A power grid energy management optimization platform based on big data

Abstract

The present invention discloses an optimized platform for power grid energy management based on big data, which relates to the technical field of power grid energy management and includes the following steps: First, deploy real-time power monitoring devices within areas with concentrated high-tech industries, and obtain real-time power consumption data of enterprises and key power-consuming devices through the power monitoring devices. Establish a data interface to allow enterprises to independently upload the power consumption data of internal key devices, improving the comprehensiveness and accuracy of the data. Through real-time power monitoring, feature extraction, and evaluation of machine learning models, the solution of the present invention realizes precise monitoring and intelligent scheduling of power consumption. For normal and surging power demands, ensure power supply stability and efficient utilization of resources respectively through load forecasting and distributed optimization algorithms, avoiding risks such as network paralysis and power outages caused by sudden power demands, and ensuring enterprise operation and power grid safety.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid energy management, and particularly to an optimization platform for power grid energy management based on big data. Background Art

[0002] An optimization platform for power grid energy management based on big data is a solution that uses large-scale data processing technology and combines intelligent algorithms to achieve efficient management and optimization of the power grid system. The platform comprehensively understands the operating state of the power grid by collecting, analyzing, and processing various data during the operation of the power grid (such as load, energy consumption, meteorological data, user electricity consumption behavior, etc.). With the help of big data analysis, the platform can accurately predict power demand, optimize the allocation of power resources, ensure the stable operation of the power supply system, improve energy utilization efficiency, and reduce energy waste. In addition, the platform can also identify potential problems and hidden dangers through in-depth mining of historical data, perform predictive maintenance in advance, and avoid large-scale power outages caused by sudden failures.

[0003] The core of the platform is intelligent optimization algorithms and data-driven decision-making. It can not only flexibly adjust according to different electricity consumption scenarios but also dynamically optimize the load balance of the power grid to ensure real-time matching of power supply and demand. In the scenario of renewable energy access, the big data platform can also adjust the power generation and grid connection strategies of clean energy according to changes in external factors such as weather, sunlight, and wind speed, improve the utilization rate of new energy, and reduce the use of fossil energy. By intelligently and automatically managing the entire power grid system, the optimization platform for power grid energy management based on big data can not only effectively improve the safety and economy of the power grid but also lay a foundation for low-carbon and environmentally friendly energy development.

[0004] The existing technologies have the following deficiencies:

[0005] The existing power grids mainly rely on the prediction of electricity demand in each region for planning and scheduling because the production and consumption of electricity must be matched in time (large-scale electrical energy is difficult to store for a long time). Through the load prediction model, the power system anticipates future demand in advance and arranges the production and distribution of resources. However, in areas where high-tech industries are concentrated, when enterprises are hit by large-scale cyberattacks (such as distributed denial-of-service attacks DDoS or ransomware infections), the server clusters need to quickly expand to divert attack traffic or restore damaged data systems. Such sudden expansions and additional computing requirements may lead to a sharp increase in regional electricity consumption beyond the predicted range. If the power system fails to schedule power supply in time, enterprises will be unable to expand their servers to cope with the attacks, resulting in network paralysis, forcing the interruption of core businesses such as e-commerce, financial services, and cloud computing. Such interruptions will not only severely impact enterprise operations, causing huge economic losses, but also damage public trust, trigger customer claims and legal lawsuits, increase financial pressure, and further endanger the market competitiveness and viability of enterprises.

[0006] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The object of the present invention is to provide a power grid energy management optimization platform based on big data. Through real-time power monitoring, feature extraction, and evaluation of machine learning models, the solution achieves precise monitoring and intelligent scheduling of power consumption. For normal and surge power demands, power supply stability and efficient resource utilization are ensured respectively through load forecasting and distributed optimization algorithms, avoiding risks such as network paralysis and power outages caused by sudden power demands, and ensuring enterprise operation and power grid safety, so as to solve the problems in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions: A power grid energy management optimization platform based on big data, including the following steps:

[0009] First, deploy real-time power monitoring devices within areas concentrated with high-tech industries. Through the power monitoring devices, real-time power consumption data of enterprises and key power-consuming equipment are obtained in real time, and a data interface is established to allow enterprises to independently upload the power consumption data of internal key equipment, improving the comprehensiveness and accuracy of the data;

[0010] Extract features from the obtained power consumption data. Based on the extracted features of the power consumption data, under the monitoring window, use a pre-learned machine learning model to conduct intelligent evaluation on the power consumption data within the area;

[0011] Based on the evaluation results of the machine learning model, divide the regional power demand into normal power demand and surge power demand;

[0012] For normal power demand, supply power continuously and stably according to the established load forecasting plan to ensure the balance between power production and consumption and maintain the normal operation of the power grid;

[0013] For surge power demand, adopt a distributed optimization algorithm. Based on the established load forecasting plan, initiate intelligent power dispatching adjustment, and at the same time adjust the power supply distribution of the remaining areas to optimize the power grid load.

[0014] Preferably, feature extraction is performed on the obtained power consumption data. The extracted features include the degree of deviation of the power factor within a short period of time and the ratio of the instantaneous peak value to the average value of the power consumption within a short period of time. Under the detection window, the degree of deviation of the power factor within a short period of time and the ratio of the instantaneous peak value to the average value of the power consumption within a short period of time are analyzed, and a power factor deviation index and an instantaneous load fluctuation index are respectively generated. The power factor deviation index and the instantaneous load fluctuation index are input into a pre-trained machine learning model, and a power consumption coefficient is generated through the machine learning model. The power consumption coefficient is used to intelligently evaluate the change of regional power consumption.

[0015] Preferably, under the detection window, after analyzing the degree of deviation of the power factor within a short period of time, a power factor deviation index is generated. The specific steps are as follows:

[0016] Under the detection window, the instantaneous power factor is calculated from the instantaneous waveforms of current and voltage. The calculation expression is as follows: PF(t) = cos(E(t)) = cos(φ V (t) - φ I (t)), where PF(t) is the instantaneous power factor, representing the instantaneous power factor of the power system at time t. E(t) is the instantaneous phase angle difference, representing the phase difference between voltage and current at time t. cos(E(t)) represents the instantaneous power factor, φ V (t) is the instantaneous phase angle of voltage, and φI(t) is the instantaneous phase angle of current;

[0017] Next, a metric for quantifying the deviation of the power factor within a short period of time is defined, called the instantaneous deviation rate. The instantaneous deviation rate measures the deviation rate of the power factor within an extremely short period of time, and represents the change of the power factor over time in the form of a derivative. The calculation expression is as follows:

[0018]

[0019] , where ΔPF(t) is the instantaneous power factor deviation rate, representing the deviation speed of the instantaneous power factor PF(t) at time t, that is, the instantaneous power factor deviation rate of the instantaneous power factor, is the derivative of the instantaneous power factor with respect to time, is the calculation process of the instantaneous power factor deviation;

[0020] To capture the overall intensity of the power factor deviation, the instantaneous power factor deviation rate ΔPF(t) is weighted and integrated, and a time weight function is introduced. The calculation expression is as follows:

[0021]

[0022] , where PFShiftStrength is the power factor offset strength, t0 and t1 are the start time and end time of the detection window, W(t) is the time weight function,

[0023] Next, define a non - linear offset discrimination function to compare the offset degree of the power factor with its critical value, and generate a function for measuring the abnormal deviation degree. The calculation expression is as follows:

[0024]

[0025] , where f(PF) is the non - linear offset discrimination function, E ref is the reference phase angle, that is, the reference phase angle under normal conditions, β is the weight coefficient, k is the non - linear amplification coefficient, cos(E ref ) is the reference instantaneous power factor;

[0026] Generate a power factor offset index based on the non - linear offset discrimination function f(PF). The calculation expression is as follows: PFΔI = ln(1 + θ·f(PF)), where PFΔI is the power factor offset index, and θ is the amplification coefficient used to adjust the sensitivity and range of the index.

[0027] Preferably, under the detection window, analyze the ratio of the instantaneous peak value to the average value of the power consumption within a short period of time to generate an instantaneous load fluctuation index. The specific steps are as follows:

[0028] Under the detection window, obtain the sampling values of current and voltage in real - time, and calibrate the obtained current sampling value as I(t) and the voltage sampling value as V(t). I(t) refers to the current sampling value at time t, and V(t) refers to the voltage sampling value at time t;

[0029] Calculate the difference in the instantaneous power change at each time point t to reflect the sudden change trend of the equipment power. The calculation expression is as follows: ΔP(t)=|I(t)·V(t)-I(t - Δt)·V(t - Δt)|, where ΔP(t) is the instantaneous change in power, representing the instantaneous change in power at time point t, Δt is the sampling time interval used to calculate the change between two adjacent time points, I(t - Δt) is the current sampling value at the previous time point of time t, and V(t - Δt) is the voltage sampling value at the previous time point of time t;

[0030] To capture the cumulative fluctuation of the instantaneous change in power within the entire detection window, use the cumulative absolute change function to avoid the mutual cancellation of positive and negative power fluctuations. The calculation expression is as follows:

[0031]

[0032] , where C P is the cumulative value of power jumps within the detection window, representing the total change in power fluctuations within the detection window, M is the total number of sampling points within the detection window, reflecting the number of data points collected within the detection window;

[0033] To accurately capture the dynamic change trend of the load, an exponential decay weight function is introduced, which can respond more sensitively to the latest power fluctuations, and then generate an instantaneous load fluctuation index. The calculation expression is as follows:

[0034]

[0035] , where I-LFX is the instantaneous load fluctuation index, e is the natural base, β is the exponential decay coefficient of the instantaneous load fluctuation index, controlling the speed of exponential decay, and M-t represents the distance between the sampling point and the current time.

[0036] Preferably, the power consumption coefficient generated after analyzing the obtained power consumption data under the monitoring window is compared with the pre-set reference threshold of the power consumption coefficient to analyze and divide the regional power demand. The division results are as follows:

[0037] If the power consumption coefficient is greater than or equal to the pre-set reference threshold of the power consumption coefficient, the regional power demand is divided into a surging power demand;

[0038] If the power consumption coefficient is less than the pre-set reference threshold of the power consumption coefficient, the regional power demand is divided into a normal power demand.

[0039] Preferably, for the surging power demand, a distributed optimization algorithm is adopted. Based on the established load forecasting plan, intelligent power dispatching adjustment is started, and at the same time, the power supply distribution of the remaining areas is adjusted. The steps to optimize the grid load are as follows:

[0040] Within the high-tech enterprise area, the power factor deviation index PFΔI and the instantaneous load fluctuation index I-LFX are obtained in real time through power monitoring equipment. Once a load surge is detected, it is necessary to determine whether to start scheduling, and the following judgment conditions are used for judgment;

[0041]

[0042] , where TriggerAdjustment is the trigger variable, θ1 is the pre-set reference threshold of the power factor deviation index PFΔI, and θ2 is the pre-set reference threshold of the instantaneous load fluctuation index I-LFX;

[0043] When the scheduling is triggered, the optimal power distribution between regions is solved to meet the surging demand at the lowest cost. The calculation expression is as follows:

[0044]

[0045] , where min is to minimize the objective function, and C i is the unit power supply cost of area i, and P i is the actual power supply of area i, represents the total power supply cost of each area, λ is the penalty coefficient, is the predicted load of area i, is the penalty term for load prediction error, aiming to reduce the deviation between the actual power supply P i and the predicted power supply , and N represents the total number of areas participating in scheduling in the power grid system;

[0046] In the distributed optimization process, it is necessary to ensure the balance between power supply and demand. At the same time, the load of each area must meet the constraint conditions. Set the power balance constraint and area load constraint conditions. The power balance constraint expression is as follows:

[0047]

[0048] , where is the total power supply capacity in the system at the current moment;

[0049] The area load constraint expression is as follows: where is the minimum load of area i, is the maximum load of area i;

[0050] Since power dispatching involves multiple areas, the distributed Lagrangian relaxation method is used to decompose the global optimization problem into several sub-problems and solve them in parallel. The calculation expression is as follows:

[0051]

[0052] , where L(P i , μ) is the Lagrangian function, is the cost and prediction error part of the power grid system, calculating the power supply cost of all areas and their prediction load errors, is the power balance constraint term, and μ is the Lagrange multiplier;

[0053] After the dispatching plan is implemented, feedback is carried out according to real-time data, and the power supply strategy is dynamically adjusted to ensure the stable operation of the power grid. The calculation expression is as follows: where is the adjusted power supply of area i, α is the dispatching coefficient, used to control the adjustment intensity of the power supply of each area, and P t is the actual total load of the system at the current moment, is the predicted total load at the current moment.

[0054] In the above technical solution, the technical effects and advantages provided by the present invention are as follows:

[0055] Through real-time power monitoring and feature extraction, and combined with the evaluation ability of the machine learning model, the present invention realizes accurate monitoring and intelligent analysis of power consumption in the region. In particular, by analyzing the power factor deviation index and the instantaneous load fluctuation index, it is possible to accurately judge whether there is a sharp increase in power demand in the region. This ability enables the power system to quickly identify sudden power demands caused by cyberattacks or server expansions and respond in a timely manner, avoiding serious consequences such as network paralysis and service interruption caused by insufficient power supply. The intelligent evaluation mechanism of the solution also ensures dynamic adjustment of load forecasting, improves the flexibility of power management, enables the power grid to more accurately adapt to the rapid changes in enterprise needs, and maintains the stable development of the regional economy and industries.

[0056] The present invention formulates differential scheduling strategies for normal power demands and surging power demands. For normal demands, the power supply is maintained smoothly through the load forecasting plan to ensure the balance between power production and consumption, effectively avoiding power waste and systematic power shortages. In the case of surging demand scenarios, the solution uses a distributed optimization algorithm to achieve intelligent power scheduling adjustment. By optimizing the power supply distribution within the region and between adjacent regions, the load pressure on a single region is minimized, avoiding the risk of power outages caused by grid overload. This distributed intelligent scheduling not only improves the utilization efficiency of power resources but also ensures the emergency power supply requirements of enterprises in the event of sudden cyberattacks, provides stable power support for enterprises, and reduces operational risks and economic losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0058] Figure 1 It is a method flow chart of an optimization platform for power grid energy management based on big data according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] Now, example embodiments will be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that this disclosure will be more complete and comprehensive, and will fully convey the concept of the example embodiments to those skilled in the art.

[0060] The present invention provides as Figure 1A power grid energy management optimization platform based on big data is shown below, including the following steps:

[0061] First, deploy real-time power monitoring devices in areas with concentrated high-tech industries. Through the power monitoring devices, real-time power consumption data of enterprises and key power-consuming equipment are obtained in real time. A data interface is established to allow enterprises to independently upload the power consumption data of internal key equipment, improving the comprehensiveness and accuracy of the data;

[0062] In areas with concentrated high-tech industries, it is crucial to deploy real-time power monitoring devices. These devices include smart meters, high-precision current sensors, and data acquisition terminals, etc., which can accurately record the real-time power consumption data of enterprises and key power-consuming equipment. These devices are usually installed at the main power consumption nodes of enterprises, data center computer rooms, server clusters, production lines, and refrigeration systems, etc., to ensure comprehensive monitoring of all important power consumption points. Through high-bandwidth networks (such as 5G, fiber optic networks), these monitoring data will be transmitted to the regional power management center in real time, providing data support for the real-time analysis and scheduling of the system.

[0063] In addition, establishing a data interface to support enterprises to independently upload the power consumption data of key equipment can further improve the comprehensiveness and accuracy of the data. For example, the control systems of internal servers, cooling systems, or automated production equipment in enterprises can directly transmit the internal energy consumption information to the power management platform. This information sharing not only helps the power grid to timely understand the power consumption situation of each enterprise, but also can improve the accuracy of prediction and scheduling. When there is an abnormal power consumption in an enterprise (such as server expansion or emergency production tasks), the management platform can immediately perceive and quickly respond, ensuring that the power supply meets the demand and preventing systemic risks caused by emergencies.

[0064] Extract the characteristics of the obtained power consumption data. Based on the extracted characteristics of the power consumption data, under the monitoring window, use the pre-learned machine learning model to intelligently evaluate the power consumption data in the region;

[0065] Extract the characteristics of the obtained power consumption data. The extracted characteristics include the degree of deviation of the power factor in a short period of time and the ratio of the instantaneous peak value to the average value of the power consumption in a short period of time. Under the detection window, analyze the degree of deviation of the power factor in a short period of time and the ratio of the instantaneous peak value to the average value of the power consumption in a short period of time, respectively generate a power factor deviation index and an instantaneous load fluctuation index, input the power factor deviation index and the instantaneous load fluctuation index into the pre-learned machine learning model, generate a power consumption coefficient through the machine learning model, and use the power consumption coefficient to intelligently evaluate the change situation of the regional power consumption;

[0066] A sudden deviation of the power factor within a short period usually indicates a sharp increase in power consumption within the region. This is because the power factor reflects the ratio between the active power (the actual consumed energy) and the apparent power (the total transmitted energy). When a large number of high-power loads (such as server clusters, refrigeration systems, electric motors, etc.) are connected in a short period, these devices usually have strong inductive or capacitive characteristics, resulting in a rapid decrease or increase in the power factor, deviating from the normal level. The sudden startup of such loads will significantly increase the reactive power demand of the power system (the energy used to maintain the electromagnetic field), causing short-term fluctuations in the power factor of the power supply system. At the same time, this kind of load is usually accompanied by a large amount of power consumption. Therefore, the drastic change in the power factor is often an important signal of a sharp increase in power consumption within the region. The deviation of the power factor may also lead to a decrease in the transmission efficiency of the power grid, making more current need to pass through the transmission line to transmit the same power, exacerbating the grid load, further confirming the rapid growth of the regional power demand.

[0067] Under the detection window, after analyzing the deviation degree of the power factor within a short period, a power factor deviation index is generated. The specific steps are as follows:

[0068] Under the detection window, calculate the instantaneous power factor from the instantaneous waveforms of current and voltage. The calculation expression is as follows: PF(t) = cos(E(t)) = cos(φ V (t) - φ I (t), where PF(t) is the instantaneous power factor, indicating the instantaneous power factor of the power system at time t. The power factor reflects the ratio between the active power (effective power) and the apparent power (total power) in the power system. E(t) is the instantaneous phase angle difference, indicating the phase difference between voltage and current at time t. The phase difference reflects the lag or lead relationship between the current and voltage waveforms. cos(E(t)) represents the instantaneous power factor, which is the cosine value of the phase angle difference and represents the utilization efficiency of electric energy. φ V (t) is the instantaneous phase angle of voltage, indicating the instantaneous phase angle of the voltage waveform at time t. φ I (t) is the instantaneous phase angle of current, indicating the instantaneous phase angle of the current waveform at time t;

[0069] By analyzing the instantaneous waveforms of current and voltage, calculate the phase angle difference between them, and thus obtain the instantaneous power factor PF(t). The power factor reflects the operating efficiency of the power system. Monitoring its changes helps to adjust the power supply in a timely manner in case of emergencies (such as load fluctuations or equipment startup) to ensure the stability and efficiency of the system.

[0070] The phase angle is the angle corresponding to the time difference between the voltage waveform and the current waveform in an AC circuit, expressed in degrees or radians. Since voltage and current usually vary in the form of sine waves in an AC system, their characteristic points such as peaks and troughs may not be synchronized (i.e., not reach simultaneously). This non-synchronization results in a phase difference between the two, and the phase angle is the specific quantification of this phase difference. The sign of the phase angle is also significant: if the current lags behind the voltage, the phase angle is positive, usually occurring in inductive loads (such as motors); if the current leads the voltage, the phase angle is negative, commonly found in capacitive loads (such as capacitors). When the phase angle is 0°, the voltage and current are completely synchronized, indicating that there is no reactive power in the system and all electrical energy is effectively converted into active power. If the phase angle is close to ±90°, it indicates that there is a large amount of reactive power in the system and the power utilization efficiency is low. Monitoring the phase angle is very important for optimizing power utilization, reducing losses, and improving system stability.

[0071] Next, define an index to quantify the power factor offset in a short period, called the instantaneous offset rate. The instantaneous offset rate measures the offset rate of the power factor in an extremely short time, representing the change of the power factor with time in the form of a derivative. The calculation expression is as follows:

[0072]

[0073] , where ΔPF(t) is the instantaneous power factor offset rate, representing the offset speed of the instantaneous power factor PF(t) at time t, that is, the instantaneous power factor offset rate of the instantaneous power factor. is the derivative of the instantaneous power factor with respect to time, which is the result of taking the time derivative of the instantaneous power factor cos(E(t)), representing the offset rate of the power factor at a certain instant. is the calculation process of the instantaneous power factor offset, which is the derivative of the instantaneous power factor cos(E(t)), reflecting the sensitivity of the power factor to the phase angle difference.

[0074] To capture the overall intensity of the power factor offset, the instantaneous power factor offset rate ΔPF(t) is weighted and integrated, and a time weight function is introduced. The calculation expression is as follows:

[0075]

[0076] , where PFShiftStrength is the power factor offset intensity, t0 and t1 are the start time and end time of the detection window, and W(t) is the time weight function, representing the weight coefficient at time t, used to assign different importance to the changes in different time periods.

[0077] Next, define a non - linear offset discrimination function. Compare the offset degree of the power factor with its critical value to generate a function for measuring the degree of abnormal deviation. The function for measuring the degree of abnormal deviation maps the offset intensity to a non - linear scale to determine whether the deviation is abnormal. The calculation expression is as follows:

[0078]

[0079] , where \(f(PF)\) is the non - linear offset discrimination function, \(E\) ref is the reference phase angle, that is, the reference phase angle under normal conditions, \(\beta\) is the weight coefficient used to control the sensitivity of the phase - angle offset in the discrimination function, \(k\) is the non - linear amplification coefficient used to control the non - linear degree in the discrimination function, and \(\cos(E\) ref ) is the reference instantaneous power factor;

[0080] Based on the non - linear offset discrimination function \(f(PF)\), generate the power - factor offset index. The calculation expression is as follows: \(PF\Delta I=\ln(1 + \theta\cdot f(PF))\), where \(PF\Delta I\) is the power - factor offset index, and \(\theta\) is the amplification coefficient used to adjust the sensitivity and range of the index;

[0081] From the calculation expression of the power - factor offset index, it can be seen that under the detection window, the larger the value of the power - factor offset index generated after analyzing the deviation degree of the power factor in a short time, the more sudden accesses of a large number of inductive or capacitive loads (such as the startup of server clusters, cooling systems or emergency equipment) exist in the area. These loads will significantly increase the reactive - power demand, causing the power factor to quickly deviate from the normal level. This usually means that the power demand in the area is surging because these devices will greatly increase the load on the power grid when starting and running. On the contrary, if the value of the power - factor offset index is small or stable within the normal range, it indicates that the power consumption situation in the area is relatively stable, the power consumption is within the predicted range, and there is no significant load growth.

[0082] A rapid increase in the ratio of the instantaneous peak to the average value of power consumption within a short period usually indicates a surge in power consumption in the area. This phenomenon shows that there has been a drastic change in power consumption within a short time, with the instantaneous peak far higher than the normal average level, reflecting that multiple high-power-consuming devices may start or expand their operations simultaneously. When enterprises in the area respond to emergencies (such as cyberattacks, emergency data processing, or production tasks), they often quickly increase the operating loads of servers, refrigeration equipment, and other key devices, resulting in a sharp surge in instantaneous power demand. The centralized startup of these loads within a short time causes a sharp peak in the power consumption curve, while the average value fails to rise synchronously due to the relatively low initial load within the time window, widening the gap between the peak and the average value. Therefore, the sharp increase in the ratio indicates that the power demand in the area is breaking through the predicted range, predicting that the load surge may continue or even further expand. If the power system fails to respond to this change in a timely manner, it may lead to risks such as grid overload or equipment downtime.

[0083] Under the detection window, an instantaneous load fluctuation index is generated after analyzing the ratio of the instantaneous peak to the average value of power consumption within a short period. The specific steps are as follows:

[0084] Under the detection window, the sampled values of current and voltage are obtained in real time, and the obtained current sampled value is calibrated as I(t), and the voltage sampled value is calibrated as V(t). I(t) refers to the current sampled value at time t, and V(t) refers to the voltage sampled value at time t;

[0085] At each time point t, the difference in the instantaneous power change is calculated to reflect the sudden change trend of the device power. Here, the non-linear relationship between current and voltage is used to capture the instantaneous jump of power. The calculation expression is as follows: ΔP(t) = |I(t)·V(t) - I(t - Δt)·V(t - Δt)|, where ΔP(t) is the instantaneous change amount of power, representing the instantaneous change amount of power at time point t, Δt is the sampling time interval used to calculate the change between two adjacent time points, I(t - Δt) is the current sampled value at the previous time point of time t, and V(t - Δt) is the voltage sampled value at the previous time point of time t;

[0086] Calculating the difference in the instantaneous power change at each time point aims to capture various sudden changes in the device power. These changes reflect the switching of the device between different operating states and may be the result of events such as device startup, shutdown, and load transfer. These differences not only reveal the operating state of the device itself but also can reflect the volatility of the enterprise's power demand and whether the power grid is facing unstable load challenges. The following details the types of differences in the instantaneous power change and their corresponding power sudden change trends:

[0087] 1. Power jump difference: Device startup or expansion trend

[0088] Definition: When a certain device or server cluster suddenly starts up, there will be an obvious instantaneous jump in the product of current and voltage (i.e., power).

[0089] Reflected trends:

[0090] Server expansion: When a data center or high-tech enterprise encounters a cyber attack, it will quickly expand the server cluster, resulting in an instantaneous surge in power.

[0091] Emergency device startup: The emergency startup of backup generators, cooling equipment, data recovery systems, etc. in emergency situations will also cause an instantaneous power jump.

[0092] Explanation:

[0093] This kind of power jump usually means that the device is switching from a low-power mode to a high-load operating state, and timely power supply adjustment is required. Otherwise, it may cause local overload of the power grid.

[0094] 2. Power sudden drop difference: Equipment shutdown or load unloading trend

[0095] Definition: When a certain device is shut down or partially unloads its load, the power will suddenly drop in a short period of time.

[0096] Reflected trends:

[0097] Planned shutdown: When an enterprise conducts equipment maintenance or finishes high-intensity tasks, some high-power-consuming equipment stops running.

[0098] Load unloading: In order to cope with the load pressure of the power grid or internal optimization of the enterprise, the system may actively unload some loads.

[0099] Explanation:

[0100] [[ID=3L]]The sudden drop in power may cause unnecessary scheduling delays in the power grid. If the scheduling system cannot respond in time, it may lead to voltage fluctuations and frequency abnormalities.

[0101] 3. Fluctuation difference: Frequent start-stop or power instability trend

[0102] Definition: The frequent start-stop of equipment or the rapid switching of operating states causes the power to fluctuate violently in a short period of time.

[0103] Reflected trends:

[0104] Server fault recovery: When the system encounters an attack or fails, the server may restart frequently.

[0105] Frequent adjustment of the temperature control system: The frequent start-stop of air conditioners or cooling systems due to changes in ambient temperature will also cause frequent power fluctuations.

[0106] Explanation:

[0107] The emergence of fluctuations in differences means that the operating state of the system is unstable. Frequent fluctuations may accelerate equipment aging and even affect the frequency and voltage stability of the power grid.

[0108] 4. Short-term peak difference: Emergency load trend

[0109] Definition: The power reaches a peak within a short time and then quickly drops back to the normal level.

[0110] Reflected trend:

[0111] Emergency system testing: Backup power generation systems and cooling systems generate short-term peak power during testing.

[0112] Instantaneous high-power consumption tasks: Data centers may execute high-energy-consuming computing tasks within a short time, such as large-scale AI model training.

[0113] Explanation:

[0114] The short-term peak difference indicates that the system requires a large amount of power instantaneously. If this demand cannot be responded to in a timely manner, it may lead to system instability or load transfer failure.

[0115] 5. Continuous load change difference: Continuous expansion or contraction trend

[0116] Definition: The power change continuously increases or decreases over a period of time rather than instantaneously.

[0117] Reflected trend:

[0118] Data center expansion: Continuously expanding servers, the load is continuously increasing, and the power is also gradually increasing.

[0119] Gradual contraction of production tasks: Enterprises gradually end high-energy-consuming tasks, and the equipment gradually stops running, resulting in a slow decrease in power.

[0120] Explanation:

[0121] Continuous load changes usually mean that the system is gradually adjusting its operating mode, and the power system needs to respond flexibly to maintain stable power supply.

[0122] Differences in instantaneous power changes can reflect various mutation trends of equipment, including equipment startup, shutdown, frequent startup and shutdown, short-term peaks, and continuous expansion or contraction. These differences are crucial for real-time monitoring of the power system, helping to identify key nodes in enterprise operations and enabling timely response to different types of load changes through an intelligent power dispatching system to avoid risks of system instability or insufficient power supply.

[0123] To capture the cumulative fluctuation of the instantaneous change in power within the entire detection window, the cumulative absolute change function is adopted to avoid the mutual cancellation of positive and negative power fluctuations. The calculation expression is as follows:

[0124]

[0125] , where C P is the cumulative value of power jumps within the detection window, representing the total change in power fluctuations within the detection window. M is the total number of sampling points within the detection window, reflecting the number of data points collected within the detection window;

[0126] To accurately capture the dynamic change trend of the load, an exponential decay weight function is introduced, which can respond more sensitively to the latest power fluctuations and then generate an instantaneous load fluctuation index. The calculation expression is as follows:

[0127]

[0128] , where I-LFX is the instantaneous load fluctuation index, e is the natural base, β is the exponential decay coefficient of the instantaneous load fluctuation index, controlling the speed of exponential decay. M-t represents the distance between the sampling point and the current time. The closer to the current moment, the higher the weight of the instantaneous load fluctuation index;

[0129] From the calculation expression of the instantaneous load fluctuation index, it can be seen that under the detection window, the larger the value of the instantaneous load fluctuation index generated after analyzing the ratio of the instantaneous peak value to the average value of the power consumption within a short period of time, the more rapidly the power consumption increases within a short period of time. This usually reflects that multiple high-energy-consuming devices start simultaneously within a short period of time or the load increases rapidly, resulting in a surge trend in the regional power demand. On the contrary, if this ratio remains at a low level and changes smoothly, the value of the instantaneous load fluctuation index is also small, indicating that the regional power consumption is in a normal or stable state without abnormal fluctuations. Therefore, the larger the value of the instantaneous load fluctuation index, the more it can reflect the sudden change in power demand within the region.

[0130] The machine learning model is not limited here. Any machine learning model that can comprehensively analyze the power factor deviation index PFΔI and the instantaneous load fluctuation index I-LFX to generate the power consumption coefficient η-PCC can be used. To implement the technical solution of the present invention, the present invention provides a specific implementation method:

[0131] The calculation formula for generating the power consumption coefficient η-PCC is as follows:

[0132]

[0133] , where k1 and k2 are the preset proportionality coefficients of the power factor deviation index PFΔI and the instantaneous load fluctuation index I-LFX respectively, and both k1 and k2 are greater than 0.

[0134] As can be seen from the calculation expression of the power consumption coefficient, under the detection window, the larger the value of the power factor deviation index generated by analyzing the deviation degree of the power factor within a short period of time, and the larger the value of the instantaneous load fluctuation index generated by analyzing the ratio of the instantaneous peak value to the average value of the power consumption within a short period of time, it indicates that the value of the power consumption coefficient generated under the monitoring window is larger, indicating that the regional power consumption is in a surge state, and vice versa indicating that the regional power consumption is in a normal state;

[0135] Based on the evaluation results of the machine learning model, the regional power demand is divided into normal power demand and surge power demand;

[0136] Compare and analyze the power consumption coefficient generated by analyzing the obtained power consumption data under the monitoring window with the pre-set reference threshold of the power consumption coefficient, and divide the regional power demand. The division results are as follows:

[0137] If the power consumption coefficient is greater than or equal to the pre-set reference threshold of the power consumption coefficient, the regional power demand is divided into surge power demand;

[0138] If the power consumption coefficient is less than the pre-set reference threshold of the power consumption coefficient, the regional power demand is divided into normal power demand;

[0139] Normal power demand refers to the demand that conforms to the daily power consumption pattern within the prediction range; Surge power demand is the power demand that exceeds the prediction range and has abnormal growth rate and amplitude.

[0140] For normal power demand, supply power continuously and stably according to the established load forecasting plan, ensure the balance between power production and consumption, and maintain the normal operation of the power grid;

[0141] Supply power continuously and stably according to the established load forecasting plan. Ensure the balance between power production and consumption, and maintain the normal operation of the power grid. Regularly update the load forecasting model, adjust the power supply plan according to the actual situation, and improve the accuracy and reliability of power supply. At the same time, maintain communication with enterprises, understand the changes in their production plans and power consumption demands, and further optimize power supply services.

[0142] For surge power demand, adopt a distributed optimization algorithm, based on the established load forecasting plan, initiate intelligent power dispatching adjustment, and at the same time adjust the power supply distribution of the remaining regions to optimize the power grid load;

[0143] For surge power demand, adopt a distributed optimization algorithm, based on the established load forecasting plan, initiate intelligent power dispatching adjustment, and at the same time adjust the power supply distribution of the remaining regions to optimize the power grid load. The steps are as follows:

[0144] In the high-tech enterprise area, the power factor deviation index PFΔI and the instantaneous load fluctuation index I-LFX are obtained in real time through power monitoring equipment. Once a load surge is detected, it is necessary to determine whether to initiate scheduling, and the following judgment conditions are used for judgment;

[0145]

[0146] , where TriggerAdjustment is the trigger variable, which is used to determine whether power scheduling adjustment needs to be triggered. When this variable is equal to 1, the system will initiate power scheduling; when it is equal to 0, the system will not make adjustments. θ1 is the preset reference threshold of the power factor deviation index PFΔI, and θ2 is the preset reference threshold of the instantaneous load fluctuation index I-LFX;

[0147] After triggering the scheduling, solve the optimal power distribution among regions to meet the surge demand at the lowest cost. The optimization goal is to reduce the power supply cost, while reducing the load deviation and ensuring the stable operation of the power grid. The calculation expression is as follows:

[0148]

[0149] , where min is the minimization objective function, and the optimization algorithm is used to find the variable combination that makes the entire expression reach the minimum value. C i is the unit power supply cost of region i, and P i is the actual power supply quantity of region i, represents the total power supply cost of each region, λ is the penalty coefficient, which is used to balance the influence of power supply cost and prediction deviation on the optimization goal, is the predicted load of region i, is the penalty term for load prediction error, and the purpose is to reduce the deviation between the actual power supply quantity P i and the predicted power supply quantity ; N represents the total number of regions participating in scheduling in the power grid system;

[0150] This objective function balances the cost and load fluctuations, ensuring that the scheduling scheme is not only economical and reasonable, but also can reduce the impact of prediction deviation on the power grid.

[0151] In the distributed optimization process, it is necessary to ensure the balance between power supply and demand. At the same time, the load of each region must meet the limit conditions. Set the power balance constraint and the regional load constraint conditions. The power balance constraint expression is as follows:

[0152]

[0153] , where is the total power supply capacity in the system at the current moment;

[0154] The regional load constraint expression is as follows: In the formula, is the minimum load of region i, is the maximum load of region i;

[0155] Through the power balance constraint and regional load constraint conditions, it is ensured that there will be no power supply and demand imbalance or regional load overload problems during the dispatching process.

[0156] Since power dispatching involves multiple regions, the distributed Lagrangian relaxation method is used to decompose the global optimization problem into several sub-problems and solve them in parallel. The calculation expression is as follows:

[0157]

[0158] , in the formula, L(P i , μ) is the Lagrangian function, representing the objective function of power dispatching optimization, is the cost and prediction error part of the power grid system, calculating the power supply cost of all regions and their predicted load errors, is the power balance constraint term, ensuring that the sum of power supplies in all regions must be equal to the total power supply capacity of the system. μ is the Lagrange multiplier, used to impose constraints to ensure the power supply balance of the system;

[0159] Through distributed solution, each region independently calculates the load distribution and coordinates at the system level to achieve the optimization of the overall power grid.

[0160] After the implementation of the dispatching plan, feedback is carried out according to real-time data, and the power supply strategy is dynamically adjusted to ensure the stable operation of the power grid. The calculation expression is as follows: In the formula, is the adjusted power supply quantity of region i, which is the final power supply quantity allocated to this region after dynamic dispatching, ensuring that the power grid optimizes the power supply of each region in a timely manner according to the actual load change to meet the new load demand. α is the dispatching coefficient, used to control the adjustment strength of the power supply quantity of each region. This coefficient determines how each region shares the adjustment task when there is a deviation in the total load of the system, P t is the actual total load of the system at the current moment, which is the total power demand obtained through real-time monitoring, is the predicted total load at the current moment;

[0161] If (actual load is higher than the prediction), then is a positive value, indicating that power supply needs to be increased. The power supply quantity of each region will be appropriately increased according to the coefficient α.

[0162] If (actual load is higher than the prediction), then It is negative, indicating that power supply needs to be reduced. The power supply of each region will be appropriately decreased according to the coefficient α to avoid waste.

[0163] It is used to implement the dynamic feedback scheduling of the power system, and adjusts the power supply of each region in real time according to the difference between the actual load and the predicted load.

[0164] Through real-time power monitoring and feature extraction, and combined with the evaluation ability of the machine learning model, the present invention realizes the accurate monitoring and intelligent analysis of power consumption in the region. Especially by analyzing the power factor offset index and the instantaneous load fluctuation index, it can accurately judge whether there is a sharp increase in power demand in the region. This ability enables the power system to quickly identify sudden power demands caused by cyberattacks or server expansion and respond in a timely manner, avoiding serious consequences such as network paralysis and service interruption caused by insufficient power supply. The intelligent evaluation mechanism of the solution also ensures the dynamic adjustment of load forecasting, improves the flexibility of power management, enables the power grid to more accurately adapt to the rapid changes in enterprise needs, and maintains the stable development of the regional economy and industry.

[0165] The present invention formulates differential scheduling strategies for normal power demand and surging power demand. For normal demand, the power supply is maintained smoothly through the load forecasting plan, ensuring the balance between power production and consumption, and effectively avoiding power waste and systematic power shortage. In the case of surging demand scenarios, the solution uses a distributed optimization algorithm to achieve intelligent power scheduling adjustment. By optimizing the power supply distribution within the region and between adjacent regions, the load pressure on a single region is minimized, avoiding the risk of power outage caused by grid overload. This distributed intelligent scheduling not only improves the utilization efficiency of power resources, but also ensures the emergency power supply demand of enterprises when encountering sudden cyberattacks, provides stable power support for enterprises, and reduces operation risks and economic losses.

[0166] Only some exemplary embodiments of the present invention have been described by way of illustration. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A power grid energy management optimization platform based on big data, characterized by: The following steps are involved: First, deploy real-time power monitoring equipment in areas where high-tech industries are concentrated. This equipment will collect real-time power consumption data from enterprises and key power-consuming equipment. A data interface will be established to allow enterprises to independently upload power consumption data from their key internal equipment, improving the comprehensiveness and accuracy of the data. Extract features from the acquired power consumption data. Based on the extracted features, use a pre-learned machine learning model to intelligently evaluate the power consumption data in the area within the monitoring window. Based on the evaluation results of the machine learning model, regional power demand is divided into normal power demand and surge power demand; In response to normal power demand, we will continue to supply power stably according to the established load forecast plan, ensure the balance between power production and consumption, and maintain the normal operation of the power grid; To address surges in power demand, a distributed optimization algorithm is used to initiate intelligent power dispatch adjustments based on established load forecast plans, while also adjusting power distribution in the remaining areas to optimize grid load. The acquired power consumption data is subjected to feature extraction. The extracted features include the degree of deviation of the power factor in a short period of time and the ratio of the instantaneous peak value to the average value of the power consumption in a short period of time. Within the detection window, the degree of deviation of the power factor in a short period of time and the ratio of the instantaneous peak value to the average value of the power consumption in a short period of time are analyzed to generate a power factor offset index and an instantaneous load fluctuation index respectively. The power factor offset index and the instantaneous load fluctuation index are input into a pre-learned machine learning model, and a power consumption coefficient is generated by the machine learning model. The power consumption coefficient is used to perform an intelligent assessment of regional power consumption changes.

2. A big data-based power grid energy management optimization platform according to claim 1, characterized in that: In the detection window, the power factor deviation degree in a short period of time is analyzed to generate the power factor deviation index. The specific steps are as follows: In the detection window, the instantaneous power factor is calculated from the instantaneous waveforms of current and voltage. The calculation expression is as follows: PF(t) = cos(E(t)) = cos(φ V (t)-φ I (t)), where PF(t) is the instantaneous power factor, which represents the instantaneous power factor of the power system at time t, E(t) is the instantaneous phase angle difference, which represents the phase difference between voltage and current at time t, cos(E(t)) represents the instantaneous power factor, φ V (t) is the instantaneous phase angle of voltage, φ I (t) is the instantaneous phase angle of the current; Next, we define an indicator to quantify the power factor deviation in a short period of time, called the instantaneous deviation rate. The instantaneous deviation rate measures the rate of deviation of the power factor in a very short period of time. It expresses the change of the power factor over time in the form of a derivative. The calculation expression is as follows: Where ΔPF(t) is the instantaneous power factor deviation rate, which indicates the deviation speed of the instantaneous power factor PF(t) at time t, that is, the instantaneous power factor deviation rate of the instantaneous power factor. is the derivative of the instantaneous power factor with time, It is the calculation process of instantaneous power factor deviation; In order to capture the overall intensity of the power factor deviation, the instantaneous power factor deviation rate ΔPF(t) is weighted and integrated, and a time weight function is introduced. The calculation expression is as follows: Where PFShiftStrength is the power factor shift strength, t0 and t1 are the detection window start time and detection window end time, W(t) is the time weight function, Next, a nonlinear offset discriminant function is defined to compare the power factor offset with its critical value to generate a function that measures the degree of abnormal deviation. The calculation expression is as follows: Where F(pF) is the nonlinear offset discriminant function, E ref is the reference phase angle, that is, the reference phase angle under normal circumstances, β is the weight coefficient, k is the nonlinear amplification coefficient, cos(E ref ) is the reference instantaneous power factor; The power factor shift index is generated based on the nonlinear shift discriminant function f(PF). The calculation expression is as follows: PFΔI = ln(1 + θ·f(PF)), where PFΔI is the power factor shift index and θ is the amplification factor used to adjust the sensitivity and range of the index.

3. The big data-based power grid energy management optimization platform according to claim 1, characterized in that: In the detection window, the ratio of the instantaneous peak value to the average value of power consumption within a short period of time is analyzed to generate the instantaneous load fluctuation index. The specific steps are as follows: In the detection window, the current and voltage sampling values are obtained in real time, and the obtained current sampling value is calibrated as I(t), and the voltage sampling value is calibrated as V(t). I(t) refers to the current sampling value at time t, and V(t) refers to the voltage sampling value at time t. At each time point t, the difference in instantaneous power change is calculated to reflect the sudden change trend of the equipment power. The calculation expression is as follows: ΔP(t) = |I(t)·V(t)-I(t-Δt)·V(t-Δt)|, where ΔP(t) is the instantaneous change in power, indicating the instantaneous change in power at time point t, Δt is the sampling time interval used to calculate the change between two adjacent time points, I(t-Δt) is the current sampling value at the time point before time t, and V(t-Δt) is the voltage sampling value at the time point before time t. In order to capture the cumulative fluctuation of the instantaneous change of power within the entire detection window, the cumulative absolute change function is used to avoid the mutual cancellation of positive and negative power fluctuations. The calculation expression is as follows: Where C P is the cumulative value of power jumps within the detection window, indicating the total change in power fluctuation within the detection window. M is the total number of sampling points within the detection window, reflecting the number of data points collected within the detection window. In order to accurately capture the dynamic change trend of the load, an exponential decay weight function is introduced to respond more sensitively to the latest power fluctuations, thereby generating an instantaneous load fluctuation index. The calculation expression is as follows: Where I-LFX is the instantaneous load fluctuation index, e is the natural base, β is the instantaneous load fluctuation index attenuation coefficient, which controls the speed of exponential decay, and Mt represents the distance between the sampling point and the current time.

4. The big data-based power grid energy management optimization platform according to claim 1, characterized in that: The power consumption coefficient generated by analyzing the power consumption data obtained in the monitoring window is compared with the pre-set power consumption coefficient reference threshold to divide the regional power demand. The division results are as follows: If the power consumption coefficient is greater than or equal to a preset power consumption coefficient reference threshold, the regional power demand is classified as a surge power demand; If the power consumption coefficient is less than a preset power consumption coefficient reference threshold, the regional power demand is classified as normal power demand.

5. The big data-based power grid energy management optimization platform according to claim 4, characterized in that: To address surges in power demand, a distributed optimization algorithm is used to initiate intelligent power dispatch adjustments based on established load forecasting plans. This also adjusts power distribution in the remaining areas. The steps for optimizing grid load are as follows: In high-tech enterprise areas, power monitoring equipment is used to obtain the power factor deviation index PFΔI and the instantaneous load fluctuation index I-LFX in real time. Once a load surge is detected, it is necessary to determine whether to initiate scheduling. The following judgment conditions are used for judgment: Where TriggerAdjustment is the trigger variable, θ1 is the preset reference threshold of the power factor deviation index PFΔI, and θ2 is the preset reference threshold of the instantaneous load fluctuation index I-LFX. When dispatch is triggered, the optimal power distribution between regions is solved to meet the surge demand at the lowest cost. The calculation expression is as follows: In the formula, min is the minimization objective function, C i is the unit electricity supply cost in region i, P i is the actual power supply in region i, represents the total power supply cost of each region, λ is the penalty coefficient, is the forecast load of region i, It is the penalty term for load forecast error, the purpose of which is to reduce the actual power supply P i and predicted power supply The deviation between them, N represents the total number of regions participating in the dispatch in the power grid system; During the distributed optimization process, it is necessary to ensure a balance between power supply and demand. At the same time, the load in each area must meet the constraints. The power balance constraint and regional load constraint are set. The power balance constraint expression is as follows: Where, is the total power supply capacity in the system at the current moment; The regional load constraint expression is as follows: Where, is the area i is the least loaded, is the maximum load in region i; Since power dispatch involves multiple regions, the distributed Lagrangian relaxation method is used to decompose the global optimization problem into several sub-problems and solve them in parallel. The calculation expression is as follows: Where, L(P i ,μ) is the Lagrangian function, It is the cost and forecast error part of the power grid system, which calculates the power supply cost and its forecast load error in all areas. is the power balance constraint term, μ is the Lagrange multiplier; After the dispatching plan is implemented, feedback is provided based on real-time data, and the power supply strategy is dynamically adjusted to ensure the stable operation of the power grid. The calculation expression is as follows: Where, is the adjusted power supply of region i, α is the scheduling coefficient, which is used to control the adjustment strength of the power supply of each region, P t is the actual total load of the system at the current moment, is the forecast total load at the current moment.

Images