencing factor of the power supply mode decision M_pow(t_i) of the time period t_i with the unique identifier i in the production scheduling time period set M on the total energy consumption cost TEC, λ represents the weight coefficient of the scheduling stability SSP, and exp represents the natural exponential function.

9. A method for monitoring and analyzing enterprise electric power energy based on industrial Internet according to claim 8, characterized in that: S5 includes S51; S51. Construct an optimization effect evaluation algorithm based on the actual electricity price C_act, the actual average energy consumption E_avg, and the predicted total energy consumption cost TEC of the optimized production schedule S_opt implemented after collection, and calculate the optimization effect evaluation value O_score; Among them, the formula of the optimization effect evaluation algorithm is as follows: ； Compare the optimization effect evaluation value O_score with the optimization effect threshold O_th, execute the iterative optimization strategy according to the comparison result, and store the data participating in the algorithm calculation in the historical data set D_his; If the optimization effect evaluation value O_score ≤ the optimization effect threshold O_th, it is determined that the optimization is qualified, and the current strategy is retained; If the optimization effect evaluation value O_score > the optimization effect threshold O_th, it is determined that the optimization is unqualified, and the model is retrained, and the weight coefficients w1 and w2 of the prediction model and the weight coefficients α and β in the reward and punishment function are corrected.

10. An enterprise power energy monitoring and analysis system based on the industrial Internet, which is applied to an enterprise power energy monitoring and analysis method based on the industrial Internet according to any one of claims 1 to 9, and is characterized in that: It includes a data acquisition and processing module, a prediction model construction module, a power supply mode decision module, an optimized production scheduling module, and an iterative optimization module; The data acquisition and processing module collects the operating power data of industrial equipment in real time through a power data acquisition device, synchronizes the order scheduling S_ord, the grid time-of-use electricity price C_gri, and the energy storage charge state SOC in the production management system data in real time, preprocesses the operating power data and the production management system data, extracts the operating power data features, constructs the current power feature vector Vec, and stores the current power feature vector Vec and the production management system data in the historical data set D_his; The prediction model construction module establishes an energy consumption prediction baseline E_base based on the historical data set D_his, constructs an energy consumption prediction model based on the energy consumption prediction baseline E_base and the current power feature vector Vec, and obtains the energy consumption prediction value E_pred through the energy consumption prediction model; The power supply mode decision module constructs a reward and punishment function based on the energy consumption prediction value E_pred and the grid time-of-use electricity price C_gri, calculates the reward and punishment function value R of the current strategy, compares it with the preset reward and punishment threshold R_th, and generates a power supply mode decision M_pow according to the comparison result; The optimized production scheduling module constructs a multi-objective scheduling optimization algorithm based on the energy consumption prediction value E_pred, the power supply mode decision M_pow, and the order scheduling S_ord, calculates the multi-objective scheduling optimization value F_s, and outputs the optimized production schedule S_opt; The iterative optimization module calculates the optimization effect evaluation value O_score by collecting the actual electricity price C_act and the actual average energy consumption E_avg of the optimized production schedule S_opt, constructing an optimization effect evaluation algorithm, comparing the O_score with the optimization effect threshold O_th, executing the iterative optimization strategy according to the comparison result, and storing the data participating in the algorithm calculation in the historical dataset D_his.

Images