A control method and system for a pumped storage power station

Abstract

The present invention relates to the technical field of power station dispatching, and specifically discloses a control method and system for a pumped-storage power station, the method comprising obtaining historical data of a pumped-storage unit, real-time power of a pumped-storage unit and historical data of power grid load, establishing a working model of a pumped-storage unit; obtaining n power grid daily load curves from the historical data of power grid load, and then obtaining the predicted load data of the power grid daily load on that day; obtaining the predicted power of the pumped-storage unit based on the predicted load data of the power grid daily load on that day and the working model of the pumped-storage unit; and adjusting the working state of the pumped-storage unit. The present invention models and predicts the power of a pumped-storage unit based on historical data, then calculates the difference between the predicted power and the actual power, and then compares the difference with a preset threshold value, and decides whether to adjust the working state of the pumped-storage unit based on the comparison result, thereby realizing automatic and real-time adjustment without relying on the experience and operation of monitoring personnel.

Description

Technical Field

[0001] The present invention relates to the technical field of power station dispatching, and in particular to a control method and system for a pumped storage power station. Background Art

[0002] With the rapid development of modern industry and electric power, the contradiction between peak and valley load in the power system has become increasingly prominent. The power system urgently needs to have stronger peak-shaving, frequency regulation, phase regulation and standby capabilities to ensure the safe, stable and economical operation of the power system. Pumped-storage power stations, with their flexible, fast, economical and reliable characteristics, have become one of the effective and economical means to solve the above problems. Pumped-storage power stations pump water to upstream reservoirs when the power grid is at low load, convert electrical energy into potential energy and store it; when the power grid is at peak load, they release water to generate electricity and provide electrical energy to the power grid. Therefore, pumped-storage power stations can convert low-value electrical energy during the low load period of the power grid into high-value electrical energy during the peak load period.

[0003] However, currently monitoring personnel manually adjust the working status of the pumped storage unit based on real-time conditions, which not only relies on the experience of the monitoring personnel, but also requires the operation of the monitoring personnel to make adjustments, and cannot meet the requirements of automatic adjustment. Summary of the invention

[0004] The object of the present invention is to provide a control method and system for a pumped storage power station to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A control method for a pumped storage power station, the method comprising:

[0007] Obtain historical data of pumped storage units, real-time power of pumped storage units and historical data of power grid load, and establish a pumped storage unit working model based on the historical data of power grid load and historical data of pumped storage units;

[0008] Find n power grid daily load data of the same date as the weather on the current day in the power grid load history data and obtain n power grid daily load curves, and obtain the predicted load data of the power grid daily load on the current day according to the n power grid daily load curves;

[0009] The predicted power of the pumped storage unit is obtained based on the predicted load data of the daily load of the power grid and the working model of the pumped storage unit;

[0010] Adjust the working state of the pumped-storage unit according to the predicted power and real-time power of the pumped-storage unit.

[0011] As a further solution of the present invention: the step of establishing a pumped storage unit working model based on the historical data of the power grid load and the historical data of the pumped storage unit includes:

[0012] Randomly select some historical data of power grid load and some historical data of pumped storage units in the same weather as the training set, and the rest of the data as the test set;

[0013] The training set is substituted into the long short-term memory neural network algorithm, SOM clustering neural network algorithm and support vector machine algorithm respectively to obtain different initial pumped storage unit working models;

[0014] The test set is substituted into different initial pumped-storage unit working models for verification. After the verification passes, different pumped-storage unit working models are obtained.

[0015] As a further solution of the present invention: the step of obtaining the predicted power of the pumped storage unit based on the predicted load data of the daily load of the power grid and the working model of the pumped storage unit includes:

[0016] The working model of each pumped storage unit is trained according to the predicted load data of the daily load of the power grid on that day, and the predicted power of the working model of each pumped storage unit is obtained;

[0017] The predicted power of all pumped storage units' working models is weighted averaged to obtain the predicted power of the pumped storage units.

[0018] As a further solution of the present invention: before the training set is substituted into the long short-term memory neural network algorithm, the SOM clustering neural network algorithm and the support vector machine algorithm, the training set is also normalized. The specific formula of the normalization process is: Where d(i) is the i-th data in the training set, is the average value of different d(i), s is the variance of different d(i), and a(i) is the normalized result of d(i).

[0019] As a further solution of the present invention: the step of adjusting the working state of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit comprises:

[0020] Calculate the power change value of the pumped storage unit based on the predicted power of the pumped storage unit and the real-time power of the pumped storage unit;

[0021] The power change value is compared with the preset power change threshold. If the power change value is greater than the preset power change threshold, the working state of the pumped storage unit is adjusted in real time.

[0022] The technical solution of the present invention also provides a control system for a pumped storage power station, the system comprising:

[0023] An acquisition module is used to acquire historical data of the pumped-storage unit, real-time power of the pumped-storage unit and historical data of power grid load, and establish a pumped-storage unit working model based on the historical data of power grid load and historical data of the pumped-storage unit;

[0024] The prediction module is used to find n pieces of grid daily load data on the same date as the weather on the current day in the grid load history data and obtain n grid daily load curves, and obtain the predicted load data of the grid daily load on the current day according to the n grid daily load curves;

[0025] A generation module is used to obtain the predicted power of the pumped storage unit based on the predicted load data of the daily load of the power grid and the working model of the pumped storage unit;

[0026] The adjustment module is used to adjust the working state of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit.

[0027] As a further solution of the present invention: the acquisition module includes:

[0028] An acquisition unit, used to acquire historical data of the pumped storage unit, real-time power of the pumped storage unit and historical data of power grid load;

[0029] A partitioning unit is used to randomly select part of the historical data of power grid load and part of the historical data of pumped storage units in the same weather as a training set, and the rest of the data as a test set;

[0030] A model generation unit is used to substitute the training set into the long short-term memory neural network algorithm, the SOM clustering neural network algorithm and the support vector machine algorithm respectively to obtain different initial pumped storage unit working models;

[0031] The model verification unit is used to substitute the test set into different initial pumped storage unit working models for verification. If the verification passes, different pumped storage unit working models are obtained.

[0032] As a further solution of the present invention: the generation module includes:

[0033] A generating unit, used for training the working model of each pumped storage unit according to the predicted load data of the daily load of the power grid on that day, and obtaining the predicted power of the working model of each pumped storage unit;

[0034] The weighting unit is used to perform weighted averaging on the predicted powers of all pumped-storage unit working models to obtain the predicted power of the pumped-storage unit.

[0035] As a further solution of the present invention: the adjustment module includes:

[0036] A calculation unit, used for calculating a power change value of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit;

[0037] The adjustment unit is used to compare the power change value with a preset power change threshold value, and if the power change value is greater than the preset power change threshold value, adjust the working state of the pumped storage unit in real time.

[0038] Compared with the prior art, the beneficial effects of the present invention are: the present invention models and predicts the power of the pumped-storage unit based on historical data, then calculates the difference between the predicted power and the actual power, and then compares the difference with a preset threshold value, and decides whether to adjust the working state of the pumped-storage unit based on the comparison result, thereby realizing automatic and real-time adjustment without relying on the experience and operation of the monitoring personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0040] Figure 1 The figure is a flow chart of the control method of the pumped storage power station.

[0041] Figure 2 This is a partial first sub-flow chart of the control method of a pumped storage power station.

[0042] Figure 3 This is a block diagram of the third sub-process of the control method for a pumped storage power station.

[0043] Figure 4 This is a fourth sub-flow chart of the control method of the pumped storage power station.

[0044] Figure 5 This is a structural block diagram of the control system of a pumped storage power station.

[0045] Figure 6 The structural block diagram of the modules is obtained in the control system of the pumped storage power station.

[0046] Figure 7 Generate a structural block diagram of the components in the control system of a pumped storage power station.

[0047] Figure 8 This is a structural block diagram of the adjustment module in the control system of a pumped storage power station. DETAILED DESCRIPTION

[0048] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] Embodiment 1: Figure 1 : is a flow chart of a control method of a pumped storage power station. In an embodiment of the present invention, a control method of a pumped storage power station includes:

[0050] Obtain historical data of pumped storage units, real-time power of pumped storage units and historical data of power grid load, and establish a pumped storage unit working model based on the historical data of power grid load and historical data of pumped storage units;

[0051] Pumped storage units generally have three working modes: pumping mode, power generation mode and idle mode. The load of the power grid at a certain moment is closely related to the power of the pumped storage unit at that moment. When the power grid is in an overload state (that is, the power generated by the power station is much lower than the load power of the power grid), the pumped storage unit is in the power generation mode, releasing water to the hydropower station in the lower reservoir for power generation to reduce the overload state of the power grid; when the power grid is in a normal state (that is, the power generated by the power station is approximately the same as the load power of the power grid), the pumped storage unit is in the idle mode and the pumped storage unit does not work; when the power grid is in a low load state (that is, the power generated by the power station is much greater than the load power of the power grid), the pumped storage unit is in the pumping mode, using the excess electricity to pump water to the upper reservoir for storage, which is convenient for subsequent reuse. The real-time power of the pumped storage unit is obtained in order to determine whether it needs to be adjusted later. Knowing the load data of the power grid and the pumped storage unit data at that time, the correlation between the two can be found, thereby constructing a working model of the pumped storage unit.

[0052] Find n power grid daily load data of the same date as the weather on the current day in the power grid load history data and obtain n power grid daily load curves, and obtain the predicted load data of the power grid daily load on the current day according to the n power grid daily load curves;

[0053] When the weather is the same, the impact of wind on pumped storage units, the impact of temperature on pumped storage units, and the impact of humidity on pumped storage units are generally similar, and the grid load is also similar at this time. The predicted load curve of the grid daily load on that day will be more accurate, and the predicted load data of the grid daily load on that day can be obtained. The predicted load curve of the grid daily load on that day is r(k), and the n grid daily load curves are h1(k), h2(k), ... hn(k) (k = 1, 2, ..., n), then bi is the fitting parameter.

[0054] The predicted power of the pumped storage unit is obtained based on the predicted load data of the daily load of the power grid and the working model of the pumped storage unit;

[0055] The predicted load data of the daily load of the power grid is input into the working model of the pumped storage unit for iteration, and the output result is the predicted power of the pumped storage unit.

[0056] Adjust the working state of the pumped-storage unit according to the predicted power and real-time power of the pumped-storage unit.

[0057] The predicted working state of the pumped-storage unit is compared with the actual working state of the pumped-storage unit, and a decision is made whether to adjust the working state of the pumped-storage unit online based on the comparison result.

[0058] Figure 2 This is a partial first sub-flow chart of the control method of a pumped storage power station, wherein the step of establishing a pumped storage unit working model based on the historical data of the power grid load and the historical data of the pumped storage unit comprises:

[0059] Randomly select some historical data of power grid load and some historical data of pumped storage units in the same weather as the training set, and the rest of the data as the test set;

[0060] The historical data of power grid load under the same weather conditions will be similar to the power grid load of the day. This part of the historical data of power grid load is used as the total sample set, and a part of the total sample set is randomly selected as the training set, so that the established model is fully representative of the samples in the total sample set. The test set is to verify whether the established model is accurate. Normalizing the training set can facilitate the subsequent data processing and speed up the network learning speed. The specific formula for normalization is: Where d(i) is the i-th data in the training set, is the average value of different d(i), s is the variance of different d(i), and a(i) is the normalized result of d(i).

[0061] The training set is substituted into the long short-term memory neural network algorithm, SOM clustering neural network algorithm and support vector machine algorithm respectively to obtain different initial pumped storage unit working models;

[0062] In order to ensure that the established model is suitable for different working situations, three algorithms are used for modeling. The core of the long short-term memory neural network algorithm is its neuron state. The neuron state is like a conveyor belt, which has a small linear effect and runs through the entire chain structure. Information can easily propagate on the conveyor belt, but the state does not change. The long short-term memory neural network algorithm has the ability to delete or add information in the neuron state. This mechanism is carefully managed by a structure called a threshold. The threshold is a way to selectively pass information. They are made of Sigmoid neural network layers and point-by-point multipliers. The Sigmoid neural network layer outputs a number between 0 and 1, describing how much information a neuron should pass. Outputting "0" means "nothing can pass", and outputting "1" means "let everything pass". The first step of the long short-term memory neural network algorithm is to decide what information should be forgotten by the neuron. This is composed of a Sigmod neural network layer called the "forget gate layer". It inputs ht-1 and xt, and then outputs a number between 0 and 1 in each neuron state of ct-1. "1" means "keep this completely" and "0" means "forget this completely". A language model that tries to predict the next word based on previous words. In this problem, the neuron state may include information about the gender of the current subject, so that the correct pronoun can be used. When we see a new subject, we forget the previous gender information, ft = δ (W f [h t-1 ,x t ]+b f ). The next step is to decide what information we want to store in the neuron, which consists of two parts. First, a sigmoid neural network layer called the "forget gate layer" determines the value we want to update. Then, a tanh layer generates a new candidate value Cts, which is added to the neuron state. In the next step, we will combine these two steps to generate an updated state value, adding the new subject's gender to the neuron state, replacing the old subject that we are about to forget, i t =δ(W i ·[h t-1 ,x t ]+b i ), Cts = tanh(W c ·[h t-1 ,x t ]+b c ). Update the old neuron state ct-1 to the new neuron state Ct. The previous step has decided what to do, and we will do it in the next step. We multiply the old state by ft, forgetting the information we decided to forget before, and then we increase it * Cts. This is the new candidate value, which is measured by how much we want to update the value of each state, Ct =f t ·C t-1 +i t ·C ts Finally, we have to decide what to output. This output is based on our neuron state, but with a filter. First, we use the Sigmund neural network layer to decide which part of the neuron state needs to be output; then we let the neuron state pass through the tanh layer (to make the output value between -1 and 1) and multiply it by the output of the Sigmund threshold. We only output what we want to output, O t =δ(W o [h t-1 ,x t ]+b o ), h t =o t tanh(C t ).

[0063] The SOM clustering neural network is an unsupervised learning neural network that maps input data to a two-dimensional topological structure through competitive learning and proximity updating, thereby achieving data clustering. The SOM clustering neural network consists of a two-dimensional structure, usually a grid-like structure, and each node is called a neuron. These neurons are arranged in a two-dimensional topological structure, and each neuron has an associated weight vector, which has the same dimension as the input data. The SOM clustering neural network is trained using competitive learning. For a given input sample, each neuron in the SOM clustering neural network calculates the distance between its weight vector and the input sample. The competition mechanism in the SOM clustering neural network is implemented based on competition between neurons. Given an input vector x, the activation value ai of neuron i can be calculated by the following formula: , wij is the weight between neuron i and input vector x, and xj is the jth component of the input vector. In order to ensure the relative importance of data in the SOM clustering neural network, the SOM clustering neural network will cluster according to the similarity and feature self-organization of the data, and classify similar fluctuation processes into one category.

[0064] The support vector machine algorithm is a two-class classification model that finds a hyperplane to divide data into one class and other classes, with the largest separation interval and is different from the perceptron. This algorithm is a commonly used algorithm and will not be described in detail in this invention. These three algorithms obtain three different initial pumped storage unit working models to prepare for the follow-up.

[0065] The test set is substituted into different initial pumped-storage unit working models for verification. After the verification passes, different pumped-storage unit working models are obtained.

[0066] The test set is used to verify whether the initial pumped-storage unit working model is accurate and to ensure the accuracy of the next step of predicted power. The verification standard is to see whether the number of iterations or the performance reaches the preset threshold. If the verification fails, the training set and the test set are randomly divided again, and the initial pumped-storage unit working model is rebuilt. If the verification passes, the initial pumped-storage unit working model can be used as the pumped-storage unit working model.

[0067] Figure 3 The third sub-flow chart of the control method of the pumped storage power station, wherein the step of obtaining the predicted power of the pumped storage unit according to the predicted load data of the daily load of the power grid and the working model of the pumped storage unit comprises:

[0068] The working model of each pumped storage unit is trained according to the predicted load data of the daily load of the power grid on that day, and the predicted power of the working model of each pumped storage unit is obtained;

[0069] By substituting the predicted load data of the day into the pumped storage unit working model obtained by the long short-term memory neural network algorithm, the pumped storage unit working model obtained by the SOM clustering neural network algorithm and the pumped storage unit working model obtained by the support vector machine algorithm, three different pumped storage unit predicted powers can be obtained.

[0070] The predicted power of all pumped storage units' working models is weighted averaged to obtain the predicted power of the pumped storage units.

[0071] Since each algorithm has different applicable situations and advantages and disadvantages, the weight scores of the pumped-storage unit working model obtained by the long short-term memory neural network algorithm, the pumped-storage unit working model obtained by the SOM clustering neural network algorithm and the pumped-storage unit working model obtained by the support vector machine algorithm are preset. The weight score of each pumped-storage unit working model divided by the sum of the weight scores of all pumped-storage unit working models is the weight coefficient of the pumped-storage unit working model. The predicted power obtained by each pumped-storage unit working model is multiplied by the weight coefficient of the pumped-storage unit working model, and these results are added together to obtain the final predicted power of the pumped-storage unit.

[0072] Figure 4 The fourth sub-flow chart of the control method of the pumped storage power station, wherein the step of adjusting the working state of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit comprises:

[0073] Calculate the power change value of the pumped storage unit based on the predicted power of the pumped storage unit and the real-time power of the pumped storage unit;

[0074] The predicted power of the pumped-storage unit is an idealized state. We also need to look at the actual real-time power of the pumped-storage unit. The difference between the two is the power change value of the pumped-storage unit.

[0075] The power change value is compared with the preset power change threshold. If the power change value is greater than the preset power change threshold, the working state of the pumped storage unit is adjusted in real time.

[0076] It costs a certain amount of money to adjust the power of the pumped-storage unit. Different pumped-storage power stations usually calculate the cost of adjusting the power of the pumped-storage unit, so as to obtain the critical value of the power change. This value is defined as the preset power change threshold. When the power change value is less than or equal to the preset power change threshold, if the working state of the pumped-storage unit is adjusted, the cost and the effect obtained are less cost-effective. At this time, the difference between the two is regarded as a reasonable difference range, and there is no need to adjust the working state of the pumped-storage unit. Only when the power change value is greater than the preset power change threshold, the difference between the two is large, and the working state of the pumped-storage unit must be adjusted, otherwise more economic costs will be incurred.

[0077] Embodiment 2: Figure 5 The structure diagram of the control system of a pumped storage power station is shown in FIG. 1 . In an embodiment of the present invention, a control system of a pumped storage power station includes:

[0078] An acquisition module is used to acquire historical data of the pumped-storage unit, real-time power of the pumped-storage unit and historical data of power grid load, and establish a pumped-storage unit working model based on the historical data of power grid load and historical data of the pumped-storage unit;

[0079] The prediction module is used to find n pieces of grid daily load data on the same date as the weather on the current day in the grid load history data and obtain n grid daily load curves, and obtain the predicted load data of the grid daily load on the current day according to the n grid daily load curves;

[0080] A generation module is used to obtain the predicted power of the pumped storage unit based on the predicted load data of the daily load of the power grid and the working model of the pumped storage unit;

[0081] The adjustment module is used to adjust the working state of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit.

[0082] The acquisition module comprises:

[0083] An acquisition unit, used to acquire historical data of the pumped storage unit, real-time power of the pumped storage unit and historical data of power grid load;

[0084] A partitioning unit is used to randomly select part of the historical data of power grid load and part of the historical data of pumped storage units in the same weather as a training set, and the rest of the data as a test set;

[0085] A model generation unit is used to substitute the training set into the long short-term memory neural network algorithm, the SOM clustering neural network algorithm and the support vector machine algorithm respectively to obtain different initial pumped storage unit working models;

[0086] The model verification unit is used to substitute the test set into different initial pumped storage unit working models for verification. If the verification passes, different pumped storage unit working models are obtained.

[0087] The generation module comprises:

[0088] A generating unit, used for training the working model of each pumped storage unit according to the predicted load data of the daily load of the power grid on that day, and obtaining the predicted power of the working model of each pumped storage unit;

[0089] The weighting unit is used to perform weighted averaging on the predicted powers of all pumped-storage unit working models to obtain the predicted power of the pumped-storage unit.

[0090] The adjustment module comprises:

[0091] A calculation unit, used for calculating a power change value of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit;

[0092] The adjustment unit is used to compare the power change value with a preset power change threshold value, and if the power change value is greater than the preset power change threshold value, adjust the working state of the pumped storage unit in real time.

[0093] The functions that can be achieved by the control method of the pumped-storage power station are all completed by a computer device, which includes one or more processors and one or more memories, and at least one program code is stored in the one or more memories. The program code is loaded and executed by the one or more processors to achieve the functions of the user behavior prediction method based on big data.

[0094] The processor takes out instructions from the memory one by one, analyzes the instructions, and then completes the corresponding operations according to the instruction requirements, generating a series of control commands, so that the various parts of the computer can automatically, continuously and coordinately move to become an organic whole, realize the input of programs, the input of data, and the calculation and output of results. The arithmetic operations or logical operations generated in this process are all completed by the operator; the memory includes a read-only memory (ROM), which is used to store computer programs, and a protection device is provided outside the memory.

[0095] Exemplarily, the computer program may be divided into one or more modules, one or more modules are stored in a memory and executed by a processor to implement the present invention. One or more modules may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in a terminal device.

[0096] Those skilled in the art will understand that the description of the above service equipment is merely an example and does not constitute a limitation on the terminal equipment. It may include more or fewer components than described above, or a combination of certain components, or different components, for example, it may include input and output devices, network access equipment, buses, etc.

[0097] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect the various parts of the entire user terminal.

[0098] The memory can be used to store computer programs and / or modules. The processor can realize various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as an information collection template display function, a product information release function, etc.); the data storage area can store data created according to the use of the berth status display system (such as product information collection templates corresponding to different product types, product information that different product providers need to release, etc.). In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0099] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the modules / units in the above-mentioned embodiment system, and can also be completed by instructing the relevant hardware through a computer program. The above-mentioned computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can realize the functions of the above-mentioned various system embodiments. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form. Computer-readable media may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0100] It should be noted that, in this article, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0101] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A control method for a pumped storage power station, characterized in that: The method comprises: Obtain historical data of pumped storage units, real-time power of pumped storage units and historical data of power grid load, and establish a pumped storage unit working model based on the historical data of power grid load and historical data of pumped storage units; In the historical data of power grid load, find n power grid daily load data of the same date as the weather of the day and obtain n power grid daily load curves. According to the n power grid daily load curves, obtain the predicted load data of the power grid daily load of the day. The predicted load curve of the power grid daily load of the day is r(k). The n power grid daily load curves are h1(k), h2(k), ... hn(k) (k = 1, 2, ..., n). Then bi is the fitting parameter; The predicted power of the pumped storage unit is obtained based on the predicted load data of the daily load of the power grid and the working model of the pumped storage unit; According to the predicted power of the pumped-storage unit and the real-time power of the pumped-storage unit, the working state of the pumped-storage unit is adjusted. The step of establishing the working model of the pumped-storage unit according to the historical data of the power grid load and the historical data of the pumped-storage unit comprises: Randomly select some historical data of power grid load and some historical data of pumped storage units in the same weather as the training set, and the rest of the data as the test set; The training set is normalized. The specific formula for normalization is: Where d(i) is the i-th data in the training set, is the average value of different d(i), s is the variance of different d(i), a(i) is the normalized result of d(i), the training set is substituted into the long short-term memory neural network algorithm, SOM clustering neural network algorithm and support vector machine algorithm respectively, and different initial pumped storage unit working models are obtained. The long short-term memory neural network algorithm is composed of a Sigmod neural network layer called the "forget gate layer", which inputs ht-1 and xt, and then outputs a number between 0 and 1 in each neuron state of ct-1, "1" means "totally keep this", "0" means "totally forget this", the tanh layer generates a new candidate value Cts, which will be added to the neuron state, and these two steps are combined to generate an updated state value, adding the gender of the new subject to the neuron state to replace the old subject that we are going to forget, i t =δ(W i ·[h t-1 ,x t ]+b i ), Cts = tanh(W c ·[h t-1 ,x t ]+b c ), update the old neuron state ct-1 to the new neuron state Ct, multiply the old state by ft, forget the information we decided to forget before, increase it*Cts, C t =f t ·C t-1 +i t ·C ts , use the Sigmod neural network layer to determine which part of the neuron state needs to be output; then let the neuron state pass through the tanh layer and multiply it by the output of the Sigmod threshold, O t =δ(W o [h t-1 ,x t ]+b o ), h t =o t tanh(C t ), each node of the SOM clustering neural network is called a neuron, and these neurons are arranged in a two-dimensional topological structure. Each neuron has a weight vector associated with it, and the dimension of this vector is the same as the dimension of the input data. Given an input vector x, the activation value ai of neuron i can be calculated by the following formula: wij is the weight between neuron i and input vector x, and xj is the jth component of the input vector; Substituting the test set into different initial pumped storage unit working models for verification, and obtaining different pumped storage unit working models after verification, the step of obtaining the predicted power of the pumped storage unit based on the predicted load data of the daily load of the power grid and the pumped storage unit working model comprises: The working model of each pumped storage unit is trained according to the predicted load data of the daily load of the power grid on that day, and the predicted power of the working model of each pumped storage unit is obtained; The predicted powers of all pumped-storage unit working models are weighted averaged to obtain the predicted power of the pumped-storage unit. The weight scores of the pumped-storage unit working models obtained by the long-short memory neural network algorithm, the pumped-storage unit working models obtained by the SOM clustering neural network algorithm and the pumped-storage unit working models obtained by the support vector machine algorithm are preset. The weight score of each pumped-storage unit working model divided by the sum of the weight scores of all pumped-storage unit working models is the weight coefficient of the pumped-storage unit working model. The predicted power obtained by each pumped-storage unit working model is multiplied by the weight coefficient of the pumped-storage unit working model, and these results are added together to obtain the final predicted power of the pumped-storage unit.

2. The control method of a pumped storage power station according to claim 1, characterized in that: The step of adjusting the working state of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit comprises: Calculate the power change value of the pumped storage unit based on the predicted power of the pumped storage unit and the real-time power of the pumped storage unit; The power change value is compared with the preset power change threshold. If the power change value is greater than the preset power change threshold, the working state of the pumped storage unit is adjusted in real time.

3. A control system for a pumped storage power station, characterized in that: The system comprises: An acquisition module is used to acquire historical data of the pumped-storage unit, real-time power of the pumped-storage unit and historical data of power grid load, and establish a pumped-storage unit working model based on the historical data of power grid load and historical data of the pumped-storage unit; The prediction module is used to find n grid daily load data of the same date as the weather on the current day in the grid load history data and obtain n grid daily load curves, and obtain the predicted load data of the grid daily load on the current day according to the n grid daily load curves. The predicted load curve of the grid daily load on the current day is r(k), and the n grid daily load curves are h1(k), h2(k), ... hn(k) (k = 1, 2, ..., n), then bi is the fitting parameter; A generation module is used to obtain the predicted power of the pumped storage unit based on the predicted load data of the daily load of the power grid and the working model of the pumped storage unit; The adjustment module is used to adjust the working state of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit. The acquisition module includes: An acquisition unit, used to acquire historical data of the pumped storage unit, real-time power of the pumped storage unit and historical data of power grid load; A partitioning unit is used to randomly select part of the historical data of power grid load and part of the historical data of pumped storage units in the same weather as a training set, and the rest of the data as a test set; The model generation unit normalizes the training set. The specific formula for normalization is: Where d(i) is the i-th data in the training set, is the average value of different d(i), s is the variance of different d(i), a(i) is the normalized result of d(i), the training set is substituted into the long short-term memory neural network algorithm, SOM clustering neural network algorithm and support vector machine algorithm respectively, and different initial pumped storage unit working models are obtained. The long short-term memory neural network algorithm is composed of a Sigmod neural network layer called the "forget gate layer", which inputs ht-1 and xt, and then outputs a number between 0 and 1 in each neuron state of ct-1, "1" means "totally keep this", "0" means "totally forget this", the tanh layer generates a new candidate value Cts, which will be added to the neuron state, and these two steps are combined to generate an updated state value, adding the gender of the new subject to the neuron state to replace the old subject that we are going to forget, i t =δ(W i ·[h t-1 ,x t ]+b i ), Cts = tanh(W c ·[h t-1 ,x t ]+b c ), update the old neuron state ct-1 to the new neuron state Ct, multiply the old state by ft, forget the information we decided to forget before, increase it * Cts, C t =f t ·C t-1 +i t ·C ts , use the Sigmod neural network layer to determine which part of the neuron state needs to be output; then let the neuron state pass through the tanh layer and multiply it by the output of the Sigmod threshold, O t =δ(W o [h t-1 ,x t ]+b o ), h t =o t tanh(C t ), each node of the SOM clustering neural network is called a neuron, and these neurons are arranged in a two-dimensional topological structure. Each neuron has a weight vector associated with it, and the dimension of this vector is the same as the dimension of the input data. Given an input vector x, the activation value ai of neuron i can be calculated by the following formula: wij is the weight between neuron i and input vector x, and xj is the jth component of the input vector; The model verification unit is used to substitute the test set into different initial pumped storage unit working models for verification. After the verification is passed, different pumped storage unit working models are obtained. The generation module includes: A generating unit, used for training the working model of each pumped storage unit according to the predicted load data of the daily load of the power grid on that day, and obtaining the predicted power of the working model of each pumped storage unit; The weighting unit is used to perform weighted averaging on the predicted powers of all pumped-storage unit working models to obtain the predicted power of the pumped-storage unit, and to preset the weight scores of the pumped-storage unit working model obtained by the long-short memory neural network algorithm, the pumped-storage unit working model obtained by the SOM clustering neural network algorithm, and the pumped-storage unit working model obtained by the support vector machine algorithm. The weight score of each pumped-storage unit working model divided by the sum of the weight scores of all pumped-storage unit working models is the weight coefficient of the pumped-storage unit working model. The predicted power obtained by each pumped-storage unit working model is multiplied by the weight coefficient of the pumped-storage unit working model, and these results are added together to obtain the final predicted power of the pumped-storage unit.

4. The control system of the pumped storage power station according to claim 3, characterized in that: The adjustment module comprises: A calculation unit, used for calculating a power change value of the pumped storage unit according to the predicted power of the pumped storage unit and the real-time power of the pumped storage unit; The adjustment unit is used to compare the power change value with a preset power change threshold value, and if the power change value is greater than the preset power change threshold value, adjust the working state of the pumped storage unit in real time.

Images